Reminds me of the old joke "90% of the code is 90% of the work. The last 10% of the code is the other 90% of the work."
I have spent almost my entire adult life (since 1986) shipping products. One of the very first things that I learned, was that "shipping" > "designing".
There's so much work in delivering products that will carry your brand, and then must be supported.
I liken it to having children. Conceiving them is fun. Delivering them is painful. Raising them, is a lifetime of work.
In my experience, the same type of thing applies to products that we ship (and charge money for).
Beautifully said. Engineering is often just a cost center and I often have the feeling management is suspicious that the engineers are just wasting time and throwing up roadblocks for nothing. This in turn makes managers always on the lookout for "the shortcut" to cut out as many of those engineers as possible.
There is a definite lack of appreciation for the often repetitive grit, toil and maintenance work required to just have profit generating working software running reliably in production.
> There's so much work in delivering products that will carry your brand, and then must be supported.
People think otherwise with AI partly because Anthropic kept telling us that they didn't have to write code or review code any more for most of their work. Their agent swarms just comb through their github, slack and wikis to figure out what to do next, and another swarm of agents just review, test, merge, deploy, A/B test, and revert the code. Boris alone merged nearly 300 PRs in the past week (or two?). So the top research labs seem have broken the productivity seal.
And then they talk about this recursively self-improving AI that is so powerful, so autonomous that they advocate that every company should be prepared to "pause" the effort. And their Fable/Mythos has this specific restriction as mentioned in their model card[1] that they are going to reject requests to tune and train models because, well you guess it, the models are too powerful to be used by mere mortals.
[1] We’ve implemented new interventions that limit Claude’s effectiveness for requests targeting frontier LLM development (for example, on building pretraining pipelines, distributed training infrastructure, or ML accelerator design). Using Claude to develop competing models already violates our Terms of Service, but enforcing this restriction through our safeguards avoids accelerating the actors most willing to violate these terms. Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts, these safeguards will not be visible to the user. Fable 5 will not fall back to a different model. Instead, the safeguards will limit effectiveness through methods such as prompt modification, steering vectors, or parameter-efficient fine-tuning (PEFT).
I think taking Anthropic or any company in this space at face value is naive at best though. AGI has been 6 months away for years now. Surely anyone can think this through: Anthropic knows what theyre doing with their public facing repositories, they know to make things enabled by their tech seem impressive. I would consider Bun etc. examples of this.
People think otherwise with AI partly because Anthropic kept telling us that they didn't have to write code or review code any more for most of their work.
Even if that were 100% true, it only collapses the coding effort to near zero. Anyone who's built and shipped a real product should know that coding is maybe 50% of the work, and on a mature product it can be much less.
Boris also says he stops using /plan, he writes loop to write prompt, and he simply asks AI to come up with solutions. He also said many times that his agents would comb their emails, slack channels, and Github issues to come up with things to do. When we combine what he has said, it's hard not to have the impression that he was implying full autonomy of their agents. The only that the engineers need to do is to build harness and to issue approvals, rejections, or suggestions.
I work on a toy project that has exactly one user (me). On its face it's fairly simple. It's a portal to my media server because I didn't like how Plex worked with regards to searching and filtering. I can look for movies or series by director, studio, publisher, etc. I can rate things, I can find highly rated things. It's great, and instead of bugging plex support to add new features, I just tell Deepseek to do it. I started it before LLms were prevalent and now that I have open code I've had Deepseek write and rewrite most of my code and implement new features.
But even with this toy project, and the target market being someone I should know very well (me), I often struggle to figure out what I want the app to do. When I go through brainstorming or grilling sessions it'll often ask me a question about how the product ought to work and I'm just like ¯\_(ツ)_/¯ give me suggestions and I'll let you know.
Genuine creativity is something LLMs struggle with and it kind of makes sense given their design. If you have a complete plan for a feature or even just an idea what the feature should do, that is enough for an LLM. But asking it to think and come up with a new feature idea by itself will always yield mostly extereme basic things you've already thought of. That creativity of "what" to build so it serves a purpose is still very difficult imo and LLMs are not good at it.
> People think otherwise with AI partly because Anthropic kept telling us that they didn't have to write code or review code any more for most of their work. Their agent swarms just comb through their github, slack and wikis to figure out what to do next, and another swarm of agents just review, test, merge, deploy, A/B test, and revert the code. Boris alone merged nearly 300 PRs in the past week (or two?).
Apart from many other issues with this, heavily subsidized subscription plans won't last forever, and if you start burning your own money on tokens in this way, you'll soon realize it's terribly inefficient.
I’ve been wondering if “you’re not google” when learning about googles software dev process applies to Anthropic. Anthropic is a company that A. Has cheap unlimited access to its models and B. Is probably largely insulated from the types of tradeoffs that the rest of industry has had to observe in the post-ZIRP era.
Like did they break through the productivity seal? Or are they willing to spend that much more on it since they see their failure as a like existential threat to humanity. I doubt it our boss sees your software the same way.
It doesn't need to be an existential threat to humanity - it's an existential threat to their business. They need agentic workflows to work for their business to become profitable. So pouring money into the "no engineers write code anymore, only agents" model is at once R&D, QA, product development, and advertising. They can spend as much of their investors' money on this as they have to because if they can't (sustainably) sell this vision to other companies, their company collapses.
Zero interest rate policy. When interest rates are Near zero you can spend money like it’s free. A lot of what we thought of as like normal engineering culture were the result of interest rates being zero.
> the safeguards will limit effectiveness through methods such as prompt modification, steering vectors, or parameter-efficient fine-tuning (PEFT)
Holy crap that is dark. I like learning about ML for fun, and now I have to assume that their model is intentionally misinforming me to sabotage my learning? It is absolutely bananas that somebody decided that was ok behavior.
I don’t see how that helps, unless you actually mean open source, rather than open weights like most people do. Without everything that goes into the model, including training data, these things are opaque.
That may very well be the case. In fact, I'm nearly certain that you're right. But it doesn't change the fact that open weight models are altogether insufficient on a number of important dimensions regarding freedom and transparency. And so often (such as the comment I replied to, I think), even technical people seem to just ignore the difference. Open weights are just weights. No amount of open-washing changes that.
The raw training data is so large that very few parties could host it for free even if there weren't copyright barriers.
But I think you could have a full open source training software pipeline that's set up to work with Wikipedia, Common Crawl, Books3, Library Genesis, Anna's Archive, and whatever other useful data sets people can name. There would just be a step where you have to provide your own copy of Library Genesis (or whatever subset of it you have managed to obtain).
This comment is not entirely on point with your comment, it circles around and above it looking for lift though.
If you're not doing work that requires your code to stay in home nation data centres, Claude for Deepseek, Deepclaude (https://github.com/aattaran/deepclaude) is a great way to get better at using Claude like tools for software development. It even does a pretty good job of putting together cover letters for job applications...
Using Deepclaude is very much cheaper than using claude... For hobby projects, I've found it useful. A recipe (for cooking) management app I've made took a couple of hours to put together and cost $US 0.5. Claude is far more expensive.
The downsides of Deepclaude for many are:-
- DeepSeek is a Chinese corporation so the Chinese Communist Party may ask for data if it wants it.
- DeepClaude isn't as fast as normal Claude, though it's still pretty fast and I think fast enough (YMMV).
- DeepClaude might not be as optimised for various code issues that Claude may be able to solve more quickly or effectively.
- The same safeguards are probably on DeepSeek, but you won't be "wasting" as much money as you might on using Claude.
Inference focused hardware (https://www.youtube.com/watch?v=nvPqHoVSenE, AI generated speech) may in the medium future cause a large enough cost/energy reduction for LLM tools like Claude to make local LLMs more attractive.
Inference focused hardware would make running Open Source models like DeepSeek on local machines far cheaper and control over safeguards would return to the end user.
Hopefully this leads to a localised LLM provision market where local businesses provide varieties of these "local" LLM services. Here, local could mean on premise through to state or nationally based LLM services. Eventually, government orgs outside of the US may demand this kind of LLM use, in the same way governments legally require data to be stored within national borders for many critical government functions.
A bloke can dream I guess...
...Could affordable inference focused hardware also cause the bottom to fall out of these stock market bending valuations for AI corps and their datacentre obsessions?... Not to mention the societal costs caused by the AI super corps building these data centres. At the moment, they're nearly making a profit... They seem almost like speculative companies... Is that a term?
Fully agree. Shipping a complete product with a functioning user acquisition funnel is much harder. It's like; you have to build the whole product first with lots of features and then you have to try to create a highly condensed overview of all those features to expose them all on the landing page.
If you can't make the visitor understand your entire complex product in 10 seconds, then you've lost them.
Your product has to be complex because that's where the software market is at. All of the low-hanging fruits have been taken by the time you identify them. Sure, someone will find a way to make money using new low-hanging fruits that arise due to technological changes but it's not going to be you. You probably don't have the business connections to make that work.
I'm not entirely sure how that dismisses the CEO's putative argument: they go big on AI precisely because shipping end-to-end is hard, so they think they shouldn't waste resources on tasks that can be automated.
The structure of a good argument would be something like: certain tasks are fundamentally human and impossible to automate (which and why?) and by pushing AI use beyond what is optimal you are actually hurting your employees ability to do those hard parts.
A weaker but still useful argument is that most everything can probably be automated, but frontier models aren't there yet.
I wouldn't say it "dismisses" their argument, but I think AI marketing encourages them to take an over-simplified view of what it takes to ship product. Most folks like a good, simple story, as opposed to the unvarnished truth.
> "There's always an easy solution to every human problem; Neat, plausible and wrong."
-- H. L. Mencken
It's like the classic scenario, where you lash-up a barely functional UI demo, and the manager cuts your development schedule by 90%, because you "already have it working." That taught me to never do a lash-up demo. If I show something to someone, it is ship-quality (but often incomplete). It's a technique that I've used for years, and is a great way to involve nontechnical stakeholders, without risking stuff like "it's already working."
All that said, I think that AI definitely could automate a lot of the repetitive stuff involved in shipping. It's just that the CEO would fire the folks that could teach it, before it can learn, because they think that what they do, is "unimportant."
How do you actually get this. I've got a product, the site is hand crafted, shows the complex product really well (and had good feedback on it) but how do I acquire the users?
It seems as the cost of creating software has plummeted, it's the actual sales side of it that's going to matter even more. I'm stuck at this point.
"How do I acquire users" is the entire function of sales and marketing. A single HN comment explaining how to do sales and marketing, which is highly dependent on your product and market (and much more difficult than technical people tend to believe), is a bit unrealistic. And a great opportunity to use Claude/ChatGPT for something other than code. There's no silver bullet but as a springboard you can think about:
Who is your ideal customer profile (look up buyer personas) -- if you're B2B figure out both the profile of the company who would buy, as well as the person who would actually buy, and the person who would actually use the software: remember that buyer != user in B2B scenarios, and you'll have to figure out if the buyer, user or both is the best path to getting a sale. If you're B2C figure out your buyer personas so you know where to advertise.
Why would people want your product; sounds like you may already have this down but be ready to explain your value proposition concisely.
How will these people hear about your product -- a SaaS that falls in the woods doesn't make a sound, you need people to learn your product exists before they can pay for it. This is the point of figuring out buyer personas, you need to meet your customers where they are, and you can't know where they are unless you know who they are. This is highly dependent on your product/personas, and could range from running LinkedIn ads to SEO to having a Bluesky brand account to going to local meetup groups or conferences and trying to get your first handful of users in-person.
I like this analogy; raising children well like delivering products well pays dividends. They’re less likely to cause problems and if they do, they tend to be smaller in scope.
Great analogy all the way through. Also the last 10% takes thre most effort/iteration to get the work done such that you don't spend a lot of time maintaining it later.
I skimmed the article, guilty, but I think what I got from it is that CEOs will CEO? No disrespect meant, I’ve seen your name here often and thoroughly enjoy the folklore that you share, but I don’t understand what context you reacted to. Cheers.
The context that they think that shipping is simple. Shipping (what you need all those annoying peons for) is really terribly difficult, and has a lot of moving parts that designers often fail to take into account, until the deployment people lock them into a restroom stall, and refuse to let them out, unless they listen.
That's common with newer engineers (and now, non-engineers). I believe that Mr. Dunning, and Mr. Kruger had something to say about it.
I also spent most of my career at hardware-oriented companies, and shipping hardware is orders of magnitude more difficult than shipping software.
Thank you. You spent some time at Apple, no? There is that “real artists ship” or “great artists steal”, but I wonder what he’d say now. Just fun to think about.
The problem is that the lines of code are riding on a stack of other dependencies that all need care and feeding. Things reach EOL. Frameworks have major breaking changes. CVEs are discovered.
You can actually get high-quality code out of them -- at least with Claude; not had a great experience with Gemini -- but for complex tasks requires riding them very, very hard and really understanding where things can go wrong and poking at them repeatedly. Iterate, iterate, iterate.
That describes my last week. What made it most annoying, was the need to release through TestFlight, because the memory issues would not appear, when tethered. Also, I was checking in constantly, because I had to revert and reset the context, several times.
Yes. It's also why working as a software (host) developer at a hardware company is difficult.
Hardware people insist on treating software the same as firmware.
Bad firmware can cause real-world, physical damage, and be impossible to fix without a hardware recall. A firmware bug can wipe out a hardware company. A software bug can be embarassing, but can also be corrected a lot more easily (as long as it is being treated differently from firmware).
> So far, LLMs seem to deliver code with "Louie Da Loan Shark"-levels of tech debt.
Maybe a couple of years ago, but these days, Opus 4.8 is frankly writing better software than what I've seen over the previous decades in non-tech enterprise. These previous two months, we've replaced so much technical debt we've been dragging along for the previous 5 years as our team went from 25 to 3 people.
This is in non-tech enterprise in Denmark and AI had absolutely no impact on us going from 25 to 3. That was all Putin and bad business decisions on the c-levels. Like keeping flexible loans to fund projects on the books when the interests rates were 0.01% because they might go to 0.001%. Anyway, I'm getting to the point where the AI does 100% of the work, but only if it's piloted by people who know what security, resource consumption and compliance is. The code itself is excellent though.
Yup. In another post, I was grumping about having to accept a truly obese bunch of code from an LLM. This particular issue is the only one I could imagine even considering accepting that much pasta, but I sort of have to, because the LLM was able to quickly solve a problem that would have taken me a couple more weeks to address.
I remember a .sig that went something along the lines of:
I hate code, and want as little as possible in my software.
In various different projects I've been involved in where we've been implementing (not developing) software solutions I've noticed that, at management level, there is little regard given to the level of maintenance required of running the software; that people are still needed; that, yes, processes are automated, but there's a helluva lot of ongoing work required to ensure that new data won't pop the automation off the rails.
It's as if the installation part is the hard bit, and after that it'll take care of itself for ... far enough into the future that it won't be <current manager>'s problem. It is solved.
... and that's just using the system, not fixing bugs and adding features.
%80 of CEOs are Meh. %10 of the CEOs add value to the company. %10 of the CEOs are actually detrimental. The %80 will always try and do what they think the top %10 CEOs do because of FOMO. The bottom %10 will do it because they know their days are numbered and hope that it might put them in the %80 keeping their jobs. A good LLM could probably replace most and not do any worse.
Good CEOs don't see their employees as an expense.
Good CEO's make such good money on their employees that everybody gets raises and bonuses, the company grows responsibly, and the stock is a good investment.
Too bad it's out of reach for so many executives.
If that's too challenging, I understand, but if they had real confidence as a business operator I don't see why so many would be kicking out anybody over AI when they could at least continue to make the same money off the same people going forward. OTOH in cases where AI is almost ideally helpful it should be no surprise if hiring is slowed, and doing the accounting it could very well add up about the same either way. But one way clearly indicates the limited vision of a lesser leader, why settle for that?
Two of the most macro giveaway characteristics are emotionalism and superstition.
Not just for CEO's and other executives, but anyone in a leadership position or with decision-making tasks to perform.
One of the legendary combinations is when superstition is used in place of technology, and emotional reactions completely prevail instead of genuine business acumen.
It's a pretty good estimate that almost every CEO who thinks it would be good if AI replaced their employees, that these CEO's fit squarely in the superstitious camp.
I would say that's just one growing subset of a much larger smorgasbord of superstitions to choose from, and some big-shots are bound to indulge a whole lot more than others :\
Honestly I am burned out on public discussion of AI. There are so many hot, low info takes on both sides. All the dumb stuff revolves around this imbecilic notion that AI will take your job.
In that sense Altman and Dario et. al. were extremely successful in their cringeworthy campaign to establish themselves as priests of a new machine god religion. Even the people who don't want it believed them.
The good news is at least this year companies are starting to get a little more thoughtful about why they're paying for AI and what specific business function it serves.
Realities:
1. AI is a tool. You don't replace a job with a tool, just like you don't replace an apple with a sock. It was always an error of classification to think this way.
2. AI is a useful tool. For every CEO who thinks it justifies mass layoffs, there are dozens of people who don't want to admit that it does have a lot of utility. Anyone who isn't figuring out where this tool can make them more productive will have a hard time down the line imho.
3. We can infer from the priors that jobs will continue to be a thing, just like they still were after the inventions of printing presses, cars, power looms, etc. but there will be some pretty massive churn, some roles maybe disappearing entirely, new ones getting created, same goes for skill sets.
4. Businesses will use this churn to the best of their ability to either reduce costs or increase output. That second part is the key the AI haters don't seem to recognize. Thanks to competition, not every business wants to just fire everyone. Yeah, America's financialized and monopolized and less competitive than it used to be, but still, plenty of businesses will repurpose their budgets to produce more and expand.
So in terms of real specific and concrete things I've seen so far.
If you're in customer support that's a pretty rough spot, a well designed RAG system can turn 20 minutes of research into 20 seconds. Customer support budgets flow downstream from actual usage/customer base so if they can do more with less people will get laid off, they don't expand scope, they automate and cut the budget.
If you're a developer, your position's a lot more secure than that, coding agents are pretty incredible but they are simply not at a point where they can think through architectural decisions, and they occasionally go off the rails and trash everything. These things need to be operated and steered. Yes that's 100% the future of the profession and I'm sorry if someone doesn't like that, being a blacksmith who forges metal things by hand is also very niche these days, kind of sad if you loved blacksmithing but it happens. Devs also have to be aware that a PM/product owner can likely do some of their job now and I think any dev whose preference was to avoid thinking about customer or business requirements is going to find his utility is shrinking. A mass shrinking of the profession is unlikely but things will change and the only rational paths forward honestly are to retire or to figure out now how you fit in, it's a career opportunity if you get ahead of the curve even.
Lastly there is a whole other domain now which for lack of a good term I'll refer to as "prompt engineering," it requires some level of systems design thinking, but basically no programming expertise. The best candidate for this work is a person who possesses business process knowledge as well as systems oriented thinking. Maybe these jobs will end up finding a home in the IT department or something. As an economy we're barely recognizing them yet but I see that in engineering we can increasingly implement places for a prompt input, and then hand some workflow/business domain decisions off to someone who understands them better and they just tweak their prompts over time to get a great system. Pretty sure new jobs will be created and formalized around this over time.
now it's closer to 95% of work can be done by AI and requires 5% mental effort, but 5% of the work requires 95% of the mental effort to finish because of all the unoptimial decisions AI has taken. I find that AI works best in small micro-service type architecture where each component has a clear goal and doesn't have interconnected parts within the same application that can break. But you do run into an issue where changes in microservice a need changes in microservice b and updating it is not ideal since it usually cascades thru the entire system or requires stacks of legacy support.
IME it’s possible to have good clear APIs, limited scopes/goals, etc in a normal (macro?) service. But it requires a level of discipline and process many teams are unwilling to engage in.
There are a lot of bad CEOs, though. It's a lot like a politician -- it's quite difficult to become a CEO, and the skills to make it to that position don't always intersect nicely with the skills necessary to actually do the job well.
CEOs do get there with lots of politics in almost all cases. It’s all about who’s ass you kiss and who’s ass you don’t and if you’re lucky with timing things might just fall into place.
I think it’s exceedingly rare that a CEO is actually competent at their job. In most cases it’s the labor class propping the company up, and in some cases the workers are doing so against the wishes of the CEO. Not that executives want to ruin the company, they’re just incompetent and therefore make terrible decisions constantly.
> It’s all about who’s ass you kiss and who’s ass you don’t ... I think it’s exceedingly rare that a CEO is actually competent at their job.
But... that is kinda the job? CEOs are, first and foremost, the public face of the company. They're the one who talk to VCs / banks, regulators, major customers, the press. They're very highly paid PR reps / fall guys that shield everyone else, including the board of directors and all the VPs and SVPs, if something goes wrong.
For most companies, especially large companies, it's not important for a CEO to be good with software engineering, business development, etc. That, at least in principle, can be handled by other parts of the hierarchy.
> it's not important for a CEO to be good with software engineering
If you are the CEO of a company, you should have expertise in whatever your company does, and If your company is primarily a software company, then you should have expertise in software engineering. You cannot effectively manage something that you don't understand.
> CEOs do get there with lots of politics in almost all cases.
I know this can be hard for engineers to sometimes accept, but relationships (aka politics) are a key part of business. Rarely is one technical solution absolutely superior to another, making purchasing decisions come down to relationships.
Politics is also about compromise and managing a bunch of differing opinions/desires, which is one of the key skills of a CEO.
I wonder if business schools could ever start actually teaching this skill. So far they just largely operate as scams, held up purely by prestige and network value
I’ve worked with CEOs in multiple large companies. I wouldn’t wish that job on my worst enemy. Nonetheless, someone needs to do that job and the intersection of difficulty and masochism is beyond what most people can do or endure. Many people try and fail. Their job, at the end of the day, is to eat an endless stream of shit sandwiches with a smile and a plan.
Much of the “competency” of a CEO in practice is to be able to accept the relentless drama and abuse without turning into an emotional wreck. Yeah, they have to make decisions, but that isn’t the part that makes the job difficult. That role takes an insane toll on the human spirit, and very few can do it for any length of time.
The cush job is often being CEO adjacent. You get most of the perks but also avoid most of the emotional abuse and drama.
This feels too soft. Each of these things has truth in it, but isn't some of this self-created? Where are these shit sandwiches coming from? A lot of these problems are the result of overpromising, breaking rules and skirting regulations, underestimating the difficulty of things they have no expertise with, asking people to solve problems with no resources, hiring more chefs rather than more cooks and dishwashers, of mismatches between good profitable product and exec exit. The idea that it makes sense for the CEO's (or really any leader's) core competency to be absorbing drama and pain is something we should think more about. Sometimes you hear that a good manager blocks and shields for their team, you have to wonder why the team always needs so much protection from their own company and processes.
I'm very sympathetic to cooperatives, have traveled/know the Mondragon people (largest coop federation), etc.
However, I think there's a reason why coops seem to succeed at smaller scales, but there are essentially no large innovative coops.
There are a few large boring coops, and some small innovative ones, but seemingly something is making the CEO/investor board model the one large innovative companies are all using.
I suspect that it's both (1) access to capital is far harder for coops, and (2) that workplace democracy and hardcore mission focus aren't fully compatible. That is, "you cannot serve two masters" without losing focus on one of them.
If a company accumulates capital, it becomes vulnerable to the principal agent problem, and coops are way more vulnerable here than centralized companies.
If a company doesn't accumulate capital, it doesn't scale in complexity. It can grow by having more people do more of the same things, but it can't move into markets that demand anything complex.
This seems hard to tease out from the fact that a) the majority of companies do not survive, b) the large, large majority of companies that do survive do not wind up being large or innovative, and c) there are far fewer coops than regular companies. If you assume equal chance of success between them, you’d still see vanishingly small numbers of large, innovative coops, because a small percentage of a small number is small.
Coops can definitely succeed and even dominate at large scales with some minimum government protections. The largest dairy producer in the world is a coop: https://en.wikipedia.org/wiki/Amul
The problem is that knowing the right people to get investment does seem to have utility coops struggle to get, I think? maybe CEOs are basically like producers on movies who are just there to network for you.
Coops tend to have better aligned incentives for employees on every step of the ladder. They'll tend to be more conservative about R&D but ensure that money that's being spent is being productive for the continuing health of the company since instead of that budget being "corporate's money pile" it's your potential profit share.
I think there's also a tendency towards longer tenure and higher value employees due to the investment in the company's future being a sort of central tenant of their attractiveness.
Generally, yes. Also the gap between the employee and executive class is a lot smaller instead of unnaturally inflated like it is in most private equity companies.
In software I can imagine a worker-owned consultancy, but not a product company. It would imply staying in one place working on one product for your whole life, which doesn't sound inspiring
A company need not be a single product, and working in a worker-owned cooperative need not be a lifetime commitment to a single firm (though cooperatives ideally will have less turnover than firms owned by capital separated from labor.)
Acting like centrally planned dictatorships is a good form of collaboration is just so off base. There's no reason to think that introducing democracy into the work place wouldn't immediately benefit both workers + customers.
If this sounds crazy the C suite + board already vote on who gets hired into the executive team, vote for the direction of the company, and vote for their compensation packages (hint, they never decrease them).
Why shouldn't workers be legally enabled to do the same? What is the justification to this? I'm curious to hear it because the only way people can justify the current system is declaring that some people are actually more deserving of prestige, money, and benefits while others deserve to suffer.
With income inequality increasing, healthcare outcomes worsening, and children literally becoming stupider isn't it time to question the current system and ask ourselves if this is the society we truly want?
There is academic research on this too if you're curious but it's mostly in English, Spanish, and Portuguese.
But yes, there isn't much "evidence" because this system hasn't been tried en masse; however if you look at our current neoliberal hellscape, it's pretty hard to imagine it doing worse. Also neoliberalism wasn't really "tried" either, it was thrusted upon us by a group of individuals that wanted it.
One thing to keep in mind is that society can change quite quickly if you want it to. I'm sure the children that died working in factories during the 1800s never imagined such a society where children are valued, cared for, educated, and protected but it did happen.
It has happen before and it can happen again. It only happened because people were willing to fight for it.
The rules are allowed to be changed at anytime if we deem so, a better world is possible.
This is a shallow dismissal of GP’s point. The point is more, “we aren’t sure it won’t work because it has never been tried,” which is much less of a straw man to argue with.
“Acting like centrally planned dictatorships is a good form of collaboration is just so off base. There's no reason to think that introducing democracy into the work place wouldn't immediately benefit both workers + customers.”
That sounds more like there’s no evidence that it won’t work than an unequivocal claim that it will.
Well... there is historical evidence that centrally planned dictatorships are not a very responsive form of government.
Now, corporations usually have the problem of competition, so if they aren't responsive (or at least responsive enough), they get out-competed by those that are. Is that enough to make them different from governments? Perhaps, but I don't know.
If you don't think there are competitions in dictatorships you are extremely sheltered. The competition in a dictatorship is whether you stay alive or not, just like in a corporation is whether you become homeless and die or keep a roof over your head.
Why does this argument never apply to neoliberalism?
That was never put to a vote but it still thrusted upon a country where the results are what you would expect: the worse income inequality ever seen in the history of the nation, life expectancy has increased, deaths of despair have reached record highs, more children go to bed hungry, healthcare is being ripped from civilians, and corporations are legally allowed to poison and kill civilians (health insurance companies with their death panels, manufacturers causing cancer valleys) with zero legal repercussions.
So yeah maybe we should actually go the extreme into the other direction, if democracy is good enough to lead nations it's good enough to run businesses. If you're a worker IDK how you would argue otherwise. Being able to keep your boss/leadership accountable by voting for them out seems like a win for every workplace metric imaginable.
Imagine how better of a company Meta would be if Zuckerberg wasn't allowed to waste billions accomplishing nothing. In a just society he would have been voted out, but in a neoliberal society he is granted an insurmountable amount of wealth.
I'm sorry but this society sucks and acting like we can't do better, be better is beyond pathetic.
Worker owned cooperatives have a variety of ways of doing this. Voting directly, electing people, etc. The main difference is that the cooperative typically doesn't buy the myth that the person making the high level decision needs to be paid 1000x the workers.
Maybe, but not necessarily for this reason. Even in a worker-owned coop, someone sets the overall direction. And how is that person going to be selected? It's still going to be largely politics.
So, we live in the economy where everything is done by labour that receive minimal acceptable wages, while all the profits are collected by the ruling class who do nothing except constantly kissing each other asses.
on the contrary, it seems to be one of the few jobs that seems to require absolutely no qualifications to have.
What you need to do to be CEO is.... convince someone to lend you money in the hope that you'll get it back to them.
I've worked under some absolutely awful people who wouldn't pass an interview anywhere, but somehow they're CEOs, because they can smarm there way into more money consistently.
>convince someone to lend you money in the hope that you'll get it back to them
And it should be noted that many of these people lending money are in a similar situation of not being required to have any qualifications. Sure, some of them have worked their way up through sound investment after sound investment, but many of them were either born into their position or simply got lucky at some point along the way. Just think of all the money investors threw away pursuing crypto and NFTs for example. Many of those investments were transparently stupid from day 1.
Often, they are good at taking things, keeping things, misdirecting and setting boundaries (especially communication boundaries). They are good at keeping their positions.
This is a broad range of skills and to actually be a CEO, you need to really hone these skills and be among the very top. To be good at those, just enough to qualify for a modest CEO role at a small start-up, you generally don't have the time to be good at anything else.
Saying that you don't need any skills is mischaracterizing it. You don't need any value-creating skills, yes, but you need significant value-capturing skills.
I can imagine a world were all companies become empty of workers and only executives remain and they would just have meetings with each other while they starve and would explain it away as a new diet they're on. There would be no petrol and they would be forced to walk to work and would say that it's their new fitness routine... And they would all believe each other.
Most of them got into a prestigious school on legacy, paid for by wealthy parents. Many were above average IQ, but by no means geniuses. They had access to computers earlier than others, due to said affluence. They seem unable or unwilling to comprehend they're overwhelmingly on average, "nepo babies" to steal a term from the world of entertainment.
I think it's private schools in general. Even those from second and third tier ones, which can filter more by means than the elite ones, find themselves atop companies. It's the natural access and the natural ability to socialize with other private school personalities. Their definition of capable leader is a particular type of leader. They can make and take jokes, but particular types of jokes. They hide each other's shortcomings, insecurities, guilt whereas people from other backgrounds, even people they like and think highly of, tend to serve as a mirror.
Getting funding is a value add, but I agree calling it "skill" misses most of what makes someone "good" at it. We've built things to overwhelmingly rely on funding gated by other private school people though. It would be nice if we could have that person with access pitching without them also being in charge of running a company, product development, or managing people. But then it would require the same of the investors. The investors would then need to evaluate products and ideas and markets. And the markets would have to reward that. Things would need to be different.
I think CS is a little unusual in that places like Berkley are ranked highly. (Interestingly, Harvard has a rather low CS rank and Zuck was in the process of transferring to psych because he can't handle anything quant, but due to FERPA no one will call him out even as he repeatedly steps outside the law)
>Getting funding is a value add, but I agree calling it "skill" misses most of what makes someone "good" at it. We've built things to overwhelmingly rely on funding gated by other private school people though.
I think part of the issue is that a big chunk of startups are straight up grift. My complaints about gatekeeping aside, I think if I had a legitimately good idea I could either submit to yCombinator or talk to contacts who've cashed in stock and get investors.
That's mostly due to the fact I'm well known for valuing authenticity though and my personal brand is such that folks would take my proposal seriously.
I met a LOT of people who went to places like Stanford who basically... they always meet the metrics, at the expense of all else.
Anyways, I think in general we try to generate too many
"startups" when we're really striving for small businesses.
But because the funding is the truly difficult part, we hyperfocus on that.
Meanwhile, I've seen plenty of startup ideas that while benefiting from the Ycombinator social network, have initial costs such that you could hit up a few rich friends who've known you a while with a solid business plan and get bootstrapped... but that would require domain knowledge and the respect of your peers, two things many founders lack ;-)
It's not difficult at all to become one, and the work involved in being a CEO is not particularly difficult in comparison to senior technical work at all. The only thing that is harder about being a CEO is the responsibility. I'm sure being the CEO of Microsoft or whatever is plenty difficult and demanding in many ways, but most CEOs are not that, and speaking just from experience most CEOs and CTOs are clueless morons.
With that said, I've been programming for 25 years and I've only been a CEO for 3, so take what I said with a pinch of salt.
I do think people overestimate titles like this a lot, though, and it really comes down to what the company actually does and what is demanding for that company at that position/role. The CTO of a some-bullshit-as-a-service company may as well be straight out of college, because they're likely doing something trivial that literally anyone (including LLMs) could put together. The CTO of a well-used and reliable streaming service that handles a meaningful part of the world's Internet traffic is obviously solving a more interesting and demanding problem, and their decisions are going to be more important.
Years of ZIRP, QE, bailouts and stymulus money muddled the waters a lot. Add to this the Old Boys (and Girls now) Networks, a culture that values getting money fast as the ultimate value, the prevalence of politics and you end up getting a boatload of bad CEOs.
There was a time when I used to recommend "Out of the Crisis", a book from Demming, to business leaders.
Problem is, "Out of the crisis" still assumes as a premise that companies compete on the quality of their products, that making money comes from actually making and selling stuff. That leaders are not the anti-intelectual morons that believe that absolutely any thing can be explained with a 15 minutes deck, that math and statistics are passtimes for weird geeks that don't add "business value" and because of that, is a book that could have steered us toward a better world in the 80s, but now it is completely useless, because its recipes can't handle the level of degradation things goto into.
He's the face of bad CEOs because people like to make up things about how bad he is. The Social Network, a major 2010 biopic about the early days of Facebook, famously cut his college sweetheart and now-wife out of the story in favor of a fabricated character arc involving an ex-girlfriend who does not exist.
Poor Marky Z. always getting a bad rap. Let s/he who hasn't made zillons getting billions of people hopelessly addicted to social media while facilitating a genocide or two and generally destroying the world order as we know it throw the first stone.
From your tone, it sounds like you may just be intending to warn me that it's cringe not to agree with any criticism of Zuckerberg? If that's so, I have to respectfully disagree; I think this is a bad attitude that leads towards being poorly informed about the world.
If you're interested in discussing the specific claims you're making, I really don't think that billions of people are hopelessly addicted to social media, and I would love to hear your basis for claiming this. My understanding (from e.g. https://www.nature.com/articles/s41598-025-27053-2) is that genuine compulsive use of social media is quite rare, and most people who describe themselves as "addicted" are just regular users who enjoy it but kinda feel like it's a waste of time.
> Meta Settles Lawsuit That Claimed Social Media Addiction Screwed Up Schools
> On Thursday, Meta settled a lawsuit brought by a Kentucky school district that claimed the tech giant’s social media platforms have created a mental health crisis at its schools.
> The case was considered the first of its kind and a bellwether (a case that is representative of a large pool of lawsuits and will be a test for future litigation). The plaintiffs argue that social media platforms have had a major negative impact on the mental health of school-age children, which in turn has caused a burden on the education system, as American schools were forced to redirect resources to counter this problem.
> The settlement comes shortly after Meta lost a key bellwether social media addiction trial. Back in March, a judge in Los Angeles ruled that Meta was liable for the adverse mental health effects a now 20-year-old suffered after getting addicted to Instagram from an early age. The representatives of the young woman argued successfully that it was Meta’s deliberate design choices, like the infinite scroll and face-altering filters on stories, that had exacerbated her addiction and subsequent mental health issues like self-harm and depression.
> The military junta in Myanmar was facilitated by Facebook to post hate speech that sought to foment sexual violence and promote genocide against the Rohingya. "Myanmar would have been a better place if Facebook had not arrived" Wynn-Williams writes.
> Wynn-Williams argued that Facebook failed to moderate hate speech against the Rohingya in Myanmar, including the use of the racial slur kalar. She noted that the company only had two Burmese language moderators, both based in Dublin, for the entire country, and claimed that one of the two moderators gave a pass to hate speech while removing pro-human rights content. She further claimed that she raised concerns that the moderator was "in cahoots with the" junta, only to have her concerns dismissed by the content team. Additionally, she claimed that her efforts to have Facebook's Community Standards rules translated into the Burmese language were resisted by the company communications team, who told her that "Myanmar isn’t a priority country" in the region.
> In the 2010s, personal data belonging to millions of Facebook users was collected by British consulting firm Cambridge Analytica for political advertising without informed consent.
> The data was collected through an app called "This Is Your Digital Life", developed by data scientist Aleksandr Kogan and his company Global Science Research in 2013. The app consisted of a series of questions to build psychological profiles on users, and collected the personal data of the users' Facebook friends via Facebook's Open Graph platform.[2] The app harvested the data of up to 87 million Facebook profiles. Cambridge Analytica used the data to analytically assist the 2016 presidential campaigns of Ted Cruz and Donald Trump.
> Other advertising agencies have been implementing various forms of psychological targeting for years and Facebook had patented a similar technology in 2012.
> Facebook manipulated the emotions of hundreds of thousands of its users, and found that they would pass on happy or sad emotions, it has said. The experiment, for which researchers did not gain specific consent, has provoked criticism from users with privacy and ethical concerns.
> For one week in 2012, Facebook skewed nearly 700,000 users’ news feeds to either be happier or sadder than normal. The experiment found that after the experiment was over users tended to post positive or negative comments according to the skew that was given to their news feed.
> The research has provoked distress because of the manipulation involved.
> Studies of real world networks show that what the researchers call ‘emotional contagion’ can be transferred through networks. But researchers say that the study is the first evidence that the effect can happen without direct interaction or nonverbal clues.
> Anyone who used the English version of Facebook automatically qualified for the experiment, the results of which were published earlier this month. Researchers analysed the words used in posts to automatically decide whether they were likely to be positive or negative, and shifted them up or down according to which group users fell into.
It's on my mind because they're releasing a sequel in October, and Aaron Sorkin has not (as far as I've seen) acknowledged that much of the original was not true nor promised to be more accurate this time around. I'm pretty confident that it's going to be about the same mix as last time, and I'm going to have to go around saying "actually the scene where Zuckerberg did suchandsuch terrible thing wasn't real", and people will respond by insinuating that it's lame for me to care.
I saw someone on Xitter say "Any CEO who wants to replace jobs with AI should first have to replace their own assistant with AI" and I think that's the perfect rule. Every AI demo is some version of a personal assistant, surely AI can do that job right?
I think we'd get zero volunteers from CEOs who have assistants.
(Note: this is not meant to be an insult to human assistants! I think they do a valuable job and should not be replaced by AI either).
AI is the new outsourcing. It is cheap in the short term, but in the medium to long term it generates many problems, like loss of know-how and competitivness, huge maintenance cost, loss of control, and dependency on other foreign companies, with misaligned goals compared to your company.
Before using AI ask yourself: Would you outsource this task, with all the risk that come with it? If yes, go for it. But if no, then don‘t use AI.
it's kind of like saying like you could replace a king with an AI. the position is a relationship to power more than it is a function with a productive output.
a bad king and a bad CEO could be replaced with a spinning top with no loss in productivity (and maybe some gain).
I still remember when I used ChatGPT the first time to write an email. I thought to myself “Oh. This sounds like 99% of the corporate communication from above”. We were joking that corporate had BossGPT for years and just didn’t tell anybody.
It also made me realize that most the so called “creatives” in marketing and PR also just repeat variations of the same few templates. Not much real creativity there.
Who takes responsibility when the AI does something unethical or illegal? Do we put the computer in jail? Or do we just look the other way like we do with human CEOs?
I think Grok on x.com is a reasonable case study. There’s the stuff we heard about and then there’s a separate category of things that weren’t newsworthy, and only the newsworthy things prompted changes to be made.
My charitable word would be average. AI means average. And an average CEO isn’t that bad. It’d be a deliberate trade in exchange for less spending on salaries. (just one very large salary)
Weirdly enough, I'd take the robot. At least we can pretend the robot doesn't know any better. The human is actively choosing to be a dick and profiting off it.
The flesh sack is choosing to be an insufferable twat, and the robot either doesn't have any choice or has a decent statistical justification for what it does.
It's pretty disappointing that people on this site of all places have no idea what CEOs do. Many of them are certainly overpaid, and like any other profession, many are not good at their jobs, but they aren't sitting around drafting memos and coming up with deciding who to fire all day.
I reckon such AIs are already in place, but by proxy. There must be CEOs somewhere who have completely offloaded what meagre amount of thinking they needed to do to some bespoke AI setup while they LARP around convincing people they are the one running things.
I do think it's more subtle. AI can replace very few jobs end to end with the same quality, maybe none. But AI can be put to work on high ROI problems. Now when the new marginal job is not obviously as high ROI as putting another 100k of tokens to work, no human gets hired.
Next, comes natural attrition in a company where a certain percentage will leave every year. Will they get replaced with a human or their budget goes into tokens?
Only when these 2 angles are exhausted, a typical company will start thinking about layoffs.
Now, some companies are already stressed: customer buy AI products instead of theirs, AI makes it easier to build what they offer, customers believe they can vibe code things. These companies will layoff first, because of AI. Not because AI will do the persons job but because the money gets spend differently.
AI will never be able to replace a human board member. I think that's what bothers technology people the most here. It's quite asymmetric. At best, LLM agents will be granted certain non voting advisory roles, but anything the llm generates will need a human to assume responsibility.
The point of a human being on the board is that they can be made to suffer the consequences of bad decisions. Executive pay might seem astronomical, but it is often commensurate with the level of stress and responsibility involved. It's easy to look at the perks and not see the struggle behind them. No one in media is going to get a lot of clicks if they publish articles about how being a CEO is actually really hard and maybe some of what they're doing is actually justified once you assume all of the same context and stress they have.
If we want to do to the executive staff what is being done to the technical staff, I'm afraid we will need to first figure out a way to make the AI experience human emotions as strongly as we do. It often sounds fun in first order terms to threaten to fire a CEO and replace them with a clanker, but have we considered the consequences of this? What would it be like to be under the management of an emotionless robot? Being managed by a robot vs managing robots are wildly different things to me.
> Executive pay might seem astronomical, but it is often commensurate with the level of stress and responsibility involved.
I call bull on that. Employees often have just as much stress, often imposed on them from those higher up, where if they don't do well they could get fired and it could mean they can't put food on the table.
Several times I've worked in environments where the stress was incredibly high, and once it even put me in the hospital after a long stretch of working 70 hour weeks to satisfy the insane deadlines my CEO boss wanted (this was at a startup).
Whereas if an executive screws up they just golden parachute to another cushy gig elsewhere, oftentimes, because even if they screw up, they're still valued as having experience running companies by other companies. And even if they cannot, they still likely got paid enough that they won't be struggling financially if they go through a period without a job.
> What would it be like to be under the management of an emotionless robot?
Depends. Is the robot looking out for our best interests? Or is the robot looking out for the shareholders?
I already ask AI to manage a lot of things, including my own activities. It all depends on whether the model can be trusted to act in our best interests.
CEOs understands that AI offers potential productivity increases. Using that productivity boost to cut staff is an unimaginative approach. Bolder approaches include using that boost to exceed the expectations of current customers, or to increase sales without proportional increase in staff, etc.
Counting token usage as a productivity metric is completely counterproductive.
I believe that effectively leveraging AI means reducing wasted tokens, not increasing them.
The visibility gap is serious.
I create local activity logs for Claude Code, but most developers have no idea what the AI is actually doing at the file or command level until they look at the logs.
Just because people have a vague idea of what someone's job is and a deep understanding of their own job, it feels like AI can replace everyone but you.
The working horse population continued to grow for almost 20 years after the introduction of the Oldsmobile in 1901: https://www.reddit.com/r/dataisbeautiful/comments/i2lmdx/num.... The number of horses didn’t fall below the 1900 level until 1930, two decades after the Model T was introduced.
Modern day CEOs of public companies are just more or less hedge fund operators looking to squeeze every last dollar out of their workforce. AI is a tantalizing, if ineffective, lever for that.
>> Yes, the tools are powerful, but a CEO who thinks they replace the work of employees is simply a bad CEO.
This is a broad generalization of employees. There will be some "routine tasks" that can be done by AI, now that is a lot more powerful.
There won't be as many employees needed for routine work - for example L1 and L2 support work. For example, many companies had ML engineers building models for various models. Companies can get that off the shelf from AI companies. They don't need a big team of model builders now.
If L2 support work is the example then I doubt we’re near replacement.
L2 issues are already involved in some way often revealing some kind of system failure, requiring context and exploration to understand, and judgement (and perhaps even system overrides) to fix.
I could see “automated L2 is the new L1” improvements, but without a big capability jump and/or a resource bonfire, I don’t think even frontier models would effectively replace good L2 staff.
They might magnify good L2 staff so fewer are needed (and maybe even help L1 staff become L2).
My agent is currently doing the work of ~6 senior engineers, based on ticket closures. These are not trivial tasks: all of them involve a mix of code, judgement, executing remote commands in a production environment, etc.
This is at a well-known tech company operating at massive scale (and resulting complexity).
L1/L2 tech support is completely dead within a couple years. The delay is only around how long it takes people to realize.
I often use AI chatbot to generate step-by-step instruction for setting or repairing up some bit of tech. It's incredibly empowering, and saved me a a lot of money that would have been spent on buying replacement tech.
You know who can't do that? People who call L1 support.
The message isn't subtle, and isn't meant to be: "we don't care how, but we expect you to stick your nose into AI tools and find some way to fit them into your workflow".
Which indicates: the management believes there are productivity gains from AI use, but adoption lags due to inertia and reluctance to change existing workflows.
> Which indicates: the management believes there are productivity gains from AI use, but adoption lags due to inertia and reluctance to change existing workflows.
Methinks adoption lags due to management's inability to align incentives such that productivity gains are rewarded.
more of "we whined and cried and screamed that we needed new budget in order to buy these tools or we would literally die. now we have them, they don't work as well as we hoped, they aren't leading to productivity gains, and they're actively alienating our workforce and users alike. we're so screwed we literally have no idea how to do reverse this."
Ideally contractors that benefit you personally (eg: your buddy who now owes you one), but definitely contractors that let you outsource the responsibility.
Even better if you get some management consultant to suggest the idea and/or do the subcontracting.
Definitely buys you a few quarters of bonus and some time to land your next gig.
Early in my career I read "Peopleware" by Tom DeMarco and Timothy Lister. Between that book Brooks' "Mythical Man Month" and Humphrey's "Managing the Software Process" it seemed that there was hope for the software industry learning some necessary lessons and growing to become a true engineering discipline. Nope. Never happened. The industry standard, despite improvements in some areas, is still a farcical shitshow with little beyond lipservice to process, predictability or proper self-evaluation. I can only describe agile, as it is practised, more of a coping mechanism than an actual methodology. Indeed the methodology is embodied mostly in the infrastructure; issue tracking, version control, code review, continuous integration with as little methodology glue between them as required to produce output.
Modern first and second tier software management seems less professional, is contributing less and is generally worse than it was twenty years ago. The quality of the engineering and program managers, their training and commitment to their craft seems really low and is not generally valued. On average team level software management has gotten worse rather than better and, given what is expected and how it is valued, this shouldn't be much of a surprise. It is truly disappointing that what could have been a valuable and productivity enhancing role became so useless.
Things aren't going to change for the better though until the dust settles somewhat on the role of AI in software and systems development and we start again to consider how software should be developed in the 21st century. Maybe it is possible that with AI doing most of the low-level work that the focus will change to building and maintaining architecture and systems. Many programmers might become more like traditional engineers doing a lot more systems work than they do today and continuing to solve problems. Lord knows though it won't be today's software management doing this work; they have nothing in the way of skills to offer to the problem.
I've been a founder since before any of this existed, same flavor of small team then and now. What I see now is that anyone can produce more but less due to their skills from scratch (slop?). A big company sees the headlines and a CEO reads that as "fewer people", or force use of LLMs and people seeing it as a preparation for AI processes taking over (and rightfully so). But frankly probably for smaller teams it reads "same people, wider surface". Imo bad CEOs existed before AI and will exist in the future...
CEOs are probably the most replaceable position, period, by human or AI. Everyone just gives them information. They don't know any information themselves.
Problem is, a CEO can fire employees, find out it was a dumb decision, then leave with a million dollar severance package. So they don't really care when they're wrong.
A story from today - I needed a small utility to remap my logitech buttons under windows without installing their horrendous GHUB. Logitech Onboard Memory Manager still required ghub to be put into onboard mode.
The solution - linux has utility called piper. I downloaded the repo and just told codex - figure out what piper is doing and create me a small utility to do it under windows. So the jolly critter started experimenting with hex commands, then pulled some other repo on which piper depended figured out how to enable said onboard mode and 10-sh minutes later I had small python script that did what i needed to do.
That would have taken probably half a day of work for a human.
There are many stupid CEO and organizations which are not committed to quality. And a lot of employees that are too set in their ways. But the instinct that underinvesting in AI is more dangerous than overinvesting is right. Doomed if you do, doomed if you don't
"To err is human, but to really foul things up requires a computer" - this is from the 60, but right now is turned into overdrive.
> I will say that I hate the term “AI psychosis” because the term is extremely misleading, and many psychologists and psychiatrists have complained that it is inaccurate and may cause more problems itself. But the general sense that CEOs are going overboard with AI is definitely happening.
It's getting exhausting how x field of experts constantly bemoan the coining of one term or another, rather than provide a decent alternative. It's not very goal-oriented thinking. Just empty complaining.
When people object to the coining of a term, what they're usually trying to say is that it just fails to refer to any coherent thing or group of things for which you could develop an alternate term.
"AI psychosis" tries to group together the phenomenon where people fall in love with an LLM-generated character, the phenomenon where people spend too much time talking philosophy with LLMs and stop expressing coherent thoughts, and the phenomenon where tech evangelists say "AI can do all of these 5 million things" when actually it can't do all of them. But what if these are all different phenomena with different causes and solutions?
We're all now on this and we will go together in it till the very end, whatever it maybe.
Look around: lots of places ressurected Lines of Code as a productivity metric for Software Teams. Companies that should have known better, as they are supposedly led by the Elite Human Capital, instituded token usage leaderboards.
We can't stop that thing. It has too much momentum. Sooner or later we would have to pay for our culture of anti-intelectualism in business anyway. If it was not this, it would be another thing.
The primary product of AI is labor displacement and consequently wage supression. This is what OpenAI and Anthropic are really selling. It didn't start with AI but AI is accelerating it.
This is what layoffs have been about since the pandemic. People in fear of losing their jobs do extra unpaid work and aren't asking for raises. The theoretical potential of AI gives companies the excuse to fire more people. The investment itself is directly used as a reason of why they need to cut back on labor.
Any sufficiently sized business can only feed the insatiable hunger for ever-increasing profits ultimately by cutting costs and raising prices. And what do we have now? High inflation and a decline in real wages. CEOs are just following this playbook.
And the result is that society is bouldering towards collapse. We're seeing the first hints of this with the youth unemployment crisis [1][2][3].
Also, who is going to buy anything when nobody has any money?
> Also, who is going to buy anything when nobody has any money?
This assumes that a mass consumer economy is necessary, when it isn't. Mass consumption is relatively new, for most of history economies functioned with just a small consuming elite and large underclass that consumed very little. We are already approaching that again in the states given that the top 10% of earners are already responsible for nearly half of all consumer spending.
There's a floor even in a mostly automated economy where some services are resistant to automation simply because the human element is the product. Luxury hospitality, personal care, etc. That billionaire is going to want a human masseuse, not a robot.
A highly automated economy could stabilize like this with a small elite population consuming luxury goods & services, served by a low-wage economic underclass human workforce.
Its certainly not a pleasant society, but its also not unsustainable given enough oppression or pacification (bread and circuses anyone?)
I don't have any sources regarding someone preferring humans for certain services over robots, just intuition there, and the fact that consuming human labor and time is itself a status signal of wealth, and the current growth of personal services.
As for connecting the dots, look at Brazil, one of the world's most unequal economies having a small consuming elite and a much, much larger low-wage service underclass. The gulf states as well. Granted, their circumstances don't map cleanly to a post-automation Western economy, but it does demonstrate that a largely bifurcated consumption based economy can exist and can be mildly stable.
Whether the US falls into that direction too will depending on politics. A bunch of mid-career knowledge workers aren't going to willingly to flip burgers in a service economy without some serious surveillance and oppression. But when thinking about it in those terms, the recent push for mass surveillance laws and tech along with the increasingly dangerous rhetoric around protests and "domestic terrorism" start to make sense.
> but it does demonstrate that a largely bifurcated consumption based economy can exist and can be mildly stable
This is not the case. These countries don’t operate/exist in isolation. They are part of global economy. But if this tech has a global scale effect, then the dots are not really connecting. You’re also forgetting the “revolt” factor.
The revolt factor is accounted for by ways of mass surveillance, and imprisonment for whatever gets labeled as domestic terrorism, like say, protesting data centers and "anti-tech sentiment" as a recent example.
This is a great article, and I agree with most of it.
The problem is that the wrong eyes are seeing it.
We need these kinds of articles to be published in places that executives read, and tailored to their audience.
AI recovery is going to be a big wave of consulting over the next several years, maybe very publicly or maybe quietly, but it's going to be a thing. That doesn't mean "all AI is bad" or any other such nonsense, it means that there are a lot of companies out there right now that are doing it wrong and will need help.
The executives that get ahead of this are going to be the winners.
So, a lot of the article makes several points that aren't necessarily new, but
> The problem tends to show up when a CEO is handed an agentic tool like Claude Code, and has it create something, which will work just fine, and thinks “oh, wait, why do we need so many people, when I can just sit here and make things work?”
> This is a bad CEO.
As described, this seems to me more like a lack of reasoning/critical thinking ability, and it's not unique to CEOs. Tracks more with a combo of "Gell-Mann Amnesia" and automation bias IMO.
> This all reminds me of cargo cult thinking: The CEO knows that somewhere in the org, employees are pecking away at computers and work gets done. So they figure that themselves pecking away with Claude Code and seeing work get done is the same thing. It’s not. All those other steps those people are handling — the ones the CEO never sees — still need to happen.
"Cargo culting" as described here by the author may be happening. But, I think it's CEOs seeing other CEOs doing layoffs and claiming it's because of AI efficiency gains. They see the other CEO's stock go up/get hyped/etc, so they decide to do it. I think it's the same thing that happens inside companies IE people see how others behave and it works, so they do the same. Effectiveness aside because that's not at all what I'm arguing, AI is just the current flavor; it is a very safe thing to "cargo cult" at the moment.
Similar to Tesla naming it's driver assistant "auto-pilot" in 2015 and your average Joe thought he would be able to sleep while the car would drive him to work.
The CEO just hear AI and think of AGI. They expect Skynet.
Aircraft autopilot is over 100 years old, and Testla autopilot is an automotive version that fits the definition by analogy to what aircaft autopilot does and what the human metaphor means (doing simple tasks aithout higher-level cognition or handling of surprising stimuli). Autopilot is not end-to-end driverless.
> the definition by analogy to what aircaft autopilot does
It's a bad analogy. Autopilot just maintains the aircraft in some state, then there's the flight director which maintains the flight path, and you can connect or disconnect the two at will. When connected the director can change the autopilot state.
To use the flight director you must fully specify your flight. The weight, the fuel, the weather, expected winds, takeoff and landing runway length, runway conditions, expected brake demand, as well as every single waypoint you're going to cross and the expected state at that crossing.
> what the human metaphor means
We learned after high levels of cockpit automation that maintaining situational awareness was still required. Pilots are freed of some stress during high workload portions of the flight, provided they planned correctly in advance, and that zero changes to their flight plan (not likely) occur.
As a result pilots are mostly told and trained to hand fly the plane during take offs and landings if the weather allows for it. You should only use high levels of automation if the situation demands it.
The point isn’t whether the aviation term "autopilot" has a technically defensible meaning, but that "Autopilot" predictably made normal people think "the car drives itself", just like "AI" now makes non-technical CEOs think "AGI employee replacement machine".
You are arguing the dictionary while I’m arguing the predictable, if not very well calculated and purposeful, misunderstanding.
and I don't know what worries me more - a burst in this bubble (and maybe some other tech stocks), or a failure of these valuations to be burst somehow, and even more concentration of capital and power around those corporations.
If the excavators only worked half the time, sometimes goes out of control bonking someone in the head and costs a billion dollars, then yeah they're pretty bad CEOs
> The problem tends to show up when a CEO is handed an agentic tool like Claude Code, and has it create something, which will work just fine, and thinks “oh, wait, why do we need so many people, when I can just sit here and make things work?”
> This is a bad CEO.
There is one and only one measure of whether a CEO is good or bad:
Does the CEO keep the majority of shareholders happy?
Since they are more often than not kept happy with money, if the AI makes the CEO ask the question above and the result is a larger return on the shareholders' investment, then that is a good CEO.
When your domain of knowledge considers Jack Welch to be a genius, there is no floor.
I think it's more of a workforce reduction technique in some areas. how much is the issue, and lots of "it depends" but even at a higher up-front cost, LLMs have higher value and less liability. Instead of offshoring/outsourcing for example, you can have the same local devs that were leading/managing the offshore teams manage LLMs instead. Or if you have 10 expensive devs making 200k, just halving that is a million/year. even if that million is spent on tokens (even double or triple that) it might not be that much outside of what they would have been willing to pay for more devs anyways if there was the need, except now they have to deal with less people to manage overall. less lawyers, less hr, less lawsuits, less hiring and promption costs, less insurance premiums. Even if the quality of the work is a whole lot less, they only care about a viable product being shipped, and how much the lower quality affects profit margins.
I'll say this, laws and regulation are sorely needed. all this hate against billionaires, ceos, ai-bros, whatever... might or might not be warranted, but it is fruitless. Redirect this energy to your law makers.
In China for example, they made it illegal to use AI for the sole purpose of replacing human workers. The CEOs aren't bad CEOs, they aren't great either, it depends on the outcome. There are scenarios where entire job roles can be replaced by LLMs, but for complex roles like developers, you always need some human devs still, but likewise, almost always less of them than before. However, although less devs might be needed to do the same work, in some cases LLMs open up possibilities that weren't there before, so more devs babysitting LLMs or working on designing/shipping more features is also a strong possibility.
I mean, companies aren't trying to simply save cost on employees, they also want to maximize profits. Less devs per feature isn't the same as less devs period.
Overall, I'll say that demand creates its own supply. The internet itself killed many job roles and reduced the number of people you needed for many others, but it's not like we have the huge unemployment many were afraid of decades ago, if anything it created lots of more jobs. LLMs simply can't do everything people can, so they're not a drop-in replacement, that alone should mean a lot in terms of supply/demand economics.
Most CEOs are not special. They are not especially smart, or skilled, or technical. The role self selects for sociopathy. That's not a quality that has any kind of linear relationship with intelligence. Quite the contrary.
A common misconception about AI is that it is intended to fully replace humans, which is incorrect. The purpose of AI is to reduce the need for human labor, and it has already been doing so. For example—though this is not an exact figure—a task that previously required 15 people might now only need 10. In no instance has the human element been completely replaced; rather, the reliance on manual labor has simply been reduced.
It's not exactly a misconception, when companies are pitching AI as a full and complete replacement for human employees. People are just reading the billboards on the side of the road.
I always found advertisements for AI to be so strange, why would you advertise your AI to the public as a danger for humanity that will also put everyone out of work? Such advertising would only appeal to sociopaths, but of course that's because it's intended to appeal to CEOs.
A "unit of work" that required X people to complete in Y time can now be done by X/Z people in Y time, where Z is whatever efficiency you are able to get out of applying AI tooling to your business.
For some companies, Z might be less than 1 though. ;-)
So you still need skilled people, just not the same amount as before, because you have different tools available to you.
This has happened before with other advancements in industrial/technological automation. It's not a new concept.
That sounds like 5 humans got replaced by AI. I don't think most people worry about whether all humans will be replaced, simply whether or not they will be replaced, or people they care about.
Or all 15 people are still employed doing the work of 22.5. Or even more people have been hired now that each person can generate 50% more value than they previously could. Or people are reallocated from the AI assisted task to another. Or some combination thereof.
Which is very short sighted. You or anyone close to you might not be replaced but it should be clear that you don't want to live in a society with 20% unemployment.
You'd think we would see a large spike in unemployment if AI was reducing the number of employees needed for jobs the way these CEO talk about AI replacing people...
Reminds me of the old joke "90% of the code is 90% of the work. The last 10% of the code is the other 90% of the work."
I have spent almost my entire adult life (since 1986) shipping products. One of the very first things that I learned, was that "shipping" > "designing".
There's so much work in delivering products that will carry your brand, and then must be supported.
I liken it to having children. Conceiving them is fun. Delivering them is painful. Raising them, is a lifetime of work.
In my experience, the same type of thing applies to products that we ship (and charge money for).
Beautifully said. Engineering is often just a cost center and I often have the feeling management is suspicious that the engineers are just wasting time and throwing up roadblocks for nothing. This in turn makes managers always on the lookout for "the shortcut" to cut out as many of those engineers as possible.
There is a definite lack of appreciation for the often repetitive grit, toil and maintenance work required to just have profit generating working software running reliably in production.
> There's so much work in delivering products that will carry your brand, and then must be supported.
People think otherwise with AI partly because Anthropic kept telling us that they didn't have to write code or review code any more for most of their work. Their agent swarms just comb through their github, slack and wikis to figure out what to do next, and another swarm of agents just review, test, merge, deploy, A/B test, and revert the code. Boris alone merged nearly 300 PRs in the past week (or two?). So the top research labs seem have broken the productivity seal.
And then they talk about this recursively self-improving AI that is so powerful, so autonomous that they advocate that every company should be prepared to "pause" the effort. And their Fable/Mythos has this specific restriction as mentioned in their model card[1] that they are going to reject requests to tune and train models because, well you guess it, the models are too powerful to be used by mere mortals.
[1] We’ve implemented new interventions that limit Claude’s effectiveness for requests targeting frontier LLM development (for example, on building pretraining pipelines, distributed training infrastructure, or ML accelerator design). Using Claude to develop competing models already violates our Terms of Service, but enforcing this restriction through our safeguards avoids accelerating the actors most willing to violate these terms. Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts, these safeguards will not be visible to the user. Fable 5 will not fall back to a different model. Instead, the safeguards will limit effectiveness through methods such as prompt modification, steering vectors, or parameter-efficient fine-tuning (PEFT).
I think taking Anthropic or any company in this space at face value is naive at best though. AGI has been 6 months away for years now. Surely anyone can think this through: Anthropic knows what theyre doing with their public facing repositories, they know to make things enabled by their tech seem impressive. I would consider Bun etc. examples of this.
Realistically, nobody intellectually honest really knows.
People think otherwise with AI partly because Anthropic kept telling us that they didn't have to write code or review code any more for most of their work.
Even if that were 100% true, it only collapses the coding effort to near zero. Anyone who's built and shipped a real product should know that coding is maybe 50% of the work, and on a mature product it can be much less.
even boris says they need people with judgment to manage the agents
i dont write code by hand anymore but shipping something people want is as hard (or maybe harder?) as its ever been
Boris also says he stops using /plan, he writes loop to write prompt, and he simply asks AI to come up with solutions. He also said many times that his agents would comb their emails, slack channels, and Github issues to come up with things to do. When we combine what he has said, it's hard not to have the impression that he was implying full autonomy of their agents. The only that the engineers need to do is to build harness and to issue approvals, rejections, or suggestions.
I work on a toy project that has exactly one user (me). On its face it's fairly simple. It's a portal to my media server because I didn't like how Plex worked with regards to searching and filtering. I can look for movies or series by director, studio, publisher, etc. I can rate things, I can find highly rated things. It's great, and instead of bugging plex support to add new features, I just tell Deepseek to do it. I started it before LLms were prevalent and now that I have open code I've had Deepseek write and rewrite most of my code and implement new features.
But even with this toy project, and the target market being someone I should know very well (me), I often struggle to figure out what I want the app to do. When I go through brainstorming or grilling sessions it'll often ask me a question about how the product ought to work and I'm just like ¯\_(ツ)_/¯ give me suggestions and I'll let you know.
Genuine creativity is something LLMs struggle with and it kind of makes sense given their design. If you have a complete plan for a feature or even just an idea what the feature should do, that is enough for an LLM. But asking it to think and come up with a new feature idea by itself will always yield mostly extereme basic things you've already thought of. That creativity of "what" to build so it serves a purpose is still very difficult imo and LLMs are not good at it.
> People think otherwise with AI partly because Anthropic kept telling us that they didn't have to write code or review code any more for most of their work. Their agent swarms just comb through their github, slack and wikis to figure out what to do next, and another swarm of agents just review, test, merge, deploy, A/B test, and revert the code. Boris alone merged nearly 300 PRs in the past week (or two?).
Apart from many other issues with this, heavily subsidized subscription plans won't last forever, and if you start burning your own money on tokens in this way, you'll soon realize it's terribly inefficient.
I’ve been wondering if “you’re not google” when learning about googles software dev process applies to Anthropic. Anthropic is a company that A. Has cheap unlimited access to its models and B. Is probably largely insulated from the types of tradeoffs that the rest of industry has had to observe in the post-ZIRP era.
Like did they break through the productivity seal? Or are they willing to spend that much more on it since they see their failure as a like existential threat to humanity. I doubt it our boss sees your software the same way.
It doesn't need to be an existential threat to humanity - it's an existential threat to their business. They need agentic workflows to work for their business to become profitable. So pouring money into the "no engineers write code anymore, only agents" model is at once R&D, QA, product development, and advertising. They can spend as much of their investors' money on this as they have to because if they can't (sustainably) sell this vision to other companies, their company collapses.
What is post-ZIRP please :-) ?
Zero interest rate policy. When interest rates are Near zero you can spend money like it’s free. A lot of what we thought of as like normal engineering culture were the result of interest rates being zero.
Tah for that.
> the safeguards will limit effectiveness through methods such as prompt modification, steering vectors, or parameter-efficient fine-tuning (PEFT)
Holy crap that is dark. I like learning about ML for fun, and now I have to assume that their model is intentionally misinforming me to sabotage my learning? It is absolutely bananas that somebody decided that was ok behavior.
time to support open source and local models
I don’t see how that helps, unless you actually mean open source, rather than open weights like most people do. Without everything that goes into the model, including training data, these things are opaque.
Actual open source is hard without a big war chest that allows you to flagrantly steal the training data.
That may very well be the case. In fact, I'm nearly certain that you're right. But it doesn't change the fact that open weight models are altogether insufficient on a number of important dimensions regarding freedom and transparency. And so often (such as the comment I replied to, I think), even technical people seem to just ignore the difference. Open weights are just weights. No amount of open-washing changes that.
The raw training data is so large that very few parties could host it for free even if there weren't copyright barriers.
But I think you could have a full open source training software pipeline that's set up to work with Wikipedia, Common Crawl, Books3, Library Genesis, Anna's Archive, and whatever other useful data sets people can name. There would just be a step where you have to provide your own copy of Library Genesis (or whatever subset of it you have managed to obtain).
Someone could write a cyberpunk Three Body Problem with this premise.
They kinda did (though it's more inspired by Trusting Trust than AI)
https://corecursive.com/coding-machines-with-don-and-krystal...
TLDR :-)
This comment is not entirely on point with your comment, it circles around and above it looking for lift though.
If you're not doing work that requires your code to stay in home nation data centres, Claude for Deepseek, Deepclaude (https://github.com/aattaran/deepclaude) is a great way to get better at using Claude like tools for software development. It even does a pretty good job of putting together cover letters for job applications...
Using Deepclaude is very much cheaper than using claude... For hobby projects, I've found it useful. A recipe (for cooking) management app I've made took a couple of hours to put together and cost $US 0.5. Claude is far more expensive.
The downsides of Deepclaude for many are:-
- DeepSeek is a Chinese corporation so the Chinese Communist Party may ask for data if it wants it.
- DeepClaude isn't as fast as normal Claude, though it's still pretty fast and I think fast enough (YMMV).
- DeepClaude might not be as optimised for various code issues that Claude may be able to solve more quickly or effectively.
- The same safeguards are probably on DeepSeek, but you won't be "wasting" as much money as you might on using Claude.
Inference focused hardware (https://www.youtube.com/watch?v=nvPqHoVSenE, AI generated speech) may in the medium future cause a large enough cost/energy reduction for LLM tools like Claude to make local LLMs more attractive.
Inference focused hardware would make running Open Source models like DeepSeek on local machines far cheaper and control over safeguards would return to the end user.
Hopefully this leads to a localised LLM provision market where local businesses provide varieties of these "local" LLM services. Here, local could mean on premise through to state or nationally based LLM services. Eventually, government orgs outside of the US may demand this kind of LLM use, in the same way governments legally require data to be stored within national borders for many critical government functions.
A bloke can dream I guess...
...Could affordable inference focused hardware also cause the bottom to fall out of these stock market bending valuations for AI corps and their datacentre obsessions?... Not to mention the societal costs caused by the AI super corps building these data centres. At the moment, they're nearly making a profit... They seem almost like speculative companies... Is that a term?
Anthropic is full of shit.
Fully agree. Shipping a complete product with a functioning user acquisition funnel is much harder. It's like; you have to build the whole product first with lots of features and then you have to try to create a highly condensed overview of all those features to expose them all on the landing page.
If you can't make the visitor understand your entire complex product in 10 seconds, then you've lost them.
Your product has to be complex because that's where the software market is at. All of the low-hanging fruits have been taken by the time you identify them. Sure, someone will find a way to make money using new low-hanging fruits that arise due to technological changes but it's not going to be you. You probably don't have the business connections to make that work.
I'm not entirely sure how that dismisses the CEO's putative argument: they go big on AI precisely because shipping end-to-end is hard, so they think they shouldn't waste resources on tasks that can be automated.
The structure of a good argument would be something like: certain tasks are fundamentally human and impossible to automate (which and why?) and by pushing AI use beyond what is optimal you are actually hurting your employees ability to do those hard parts.
A weaker but still useful argument is that most everything can probably be automated, but frontier models aren't there yet.
I wouldn't say it "dismisses" their argument, but I think AI marketing encourages them to take an over-simplified view of what it takes to ship product. Most folks like a good, simple story, as opposed to the unvarnished truth.
> "There's always an easy solution to every human problem; Neat, plausible and wrong."
-- H. L. Mencken
It's like the classic scenario, where you lash-up a barely functional UI demo, and the manager cuts your development schedule by 90%, because you "already have it working." That taught me to never do a lash-up demo. If I show something to someone, it is ship-quality (but often incomplete). It's a technique that I've used for years, and is a great way to involve nontechnical stakeholders, without risking stuff like "it's already working."
All that said, I think that AI definitely could automate a lot of the repetitive stuff involved in shipping. It's just that the CEO would fire the folks that could teach it, before it can learn, because they think that what they do, is "unimportant."
I hate to use a throwaway, but this bit:
> with a functioning user acquisition funnel
How do you actually get this. I've got a product, the site is hand crafted, shows the complex product really well (and had good feedback on it) but how do I acquire the users?
It seems as the cost of creating software has plummeted, it's the actual sales side of it that's going to matter even more. I'm stuck at this point.
"How do I acquire users" is the entire function of sales and marketing. A single HN comment explaining how to do sales and marketing, which is highly dependent on your product and market (and much more difficult than technical people tend to believe), is a bit unrealistic. And a great opportunity to use Claude/ChatGPT for something other than code. There's no silver bullet but as a springboard you can think about:
Who is your ideal customer profile (look up buyer personas) -- if you're B2B figure out both the profile of the company who would buy, as well as the person who would actually buy, and the person who would actually use the software: remember that buyer != user in B2B scenarios, and you'll have to figure out if the buyer, user or both is the best path to getting a sale. If you're B2C figure out your buyer personas so you know where to advertise.
Why would people want your product; sounds like you may already have this down but be ready to explain your value proposition concisely.
How will these people hear about your product -- a SaaS that falls in the woods doesn't make a sound, you need people to learn your product exists before they can pay for it. This is the point of figuring out buyer personas, you need to meet your customers where they are, and you can't know where they are unless you know who they are. This is highly dependent on your product/personas, and could range from running LinkedIn ads to SEO to having a Bluesky brand account to going to local meetup groups or conferences and trying to get your first handful of users in-person.
Get a dozen users word of mouth? They will tell friends? Won’t scale forever but it gets you going.
Sorry to burst your bubble but the cost of creating software has not, bloatware definitely has.
I like this analogy; raising children well like delivering products well pays dividends. They’re less likely to cause problems and if they do, they tend to be smaller in scope.
> 90% of the code is 90% of the work. The last 10% of the code is the other 90% of the work.
Don't think I've heard that one but certainly rings true to my experience.
Reminds me of "ninety percent of the game is half mental"
I've heard it as "once you think you're 90% done, you're really halfway done."
Tangential: it's always made me wonder about teams that believe "80% effort" is optimal.
Great analogy all the way through. Also the last 10% takes thre most effort/iteration to get the work done such that you don't spend a lot of time maintaining it later.
I skimmed the article, guilty, but I think what I got from it is that CEOs will CEO? No disrespect meant, I’ve seen your name here often and thoroughly enjoy the folklore that you share, but I don’t understand what context you reacted to. Cheers.
The context that they think that shipping is simple. Shipping (what you need all those annoying peons for) is really terribly difficult, and has a lot of moving parts that designers often fail to take into account, until the deployment people lock them into a restroom stall, and refuse to let them out, unless they listen.
That's common with newer engineers (and now, non-engineers). I believe that Mr. Dunning, and Mr. Kruger had something to say about it.
I also spent most of my career at hardware-oriented companies, and shipping hardware is orders of magnitude more difficult than shipping software.
Thank you. You spent some time at Apple, no? There is that “real artists ship” or “great artists steal”, but I wonder what he’d say now. Just fun to think about.
Not as an employee, but I've been an Apple developer since '86. Had fairly intimate relationships with them, at various points in my career.
> Conceiving them is fun. Delivering them is painful. Raising them, is a lifetime of work.
Then there's the technical debt!
Shipping is frankly the easy part. It's the operating overhead that often breaks you.
I liken it to free puppies.
This is true.
I have always prided myself on writing concise, high-Quality code, because it tends to be quite debt-free.
So far, LLMs seem to deliver code with "Louie Da Loan Shark"-levels of tech debt.
The problem is that the lines of code are riding on a stack of other dependencies that all need care and feeding. Things reach EOL. Frameworks have major breaking changes. CVEs are discovered.
You can actually get high-quality code out of them -- at least with Claude; not had a great experience with Gemini -- but for complex tasks requires riding them very, very hard and really understanding where things can go wrong and poking at them repeatedly. Iterate, iterate, iterate.
> Iterate, iterate, iterate.
That describes my last week. What made it most annoying, was the need to release through TestFlight, because the memory issues would not appear, when tethered. Also, I was checking in constantly, because I had to revert and reset the context, several times.
You mentioned in another part of this thread that you worked in hardware mostly?
It seems like the cost of changing hardware code is high enough to still insist on building it high quality, is that accurate?
Yes. It's also why working as a software (host) developer at a hardware company is difficult.
Hardware people insist on treating software the same as firmware.
Bad firmware can cause real-world, physical damage, and be impossible to fix without a hardware recall. A firmware bug can wipe out a hardware company. A software bug can be embarassing, but can also be corrected a lot more easily (as long as it is being treated differently from firmware).
> So far, LLMs seem to deliver code with "Louie Da Loan Shark"-levels of tech debt.
Maybe a couple of years ago, but these days, Opus 4.8 is frankly writing better software than what I've seen over the previous decades in non-tech enterprise. These previous two months, we've replaced so much technical debt we've been dragging along for the previous 5 years as our team went from 25 to 3 people.
This is in non-tech enterprise in Denmark and AI had absolutely no impact on us going from 25 to 3. That was all Putin and bad business decisions on the c-levels. Like keeping flexible loans to fund projects on the books when the interests rates were 0.01% because they might go to 0.001%. Anyway, I'm getting to the point where the AI does 100% of the work, but only if it's piloted by people who know what security, resource consumption and compliance is. The code itself is excellent though.
> Maybe a couple of years ago, but these days, Opus 4.8...
No. Opus 4.8 still writes bad code, just like every other time AI boosters have claimed "but the newest models are really good".
Every line of code is a liability
Yup. In another post, I was grumping about having to accept a truly obese bunch of code from an LLM. This particular issue is the only one I could imagine even considering accepting that much pasta, but I sort of have to, because the LLM was able to quickly solve a problem that would have taken me a couple more weeks to address.
I remember a .sig that went something along the lines of:
Amen. Or I should say RIP.
> I liken it to having children. Conceiving them is fun. Delivering them is painful. Raising them, is a lifetime of work.
I am not a children person. But I love this analogy.
To deliver something nice, we also must accept some suffering.
In various different projects I've been involved in where we've been implementing (not developing) software solutions I've noticed that, at management level, there is little regard given to the level of maintenance required of running the software; that people are still needed; that, yes, processes are automated, but there's a helluva lot of ongoing work required to ensure that new data won't pop the automation off the rails.
It's as if the installation part is the hard bit, and after that it'll take care of itself for ... far enough into the future that it won't be <current manager>'s problem. It is solved.
... and that's just using the system, not fixing bugs and adding features.
%80 of CEOs are Meh. %10 of the CEOs add value to the company. %10 of the CEOs are actually detrimental. The %80 will always try and do what they think the top %10 CEOs do because of FOMO. The bottom %10 will do it because they know their days are numbered and hope that it might put them in the %80 keeping their jobs. A good LLM could probably replace most and not do any worse.
Good CEOs don't see their employees as an expense.
Those percent estimates look about right :0
Good CEO's make such good money on their employees that everybody gets raises and bonuses, the company grows responsibly, and the stock is a good investment.
Too bad it's out of reach for so many executives.
If that's too challenging, I understand, but if they had real confidence as a business operator I don't see why so many would be kicking out anybody over AI when they could at least continue to make the same money off the same people going forward. OTOH in cases where AI is almost ideally helpful it should be no surprise if hiring is slowed, and doing the accounting it could very well add up about the same either way. But one way clearly indicates the limited vision of a lesser leader, why settle for that?
Two of the most macro giveaway characteristics are emotionalism and superstition.
Not just for CEO's and other executives, but anyone in a leadership position or with decision-making tasks to perform.
One of the legendary combinations is when superstition is used in place of technology, and emotional reactions completely prevail instead of genuine business acumen.
It's a pretty good estimate that almost every CEO who thinks it would be good if AI replaced their employees, that these CEO's fit squarely in the superstitious camp.
I would say that's just one growing subset of a much larger smorgasbord of superstitions to choose from, and some big-shots are bound to indulge a whole lot more than others :\
Honestly I am burned out on public discussion of AI. There are so many hot, low info takes on both sides. All the dumb stuff revolves around this imbecilic notion that AI will take your job.
In that sense Altman and Dario et. al. were extremely successful in their cringeworthy campaign to establish themselves as priests of a new machine god religion. Even the people who don't want it believed them.
The good news is at least this year companies are starting to get a little more thoughtful about why they're paying for AI and what specific business function it serves.
Realities:
1. AI is a tool. You don't replace a job with a tool, just like you don't replace an apple with a sock. It was always an error of classification to think this way.
2. AI is a useful tool. For every CEO who thinks it justifies mass layoffs, there are dozens of people who don't want to admit that it does have a lot of utility. Anyone who isn't figuring out where this tool can make them more productive will have a hard time down the line imho.
3. We can infer from the priors that jobs will continue to be a thing, just like they still were after the inventions of printing presses, cars, power looms, etc. but there will be some pretty massive churn, some roles maybe disappearing entirely, new ones getting created, same goes for skill sets.
4. Businesses will use this churn to the best of their ability to either reduce costs or increase output. That second part is the key the AI haters don't seem to recognize. Thanks to competition, not every business wants to just fire everyone. Yeah, America's financialized and monopolized and less competitive than it used to be, but still, plenty of businesses will repurpose their budgets to produce more and expand.
So in terms of real specific and concrete things I've seen so far.
If you're in customer support that's a pretty rough spot, a well designed RAG system can turn 20 minutes of research into 20 seconds. Customer support budgets flow downstream from actual usage/customer base so if they can do more with less people will get laid off, they don't expand scope, they automate and cut the budget.
If you're a developer, your position's a lot more secure than that, coding agents are pretty incredible but they are simply not at a point where they can think through architectural decisions, and they occasionally go off the rails and trash everything. These things need to be operated and steered. Yes that's 100% the future of the profession and I'm sorry if someone doesn't like that, being a blacksmith who forges metal things by hand is also very niche these days, kind of sad if you loved blacksmithing but it happens. Devs also have to be aware that a PM/product owner can likely do some of their job now and I think any dev whose preference was to avoid thinking about customer or business requirements is going to find his utility is shrinking. A mass shrinking of the profession is unlikely but things will change and the only rational paths forward honestly are to retire or to figure out now how you fit in, it's a career opportunity if you get ahead of the curve even.
Lastly there is a whole other domain now which for lack of a good term I'll refer to as "prompt engineering," it requires some level of systems design thinking, but basically no programming expertise. The best candidate for this work is a person who possesses business process knowledge as well as systems oriented thinking. Maybe these jobs will end up finding a home in the IT department or something. As an economy we're barely recognizing them yet but I see that in engineering we can increasingly implement places for a prompt input, and then hand some workflow/business domain decisions off to someone who understands them better and they just tweak their prompts over time to get a great system. Pretty sure new jobs will be created and formalized around this over time.
I think this is the most level-headed take I've read in a while, sums up perfectly the HN discourse on "AI taking our jobs", thanks!
Agreed.
It is a fairly good summary of my PoV.
Thanks guys!
It's happening! We can see it! The future is somewhat legible and in it we don't all die! (Well we do, but mostly for the usual reasons)
We just gotta keep getting out of bed and making a smart choice or two between now and then!
now it's closer to 95% of work can be done by AI and requires 5% mental effort, but 5% of the work requires 95% of the mental effort to finish because of all the unoptimial decisions AI has taken. I find that AI works best in small micro-service type architecture where each component has a clear goal and doesn't have interconnected parts within the same application that can break. But you do run into an issue where changes in microservice a need changes in microservice b and updating it is not ideal since it usually cascades thru the entire system or requires stacks of legacy support.
IME it’s possible to have good clear APIs, limited scopes/goals, etc in a normal (macro?) service. But it requires a level of discipline and process many teams are unwilling to engage in.
There are a lot of bad CEOs, though. It's a lot like a politician -- it's quite difficult to become a CEO, and the skills to make it to that position don't always intersect nicely with the skills necessary to actually do the job well.
CEOs do get there with lots of politics in almost all cases. It’s all about who’s ass you kiss and who’s ass you don’t and if you’re lucky with timing things might just fall into place.
I think it’s exceedingly rare that a CEO is actually competent at their job. In most cases it’s the labor class propping the company up, and in some cases the workers are doing so against the wishes of the CEO. Not that executives want to ruin the company, they’re just incompetent and therefore make terrible decisions constantly.
> It’s all about who’s ass you kiss and who’s ass you don’t ... I think it’s exceedingly rare that a CEO is actually competent at their job.
But... that is kinda the job? CEOs are, first and foremost, the public face of the company. They're the one who talk to VCs / banks, regulators, major customers, the press. They're very highly paid PR reps / fall guys that shield everyone else, including the board of directors and all the VPs and SVPs, if something goes wrong.
For most companies, especially large companies, it's not important for a CEO to be good with software engineering, business development, etc. That, at least in principle, can be handled by other parts of the hierarchy.
> it's not important for a CEO to be good with software engineering
If you are the CEO of a company, you should have expertise in whatever your company does, and If your company is primarily a software company, then you should have expertise in software engineering. You cannot effectively manage something that you don't understand.
> CEOs do get there with lots of politics in almost all cases.
I know this can be hard for engineers to sometimes accept, but relationships (aka politics) are a key part of business. Rarely is one technical solution absolutely superior to another, making purchasing decisions come down to relationships.
Politics is also about compromise and managing a bunch of differing opinions/desires, which is one of the key skills of a CEO.
Knowing which ass to kiss at the right time is an important skill not everyone has.
Kissing ass: $1
Knowing which ass(es) to kiss when: $9,999,999
And that's how CEOs justify their exorbitant compensation
I wonder if business schools could ever start actually teaching this skill. So far they just largely operate as scams, held up purely by prestige and network value
The "network value" of a business school could be reinterpreted as "can the business school introduce you to some of the right people's asses to kiss"
not a skill i'm interested in, lol..
I’ve worked with CEOs in multiple large companies. I wouldn’t wish that job on my worst enemy. Nonetheless, someone needs to do that job and the intersection of difficulty and masochism is beyond what most people can do or endure. Many people try and fail. Their job, at the end of the day, is to eat an endless stream of shit sandwiches with a smile and a plan.
Much of the “competency” of a CEO in practice is to be able to accept the relentless drama and abuse without turning into an emotional wreck. Yeah, they have to make decisions, but that isn’t the part that makes the job difficult. That role takes an insane toll on the human spirit, and very few can do it for any length of time.
The cush job is often being CEO adjacent. You get most of the perks but also avoid most of the emotional abuse and drama.
This feels too soft. Each of these things has truth in it, but isn't some of this self-created? Where are these shit sandwiches coming from? A lot of these problems are the result of overpromising, breaking rules and skirting regulations, underestimating the difficulty of things they have no expertise with, asking people to solve problems with no resources, hiring more chefs rather than more cooks and dishwashers, of mismatches between good profitable product and exec exit. The idea that it makes sense for the CEO's (or really any leader's) core competency to be absorbing drama and pain is something we should think more about. Sometimes you hear that a good manager blocks and shields for their team, you have to wonder why the team always needs so much protection from their own company and processes.
Thank you for busting the myth of worker self-management.
> Much of the “competency” of a CEO in practice is to be able to accept the relentless drama and abuse without turning into an emotional wreck.
So it self selects for sociopaths. Good to know
You’re making the case for worker-owned cooperatives. Love it — we need more of them!
I'm very sympathetic to cooperatives, have traveled/know the Mondragon people (largest coop federation), etc.
However, I think there's a reason why coops seem to succeed at smaller scales, but there are essentially no large innovative coops.
There are a few large boring coops, and some small innovative ones, but seemingly something is making the CEO/investor board model the one large innovative companies are all using.
I suspect that it's both (1) access to capital is far harder for coops, and (2) that workplace democracy and hardcore mission focus aren't fully compatible. That is, "you cannot serve two masters" without losing focus on one of them.
If a company accumulates capital, it becomes vulnerable to the principal agent problem, and coops are way more vulnerable here than centralized companies.
If a company doesn't accumulate capital, it doesn't scale in complexity. It can grow by having more people do more of the same things, but it can't move into markets that demand anything complex.
This seems hard to tease out from the fact that a) the majority of companies do not survive, b) the large, large majority of companies that do survive do not wind up being large or innovative, and c) there are far fewer coops than regular companies. If you assume equal chance of success between them, you’d still see vanishingly small numbers of large, innovative coops, because a small percentage of a small number is small.
Coops can definitely succeed and even dominate at large scales with some minimum government protections. The largest dairy producer in the world is a coop: https://en.wikipedia.org/wiki/Amul
The problem is that knowing the right people to get investment does seem to have utility coops struggle to get, I think? maybe CEOs are basically like producers on movies who are just there to network for you.
What are the concrete benefits?
Do they tend to make greater revenue or profits? Pay higher wages and offer greater benefits to employees?
Coops tend to have better aligned incentives for employees on every step of the ladder. They'll tend to be more conservative about R&D but ensure that money that's being spent is being productive for the continuing health of the company since instead of that budget being "corporate's money pile" it's your potential profit share.
I think there's also a tendency towards longer tenure and higher value employees due to the investment in the company's future being a sort of central tenant of their attractiveness.
Any data to support those claims?
Not usually the kind of data that the greediest capitalists would appreciate.
There are some case studies here, but it's only one professor in a largely non-capitalist approach overall:
https://www.reddit.com/r/Cooperative/comments/1bm5s5s/richar...
Generally, yes. Also the gap between the employee and executive class is a lot smaller instead of unnaturally inflated like it is in most private equity companies.
In software I can imagine a worker-owned consultancy, but not a product company. It would imply staying in one place working on one product for your whole life, which doesn't sound inspiring
A company need not be a single product, and working in a worker-owned cooperative need not be a lifetime commitment to a single firm (though cooperatives ideally will have less turnover than firms owned by capital separated from labor.)
No, it implies that you give workers the means to dictate the direction of the company. That is what workplace democracy does:
https://en.wikipedia.org/wiki/Workplace_democracy
Acting like centrally planned dictatorships is a good form of collaboration is just so off base. There's no reason to think that introducing democracy into the work place wouldn't immediately benefit both workers + customers.
If this sounds crazy the C suite + board already vote on who gets hired into the executive team, vote for the direction of the company, and vote for their compensation packages (hint, they never decrease them).
Why shouldn't workers be legally enabled to do the same? What is the justification to this? I'm curious to hear it because the only way people can justify the current system is declaring that some people are actually more deserving of prestige, money, and benefits while others deserve to suffer.
With income inequality increasing, healthcare outcomes worsening, and children literally becoming stupider isn't it time to question the current system and ask ourselves if this is the society we truly want?
Where’s your real world evidence of all these benefits of coops?
Because I would love for it to be true.
Well workplace democracy has only been tried in a few corporations. If you want an interesting business case look into Semco Partners in Brazil:
https://en.wikipedia.org/wiki/Ricardo_Semler#Semco_1990%E2%8...
There is academic research on this too if you're curious but it's mostly in English, Spanish, and Portuguese.
But yes, there isn't much "evidence" because this system hasn't been tried en masse; however if you look at our current neoliberal hellscape, it's pretty hard to imagine it doing worse. Also neoliberalism wasn't really "tried" either, it was thrusted upon us by a group of individuals that wanted it.
One thing to keep in mind is that society can change quite quickly if you want it to. I'm sure the children that died working in factories during the 1800s never imagined such a society where children are valued, cared for, educated, and protected but it did happen.
It has happen before and it can happen again. It only happened because people were willing to fight for it.
The rules are allowed to be changed at anytime if we deem so, a better world is possible.
“We are sure it will work because it’s never been tried!”
I believe the Germans have had success with including labor representatives on corporate boards. Maybe we can start there.
This is a shallow dismissal of GP’s point. The point is more, “we aren’t sure it won’t work because it has never been tried,” which is much less of a straw man to argue with.
No.
There was an unequivocal claim that it will work better than our current system.
They said
“Acting like centrally planned dictatorships is a good form of collaboration is just so off base. There's no reason to think that introducing democracy into the work place wouldn't immediately benefit both workers + customers.”
That sounds more like there’s no evidence that it won’t work than an unequivocal claim that it will.
Well... there is historical evidence that centrally planned dictatorships are not a very responsive form of government.
Now, corporations usually have the problem of competition, so if they aren't responsive (or at least responsive enough), they get out-competed by those that are. Is that enough to make them different from governments? Perhaps, but I don't know.
If you don't think there are competitions in dictatorships you are extremely sheltered. The competition in a dictatorship is whether you stay alive or not, just like in a corporation is whether you become homeless and die or keep a roof over your head.
That's just neoliberalism baby!
Could you quote that? I don’t see it.
Why does this argument never apply to neoliberalism?
That was never put to a vote but it still thrusted upon a country where the results are what you would expect: the worse income inequality ever seen in the history of the nation, life expectancy has increased, deaths of despair have reached record highs, more children go to bed hungry, healthcare is being ripped from civilians, and corporations are legally allowed to poison and kill civilians (health insurance companies with their death panels, manufacturers causing cancer valleys) with zero legal repercussions.
So yeah maybe we should actually go the extreme into the other direction, if democracy is good enough to lead nations it's good enough to run businesses. If you're a worker IDK how you would argue otherwise. Being able to keep your boss/leadership accountable by voting for them out seems like a win for every workplace metric imaginable.
Imagine how better of a company Meta would be if Zuckerberg wasn't allowed to waste billions accomplishing nothing. In a just society he would have been voted out, but in a neoliberal society he is granted an insurmountable amount of wealth.
I'm sorry but this society sucks and acting like we can't do better, be better is beyond pathetic.
Don’t you still need someone to make high level decisions?
Worker owned cooperatives have a variety of ways of doing this. Voting directly, electing people, etc. The main difference is that the cooperative typically doesn't buy the myth that the person making the high level decision needs to be paid 1000x the workers.
No, look at functioning democracies. They don't need authoritarian rulers, only those that want to be authoritarians argue elsewise.
> They don't need authoritarian rulers
Yet they seem still inclined to elect them.
Maybe, but not necessarily for this reason. Even in a worker-owned coop, someone sets the overall direction. And how is that person going to be selected? It's still going to be largely politics.
Politics among the workers vs politics among the shareholders.
So, we live in the economy where everything is done by labour that receive minimal acceptable wages, while all the profits are collected by the ruling class who do nothing except constantly kissing each other asses.
> it's quite difficult to become a CEO
on the contrary, it seems to be one of the few jobs that seems to require absolutely no qualifications to have.
What you need to do to be CEO is.... convince someone to lend you money in the hope that you'll get it back to them.
I've worked under some absolutely awful people who wouldn't pass an interview anywhere, but somehow they're CEOs, because they can smarm there way into more money consistently.
>convince someone to lend you money in the hope that you'll get it back to them
And it should be noted that many of these people lending money are in a similar situation of not being required to have any qualifications. Sure, some of them have worked their way up through sound investment after sound investment, but many of them were either born into their position or simply got lucky at some point along the way. Just think of all the money investors threw away pursuing crypto and NFTs for example. Many of those investments were transparently stupid from day 1.
To be fair, raising money takes a certain skill, that few people possess; and in many cases, it’s essential for a startup to even exist.
> To be fair, raising money takes connections, that few people possess; and in many cases, it’s essential for a startup to even exist.
FTFY
>> To be fair, building connections takes skill, that few people possess; and in many cases, it’s essential for a startup to even exist.
FTFY
Often, they are good at taking things, keeping things, misdirecting and setting boundaries (especially communication boundaries). They are good at keeping their positions.
This is a broad range of skills and to actually be a CEO, you need to really hone these skills and be among the very top. To be good at those, just enough to qualify for a modest CEO role at a small start-up, you generally don't have the time to be good at anything else.
Saying that you don't need any skills is mischaracterizing it. You don't need any value-creating skills, yes, but you need significant value-capturing skills.
I can imagine a world were all companies become empty of workers and only executives remain and they would just have meetings with each other while they starve and would explain it away as a new diet they're on. There would be no petrol and they would be forced to walk to work and would say that it's their new fitness routine... And they would all believe each other.
What skills? I've met several.
Most of them got into a prestigious school on legacy, paid for by wealthy parents. Many were above average IQ, but by no means geniuses. They had access to computers earlier than others, due to said affluence. They seem unable or unwilling to comprehend they're overwhelmingly on average, "nepo babies" to steal a term from the world of entertainment.
I think it's private schools in general. Even those from second and third tier ones, which can filter more by means than the elite ones, find themselves atop companies. It's the natural access and the natural ability to socialize with other private school personalities. Their definition of capable leader is a particular type of leader. They can make and take jokes, but particular types of jokes. They hide each other's shortcomings, insecurities, guilt whereas people from other backgrounds, even people they like and think highly of, tend to serve as a mirror.
Getting funding is a value add, but I agree calling it "skill" misses most of what makes someone "good" at it. We've built things to overwhelmingly rely on funding gated by other private school people though. It would be nice if we could have that person with access pitching without them also being in charge of running a company, product development, or managing people. But then it would require the same of the investors. The investors would then need to evaluate products and ideas and markets. And the markets would have to reward that. Things would need to be different.
>I think it's private schools in general.
I think CS is a little unusual in that places like Berkley are ranked highly. (Interestingly, Harvard has a rather low CS rank and Zuck was in the process of transferring to psych because he can't handle anything quant, but due to FERPA no one will call him out even as he repeatedly steps outside the law)
>Getting funding is a value add, but I agree calling it "skill" misses most of what makes someone "good" at it. We've built things to overwhelmingly rely on funding gated by other private school people though.
I think part of the issue is that a big chunk of startups are straight up grift. My complaints about gatekeeping aside, I think if I had a legitimately good idea I could either submit to yCombinator or talk to contacts who've cashed in stock and get investors.
That's mostly due to the fact I'm well known for valuing authenticity though and my personal brand is such that folks would take my proposal seriously.
I met a LOT of people who went to places like Stanford who basically... they always meet the metrics, at the expense of all else.
Anyways, I think in general we try to generate too many "startups" when we're really striving for small businesses.
But because the funding is the truly difficult part, we hyperfocus on that.
Meanwhile, I've seen plenty of startup ideas that while benefiting from the Ycombinator social network, have initial costs such that you could hit up a few rich friends who've known you a while with a solid business plan and get bootstrapped... but that would require domain knowledge and the respect of your peers, two things many founders lack ;-)
It's not difficult at all to become one, and the work involved in being a CEO is not particularly difficult in comparison to senior technical work at all. The only thing that is harder about being a CEO is the responsibility. I'm sure being the CEO of Microsoft or whatever is plenty difficult and demanding in many ways, but most CEOs are not that, and speaking just from experience most CEOs and CTOs are clueless morons.
With that said, I've been programming for 25 years and I've only been a CEO for 3, so take what I said with a pinch of salt.
I do think people overestimate titles like this a lot, though, and it really comes down to what the company actually does and what is demanding for that company at that position/role. The CTO of a some-bullshit-as-a-service company may as well be straight out of college, because they're likely doing something trivial that literally anyone (including LLMs) could put together. The CTO of a well-used and reliable streaming service that handles a meaningful part of the world's Internet traffic is obviously solving a more interesting and demanding problem, and their decisions are going to be more important.
Years of ZIRP, QE, bailouts and stymulus money muddled the waters a lot. Add to this the Old Boys (and Girls now) Networks, a culture that values getting money fast as the ultimate value, the prevalence of politics and you end up getting a boatload of bad CEOs.
There was a time when I used to recommend "Out of the Crisis", a book from Demming, to business leaders.
Problem is, "Out of the crisis" still assumes as a premise that companies compete on the quality of their products, that making money comes from actually making and selling stuff. That leaders are not the anti-intelectual morons that believe that absolutely any thing can be explained with a 15 minutes deck, that math and statistics are passtimes for weird geeks that don't add "business value" and because of that, is a book that could have steered us toward a better world in the 80s, but now it is completely useless, because its recipes can't handle the level of degradation things goto into.
what about zukerberg he didnt have to do any politics to get to ceo. yet he is the face of ai layoffs and bad ceo.
What about him?
doesn't meanThe Winkelvoss twins would beg to differ.
He's the face of bad CEOs because people like to make up things about how bad he is. The Social Network, a major 2010 biopic about the early days of Facebook, famously cut his college sweetheart and now-wife out of the story in favor of a fabricated character arc involving an ex-girlfriend who does not exist.
Poor Marky Z. always getting a bad rap. Let s/he who hasn't made zillons getting billions of people hopelessly addicted to social media while facilitating a genocide or two and generally destroying the world order as we know it throw the first stone.
From your tone, it sounds like you may just be intending to warn me that it's cringe not to agree with any criticism of Zuckerberg? If that's so, I have to respectfully disagree; I think this is a bad attitude that leads towards being poorly informed about the world.
If you're interested in discussing the specific claims you're making, I really don't think that billions of people are hopelessly addicted to social media, and I would love to hear your basis for claiming this. My understanding (from e.g. https://www.nature.com/articles/s41598-025-27053-2) is that genuine compulsive use of social media is quite rare, and most people who describe themselves as "addicted" are just regular users who enjoy it but kinda feel like it's a waste of time.
https://gizmodo.com/meta-settles-lawsuit-that-claimed-social...
> Meta Settles Lawsuit That Claimed Social Media Addiction Screwed Up Schools
> On Thursday, Meta settled a lawsuit brought by a Kentucky school district that claimed the tech giant’s social media platforms have created a mental health crisis at its schools.
> The case was considered the first of its kind and a bellwether (a case that is representative of a large pool of lawsuits and will be a test for future litigation). The plaintiffs argue that social media platforms have had a major negative impact on the mental health of school-age children, which in turn has caused a burden on the education system, as American schools were forced to redirect resources to counter this problem.
> The settlement comes shortly after Meta lost a key bellwether social media addiction trial. Back in March, a judge in Los Angeles ruled that Meta was liable for the adverse mental health effects a now 20-year-old suffered after getting addicted to Instagram from an early age. The representatives of the young woman argued successfully that it was Meta’s deliberate design choices, like the infinite scroll and face-altering filters on stories, that had exacerbated her addiction and subsequent mental health issues like self-harm and depression.
---
https://en.wikipedia.org/wiki/Careless_People#Myanmar_genoci...
> The military junta in Myanmar was facilitated by Facebook to post hate speech that sought to foment sexual violence and promote genocide against the Rohingya. "Myanmar would have been a better place if Facebook had not arrived" Wynn-Williams writes.
> Wynn-Williams argued that Facebook failed to moderate hate speech against the Rohingya in Myanmar, including the use of the racial slur kalar. She noted that the company only had two Burmese language moderators, both based in Dublin, for the entire country, and claimed that one of the two moderators gave a pass to hate speech while removing pro-human rights content. She further claimed that she raised concerns that the moderator was "in cahoots with the" junta, only to have her concerns dismissed by the content team. Additionally, she claimed that her efforts to have Facebook's Community Standards rules translated into the Burmese language were resisted by the company communications team, who told her that "Myanmar isn’t a priority country" in the region.
---
https://en.wikipedia.org/wiki/Facebook%E2%80%93Cambridge_Ana...
> In the 2010s, personal data belonging to millions of Facebook users was collected by British consulting firm Cambridge Analytica for political advertising without informed consent.
> The data was collected through an app called "This Is Your Digital Life", developed by data scientist Aleksandr Kogan and his company Global Science Research in 2013. The app consisted of a series of questions to build psychological profiles on users, and collected the personal data of the users' Facebook friends via Facebook's Open Graph platform.[2] The app harvested the data of up to 87 million Facebook profiles. Cambridge Analytica used the data to analytically assist the 2016 presidential campaigns of Ted Cruz and Donald Trump.
> Other advertising agencies have been implementing various forms of psychological targeting for years and Facebook had patented a similar technology in 2012.
---
https://www.the-independent.com/tech/facebook-manipulated-us...
> Facebook manipulated the emotions of hundreds of thousands of its users, and found that they would pass on happy or sad emotions, it has said. The experiment, for which researchers did not gain specific consent, has provoked criticism from users with privacy and ethical concerns.
> For one week in 2012, Facebook skewed nearly 700,000 users’ news feeds to either be happier or sadder than normal. The experiment found that after the experiment was over users tended to post positive or negative comments according to the skew that was given to their news feed.
> The research has provoked distress because of the manipulation involved.
> Studies of real world networks show that what the researchers call ‘emotional contagion’ can be transferred through networks. But researchers say that the study is the first evidence that the effect can happen without direct interaction or nonverbal clues.
> Anyone who used the English version of Facebook automatically qualified for the experiment, the results of which were published earlier this month. Researchers analysed the words used in posts to automatically decide whether they were likely to be positive or negative, and shifted them up or down according to which group users fell into.
Yeah and Justin Timberlake was a co star. You’re really beefing about a movie almost 2 decades old?
It's on my mind because they're releasing a sequel in October, and Aaron Sorkin has not (as far as I've seen) acknowledged that much of the original was not true nor promised to be more accurate this time around. I'm pretty confident that it's going to be about the same mix as last time, and I'm going to have to go around saying "actually the scene where Zuckerberg did suchandsuch terrible thing wasn't real", and people will respond by insinuating that it's lame for me to care.
If you’re looking for an honest depiction of events out of Hollywood, you’re going to be persistently disappointed.
> it's quite difficult to become a CEO
It literally just requires filing an LLC or Corporation. There are several SaaS companies that will do it for you.
I saw someone on Xitter say "Any CEO who wants to replace jobs with AI should first have to replace their own assistant with AI" and I think that's the perfect rule. Every AI demo is some version of a personal assistant, surely AI can do that job right?
I think we'd get zero volunteers from CEOs who have assistants.
(Note: this is not meant to be an insult to human assistants! I think they do a valuable job and should not be replaced by AI either).
It's the eating your own dog food.
The Open AI guy said it's now better than doctors (he said it, not me lol). He's replaced his doctor, right?
What he said (https://economictimes.indiatimes.com/tech/artificial-intelli...) is that he doesn't want AI to replace his doctor even if it is better at diagnosis. He wants both and he expects most people to agree.
I donno
CEOs get signals from multiple channels that cantbe accessed by AI or ordinary workers
so that alone becomes a moat
same with sales jobs theres a plane that AI cannot observe (yet)
AI is the new outsourcing. It is cheap in the short term, but in the medium to long term it generates many problems, like loss of know-how and competitivness, huge maintenance cost, loss of control, and dependency on other foreign companies, with misaligned goals compared to your company.
Before using AI ask yourself: Would you outsource this task, with all the risk that come with it? If yes, go for it. But if no, then don‘t use AI.
A custom-built AI would be pretty good at replacing a CEO. Think of all the things a company could do if they reduced overhead by that much?
You don't need custom built anything, ChatGPT could generate corporate initiatives and PR statements all day and no one would notice.
100% been saying this for a while now. the main thing AI will be able to replace is a C suite.
maybe this is the way forward. Imagine how many tokens one could burn given C suite salaries.
I think it will handle communication between the C suit and the underlings.
it's kind of like saying like you could replace a king with an AI. the position is a relationship to power more than it is a function with a productive output.
a bad king and a bad CEO could be replaced with a spinning top with no loss in productivity (and maybe some gain).
I still remember when I used ChatGPT the first time to write an email. I thought to myself “Oh. This sounds like 99% of the corporate communication from above”. We were joking that corporate had BossGPT for years and just didn’t tell anybody.
It also made me realize that most the so called “creatives” in marketing and PR also just repeat variations of the same few templates. Not much real creativity there.
Our CEO bragged about him "just building stuff" now (vibe coding) while outsourcing his other tasks to AI.
I don't think that's the flex he thinks it is.
The hallucinations would be a feature in this case.
and people actully MIND if your CEO is a liar.
But AI is expected to lie^H^H^H hallucinate
You can get a lot of tokens for a CEO. I'd say it's worth a try.
I am building a project now and then will create an AI to manage it and be the CEO.
The code is human + AI, the management is only AI
Who takes responsibility when the AI does something unethical or illegal? Do we put the computer in jail? Or do we just look the other way like we do with human CEOs?
I think Grok on x.com is a reasonable case study. There’s the stuff we heard about and then there’s a separate category of things that weren’t newsworthy, and only the newsworthy things prompted changes to be made.
It sounds like you already know the answer.
To be fair here, CEO dont take responsibility and dont go to jail. In rare case they do, they will be pardoned.
Remember the printer in Office Space?
lol CEOs in jail.
Robotics aren't there yet, it needs to go on golf playdates with investors and board members.
Maybe a virtual / Zoom agent could allow the elderly investor to stay in the game. When you get too old to go golfing you can still stay in the game.
AI means mediocre. Mediocre companies are pretty bad to work in. So it would guarantee a soulless and pointless company imo
My charitable word would be average. AI means average. And an average CEO isn’t that bad. It’d be a deliberate trade in exchange for less spending on salaries. (just one very large salary)
https://replaceyourboss.ai
(discussed 6 months ago https://news.ycombinator.com/item?id=46072002)
While I agree that AIs would do a good job…
Would you rather take instructions from a ruthless robot or ruthless flesh sack?
Weirdly enough, I'd take the robot. At least we can pretend the robot doesn't know any better. The human is actively choosing to be a dick and profiting off it.
The flesh sack is choosing to be an insufferable twat, and the robot either doesn't have any choice or has a decent statistical justification for what it does.
A robot with access to all of the company data and mediocre decision making wouldn’t be terrible.
It's pretty disappointing that people on this site of all places have no idea what CEOs do. Many of them are certainly overpaid, and like any other profession, many are not good at their jobs, but they aren't sitting around drafting memos and coming up with deciding who to fire all day.
I reckon such AIs are already in place, but by proxy. There must be CEOs somewhere who have completely offloaded what meagre amount of thinking they needed to do to some bespoke AI setup while they LARP around convincing people they are the one running things.
I posited that one as a question.
Naive and stupid, but it was downvoted and flagged away into oblivion with zero chance for a conversation.
I do think it's more subtle. AI can replace very few jobs end to end with the same quality, maybe none. But AI can be put to work on high ROI problems. Now when the new marginal job is not obviously as high ROI as putting another 100k of tokens to work, no human gets hired.
Next, comes natural attrition in a company where a certain percentage will leave every year. Will they get replaced with a human or their budget goes into tokens?
Only when these 2 angles are exhausted, a typical company will start thinking about layoffs.
Now, some companies are already stressed: customer buy AI products instead of theirs, AI makes it easier to build what they offer, customers believe they can vibe code things. These companies will layoff first, because of AI. Not because AI will do the persons job but because the money gets spend differently.
AI will never be able to replace a human board member. I think that's what bothers technology people the most here. It's quite asymmetric. At best, LLM agents will be granted certain non voting advisory roles, but anything the llm generates will need a human to assume responsibility.
The point of a human being on the board is that they can be made to suffer the consequences of bad decisions. Executive pay might seem astronomical, but it is often commensurate with the level of stress and responsibility involved. It's easy to look at the perks and not see the struggle behind them. No one in media is going to get a lot of clicks if they publish articles about how being a CEO is actually really hard and maybe some of what they're doing is actually justified once you assume all of the same context and stress they have.
If we want to do to the executive staff what is being done to the technical staff, I'm afraid we will need to first figure out a way to make the AI experience human emotions as strongly as we do. It often sounds fun in first order terms to threaten to fire a CEO and replace them with a clanker, but have we considered the consequences of this? What would it be like to be under the management of an emotionless robot? Being managed by a robot vs managing robots are wildly different things to me.
> Executive pay might seem astronomical, but it is often commensurate with the level of stress and responsibility involved.
I call bull on that. Employees often have just as much stress, often imposed on them from those higher up, where if they don't do well they could get fired and it could mean they can't put food on the table.
Several times I've worked in environments where the stress was incredibly high, and once it even put me in the hospital after a long stretch of working 70 hour weeks to satisfy the insane deadlines my CEO boss wanted (this was at a startup).
Whereas if an executive screws up they just golden parachute to another cushy gig elsewhere, oftentimes, because even if they screw up, they're still valued as having experience running companies by other companies. And even if they cannot, they still likely got paid enough that they won't be struggling financially if they go through a period without a job.
> What would it be like to be under the management of an emotionless robot?
Depends. Is the robot looking out for our best interests? Or is the robot looking out for the shareholders?
I already ask AI to manage a lot of things, including my own activities. It all depends on whether the model can be trusted to act in our best interests.
CEOs understands that AI offers potential productivity increases. Using that productivity boost to cut staff is an unimaginative approach. Bolder approaches include using that boost to exceed the expectations of current customers, or to increase sales without proportional increase in staff, etc.
There are a LOT of bad CEOs.
There are also a LOT of bad software developers.
When they meet, the software developer is fired.
The CEO exits after a while, after exercising their stock options...
bad CEO + bad software developer = new CTO.
The token leaderboard example was shocking.
Counting token usage as a productivity metric is completely counterproductive. I believe that effectively leveraging AI means reducing wasted tokens, not increasing them.
The visibility gap is serious. I create local activity logs for Claude Code, but most developers have no idea what the AI is actually doing at the file or command level until they look at the logs.
If AI makes you more capable, it’s basically like having a capital injection.
CEOs that look at that and think they need to reduce headcount seem to also be signaling they don’t know what to do with increased resources
But the number up, what wrong? Less word more time, me good CEO.
Why can't we get AI models that replace these CEOs? I bet they're pretty good at running a company.
What about employees who think AI replaces their CEO
Upper management material written all over them.
Maybe the CEOs who think AI replaces their employees should be replaced by AI?
I keep seeing this sentiment in general and in this thread.
I wrote about it here too:
https://blog.nilesh.io/post/llms-and-jobs
Just because people have a vague idea of what someone's job is and a deep understanding of their own job, it feels like AI can replace everyone but you.
The working horse population continued to grow for almost 20 years after the introduction of the Oldsmobile in 1901: https://www.reddit.com/r/dataisbeautiful/comments/i2lmdx/num.... The number of horses didn’t fall below the 1900 level until 1930, two decades after the Model T was introduced.
Wait, tech CEOs don’t understand why employees are valuable?
Astronaut holding up gun to other astronaut
Always have been.
Modern day CEOs of public companies are just more or less hedge fund operators looking to squeeze every last dollar out of their workforce. AI is a tantalizing, if ineffective, lever for that.
I'm don't know what makes a bad CEO but I've definitely worked with people who could be replaced by a current-gen AI.
I have unfortunately worked with people who don't understand git. I mean Post-covid, to quell some of you funny guys.
>> Yes, the tools are powerful, but a CEO who thinks they replace the work of employees is simply a bad CEO.
This is a broad generalization of employees. There will be some "routine tasks" that can be done by AI, now that is a lot more powerful.
There won't be as many employees needed for routine work - for example L1 and L2 support work. For example, many companies had ML engineers building models for various models. Companies can get that off the shelf from AI companies. They don't need a big team of model builders now.
If L2 support work is the example then I doubt we’re near replacement.
L2 issues are already involved in some way often revealing some kind of system failure, requiring context and exploration to understand, and judgement (and perhaps even system overrides) to fix.
I could see “automated L2 is the new L1” improvements, but without a big capability jump and/or a resource bonfire, I don’t think even frontier models would effectively replace good L2 staff.
They might magnify good L2 staff so fewer are needed (and maybe even help L1 staff become L2).
My agent is currently doing the work of ~6 senior engineers, based on ticket closures. These are not trivial tasks: all of them involve a mix of code, judgement, executing remote commands in a production environment, etc.
This is at a well-known tech company operating at massive scale (and resulting complexity).
L1/L2 tech support is completely dead within a couple years. The delay is only around how long it takes people to realize.
I often use AI chatbot to generate step-by-step instruction for setting or repairing up some bit of tech. It's incredibly empowering, and saved me a a lot of money that would have been spent on buying replacement tech.
You know who can't do that? People who call L1 support.
Wow the token leaderboard idea is nuts. It's similar to trying to measure the productivity of software engineers based on number of lines of code.
The message isn't subtle, and isn't meant to be: "we don't care how, but we expect you to stick your nose into AI tools and find some way to fit them into your workflow".
Which indicates: the management believes there are productivity gains from AI use, but adoption lags due to inertia and reluctance to change existing workflows.
> Which indicates: the management believes there are productivity gains from AI use, but adoption lags due to inertia and reluctance to change existing workflows.
Methinks adoption lags due to management's inability to align incentives such that productivity gains are rewarded.
more of "we whined and cried and screamed that we needed new budget in order to buy these tools or we would literally die. now we have them, they don't work as well as we hoped, they aren't leading to productivity gains, and they're actively alienating our workforce and users alike. we're so screwed we literally have no idea how to do reverse this."
It’s hilarious to me that when you stop investing in juniors and seniors who use your AIs retire, what are they going to do then?
You bring in contractors.
Ideally contractors that benefit you personally (eg: your buddy who now owes you one), but definitely contractors that let you outsource the responsibility.
Even better if you get some management consultant to suggest the idea and/or do the subcontracting.
Definitely buys you a few quarters of bonus and some time to land your next gig.
Complain how nobody wants to work and/or blame foreigners. The usual.
Who wants to be CEO for that long? The company will be sold off to a larger conglomerate long before that happens.
Get your state governor to go on tv and ask for people who know how to program Cobalt.
Early in my career I read "Peopleware" by Tom DeMarco and Timothy Lister. Between that book Brooks' "Mythical Man Month" and Humphrey's "Managing the Software Process" it seemed that there was hope for the software industry learning some necessary lessons and growing to become a true engineering discipline. Nope. Never happened. The industry standard, despite improvements in some areas, is still a farcical shitshow with little beyond lipservice to process, predictability or proper self-evaluation. I can only describe agile, as it is practised, more of a coping mechanism than an actual methodology. Indeed the methodology is embodied mostly in the infrastructure; issue tracking, version control, code review, continuous integration with as little methodology glue between them as required to produce output.
Modern first and second tier software management seems less professional, is contributing less and is generally worse than it was twenty years ago. The quality of the engineering and program managers, their training and commitment to their craft seems really low and is not generally valued. On average team level software management has gotten worse rather than better and, given what is expected and how it is valued, this shouldn't be much of a surprise. It is truly disappointing that what could have been a valuable and productivity enhancing role became so useless.
Things aren't going to change for the better though until the dust settles somewhat on the role of AI in software and systems development and we start again to consider how software should be developed in the 21st century. Maybe it is possible that with AI doing most of the low-level work that the focus will change to building and maintaining architecture and systems. Many programmers might become more like traditional engineers doing a lot more systems work than they do today and continuing to solve problems. Lord knows though it won't be today's software management doing this work; they have nothing in the way of skills to offer to the problem.
I've had a urge to give that book to a couple of bosses when leaving the company (mostly because of said bosses).
I think sooner rather than later, AI will replace CEOs.
The CEOs that think AI replaces their employees are the same that at the same time don't want to pay the AI costs.
I've been a founder since before any of this existed, same flavor of small team then and now. What I see now is that anyone can produce more but less due to their skills from scratch (slop?). A big company sees the headlines and a CEO reads that as "fewer people", or force use of LLMs and people seeing it as a preparation for AI processes taking over (and rightfully so). But frankly probably for smaller teams it reads "same people, wider surface". Imo bad CEOs existed before AI and will exist in the future...
why shouldn't AI then replace the CEOs?
CEOs are probably the most replaceable position, period, by human or AI. Everyone just gives them information. They don't know any information themselves.
Problem is, a CEO can fire employees, find out it was a dumb decision, then leave with a million dollar severance package. So they don't really care when they're wrong.
A story from today - I needed a small utility to remap my logitech buttons under windows without installing their horrendous GHUB. Logitech Onboard Memory Manager still required ghub to be put into onboard mode.
The solution - linux has utility called piper. I downloaded the repo and just told codex - figure out what piper is doing and create me a small utility to do it under windows. So the jolly critter started experimenting with hex commands, then pulled some other repo on which piper depended figured out how to enable said onboard mode and 10-sh minutes later I had small python script that did what i needed to do.
That would have taken probably half a day of work for a human.
There are many stupid CEO and organizations which are not committed to quality. And a lot of employees that are too set in their ways. But the instinct that underinvesting in AI is more dangerous than overinvesting is right. Doomed if you do, doomed if you don't
"To err is human, but to really foul things up requires a computer" - this is from the 60, but right now is turned into overdrive.
I'm not sure I've ever met a good ceo...
> I will say that I hate the term “AI psychosis” because the term is extremely misleading, and many psychologists and psychiatrists have complained that it is inaccurate and may cause more problems itself. But the general sense that CEOs are going overboard with AI is definitely happening.
It's getting exhausting how x field of experts constantly bemoan the coining of one term or another, rather than provide a decent alternative. It's not very goal-oriented thinking. Just empty complaining.
When people object to the coining of a term, what they're usually trying to say is that it just fails to refer to any coherent thing or group of things for which you could develop an alternate term.
"AI psychosis" tries to group together the phenomenon where people fall in love with an LLM-generated character, the phenomenon where people spend too much time talking philosophy with LLMs and stop expressing coherent thoughts, and the phenomenon where tech evangelists say "AI can do all of these 5 million things" when actually it can't do all of them. But what if these are all different phenomena with different causes and solutions?
is it inaccurate because "AI" is not really AI? ;)
It is useless to preach against the wind.
We're all now on this and we will go together in it till the very end, whatever it maybe.
Look around: lots of places ressurected Lines of Code as a productivity metric for Software Teams. Companies that should have known better, as they are supposedly led by the Elite Human Capital, instituded token usage leaderboards.
We can't stop that thing. It has too much momentum. Sooner or later we would have to pay for our culture of anti-intelectualism in business anyway. If it was not this, it would be another thing.
CEOs see that AI can do their job and assume it can do everyone else's too.
The primary product of AI is labor displacement and consequently wage supression. This is what OpenAI and Anthropic are really selling. It didn't start with AI but AI is accelerating it.
This is what layoffs have been about since the pandemic. People in fear of losing their jobs do extra unpaid work and aren't asking for raises. The theoretical potential of AI gives companies the excuse to fire more people. The investment itself is directly used as a reason of why they need to cut back on labor.
Any sufficiently sized business can only feed the insatiable hunger for ever-increasing profits ultimately by cutting costs and raising prices. And what do we have now? High inflation and a decline in real wages. CEOs are just following this playbook.
And the result is that society is bouldering towards collapse. We're seeing the first hints of this with the youth unemployment crisis [1][2][3].
Also, who is going to buy anything when nobody has any money?
[1]: https://www.americanprogress.org/article/americas-10-million...
[2]: https://www.brookings.edu/articles/twelve-ways-to-fix-the-yo...
[3]: https://www.bbc.com/news/articles/cy026x9jpd0o
> Also, who is going to buy anything when nobody has any money?
This assumes that a mass consumer economy is necessary, when it isn't. Mass consumption is relatively new, for most of history economies functioned with just a small consuming elite and large underclass that consumed very little. We are already approaching that again in the states given that the top 10% of earners are already responsible for nearly half of all consumer spending.
There's a floor even in a mostly automated economy where some services are resistant to automation simply because the human element is the product. Luxury hospitality, personal care, etc. That billionaire is going to want a human masseuse, not a robot.
A highly automated economy could stabilize like this with a small elite population consuming luxury goods & services, served by a low-wage economic underclass human workforce.
Its certainly not a pleasant society, but its also not unsustainable given enough oppression or pacification (bread and circuses anyone?)
Care to share some sources that corroborate any of these claims and explains how the dots connect?
Consumption by the top 10% of earners: https://www.wsj.com/economy/consumers/us-economy-strength-ri...
As for services increasingly unaffordable by the workers providing them, that's Baumol's cost disease (https://en.wikipedia.org/wiki/Baumol_effect)
I don't have any sources regarding someone preferring humans for certain services over robots, just intuition there, and the fact that consuming human labor and time is itself a status signal of wealth, and the current growth of personal services.
As for connecting the dots, look at Brazil, one of the world's most unequal economies having a small consuming elite and a much, much larger low-wage service underclass. The gulf states as well. Granted, their circumstances don't map cleanly to a post-automation Western economy, but it does demonstrate that a largely bifurcated consumption based economy can exist and can be mildly stable.
Whether the US falls into that direction too will depending on politics. A bunch of mid-career knowledge workers aren't going to willingly to flip burgers in a service economy without some serious surveillance and oppression. But when thinking about it in those terms, the recent push for mass surveillance laws and tech along with the increasingly dangerous rhetoric around protests and "domestic terrorism" start to make sense.
> but it does demonstrate that a largely bifurcated consumption based economy can exist and can be mildly stable
This is not the case. These countries don’t operate/exist in isolation. They are part of global economy. But if this tech has a global scale effect, then the dots are not really connecting. You’re also forgetting the “revolt” factor.
The revolt factor is accounted for by ways of mass surveillance, and imprisonment for whatever gets labeled as domestic terrorism, like say, protesting data centers and "anti-tech sentiment" as a recent example.
it necessary for ceos doing layoffs. i thought thats context we are talking about.
This is a great article, and I agree with most of it.
The problem is that the wrong eyes are seeing it.
We need these kinds of articles to be published in places that executives read, and tailored to their audience.
AI recovery is going to be a big wave of consulting over the next several years, maybe very publicly or maybe quietly, but it's going to be a thing. That doesn't mean "all AI is bad" or any other such nonsense, it means that there are a lot of companies out there right now that are doing it wrong and will need help.
The executives that get ahead of this are going to be the winners.
So, a lot of the article makes several points that aren't necessarily new, but
> The problem tends to show up when a CEO is handed an agentic tool like Claude Code, and has it create something, which will work just fine, and thinks “oh, wait, why do we need so many people, when I can just sit here and make things work?”
> This is a bad CEO.
As described, this seems to me more like a lack of reasoning/critical thinking ability, and it's not unique to CEOs. Tracks more with a combo of "Gell-Mann Amnesia" and automation bias IMO.
> This all reminds me of cargo cult thinking: The CEO knows that somewhere in the org, employees are pecking away at computers and work gets done. So they figure that themselves pecking away with Claude Code and seeing work get done is the same thing. It’s not. All those other steps those people are handling — the ones the CEO never sees — still need to happen.
"Cargo culting" as described here by the author may be happening. But, I think it's CEOs seeing other CEOs doing layoffs and claiming it's because of AI efficiency gains. They see the other CEO's stock go up/get hyped/etc, so they decide to do it. I think it's the same thing that happens inside companies IE people see how others behave and it works, so they do the same. Effectiveness aside because that's not at all what I'm arguing, AI is just the current flavor; it is a very safe thing to "cargo cult" at the moment.
It's our fault for stupidly naming everything AI:
A* search -> AI
Backtracking -> AI
Neural Networks -> AI
Fuzzy Logic -> AI
Genetic Algorithm -> AI
Deep Learning -> AI
Generative "AI" -> AI
Similar to Tesla naming it's driver assistant "auto-pilot" in 2015 and your average Joe thought he would be able to sleep while the car would drive him to work.
The CEO just hear AI and think of AGI. They expect Skynet.
Aircraft autopilot is over 100 years old, and Testla autopilot is an automotive version that fits the definition by analogy to what aircaft autopilot does and what the human metaphor means (doing simple tasks aithout higher-level cognition or handling of surprising stimuli). Autopilot is not end-to-end driverless.
> the definition by analogy to what aircaft autopilot does
It's a bad analogy. Autopilot just maintains the aircraft in some state, then there's the flight director which maintains the flight path, and you can connect or disconnect the two at will. When connected the director can change the autopilot state.
To use the flight director you must fully specify your flight. The weight, the fuel, the weather, expected winds, takeoff and landing runway length, runway conditions, expected brake demand, as well as every single waypoint you're going to cross and the expected state at that crossing.
> what the human metaphor means
We learned after high levels of cockpit automation that maintaining situational awareness was still required. Pilots are freed of some stress during high workload portions of the flight, provided they planned correctly in advance, and that zero changes to their flight plan (not likely) occur.
As a result pilots are mostly told and trained to hand fly the plane during take offs and landings if the weather allows for it. You should only use high levels of automation if the situation demands it.
The point isn’t whether the aviation term "autopilot" has a technically defensible meaning, but that "Autopilot" predictably made normal people think "the car drives itself", just like "AI" now makes non-technical CEOs think "AGI employee replacement machine".
You are arguing the dictionary while I’m arguing the predictable, if not very well calculated and purposeful, misunderstanding.
Previously:
Tech CEOs are apparently suffering from AI psychosis
https://news.ycombinator.com/item?id=48295679
I believe there are entire companies right now under AI psychosis
https://news.ycombinator.com/item?id=48153379
So much of this hype feels like astroturf in preparation for the upcoming IPOs:
https://tomtunguz.com/spacex-openai-anthropic-ipo-2026/
and I don't know what worries me more - a burst in this bubble (and maybe some other tech stocks), or a failure of these valuations to be burst somehow, and even more concentration of capital and power around those corporations.
We're laughing now but the trendlines point to a future where this is true. That is the most realistic take as of right now.
Cue rationalizations claiming that it isn't:
Future? Try present.
CEOs who think that excavators replace their hand-diggers are just bad CEOs
If the excavators only worked half the time, sometimes goes out of control bonking someone in the head and costs a billion dollars, then yeah they're pretty bad CEOs
> The problem tends to show up when a CEO is handed an agentic tool like Claude Code, and has it create something, which will work just fine, and thinks “oh, wait, why do we need so many people, when I can just sit here and make things work?”
> This is a bad CEO.
There is one and only one measure of whether a CEO is good or bad:
Does the CEO keep the majority of shareholders happy?
Since they are more often than not kept happy with money, if the AI makes the CEO ask the question above and the result is a larger return on the shareholders' investment, then that is a good CEO.
When your domain of knowledge considers Jack Welch to be a genius, there is no floor.
I think it's more of a workforce reduction technique in some areas. how much is the issue, and lots of "it depends" but even at a higher up-front cost, LLMs have higher value and less liability. Instead of offshoring/outsourcing for example, you can have the same local devs that were leading/managing the offshore teams manage LLMs instead. Or if you have 10 expensive devs making 200k, just halving that is a million/year. even if that million is spent on tokens (even double or triple that) it might not be that much outside of what they would have been willing to pay for more devs anyways if there was the need, except now they have to deal with less people to manage overall. less lawyers, less hr, less lawsuits, less hiring and promption costs, less insurance premiums. Even if the quality of the work is a whole lot less, they only care about a viable product being shipped, and how much the lower quality affects profit margins.
I'll say this, laws and regulation are sorely needed. all this hate against billionaires, ceos, ai-bros, whatever... might or might not be warranted, but it is fruitless. Redirect this energy to your law makers.
In China for example, they made it illegal to use AI for the sole purpose of replacing human workers. The CEOs aren't bad CEOs, they aren't great either, it depends on the outcome. There are scenarios where entire job roles can be replaced by LLMs, but for complex roles like developers, you always need some human devs still, but likewise, almost always less of them than before. However, although less devs might be needed to do the same work, in some cases LLMs open up possibilities that weren't there before, so more devs babysitting LLMs or working on designing/shipping more features is also a strong possibility.
I mean, companies aren't trying to simply save cost on employees, they also want to maximize profits. Less devs per feature isn't the same as less devs period.
Overall, I'll say that demand creates its own supply. The internet itself killed many job roles and reduced the number of people you needed for many others, but it's not like we have the huge unemployment many were afraid of decades ago, if anything it created lots of more jobs. LLMs simply can't do everything people can, so they're not a drop-in replacement, that alone should mean a lot in terms of supply/demand economics.
Most CEOs are not special. They are not especially smart, or skilled, or technical. The role self selects for sociopathy. That's not a quality that has any kind of linear relationship with intelligence. Quite the contrary.
Sociopathy is a valuable skill in today's competitive marketplace.
A common misconception about AI is that it is intended to fully replace humans, which is incorrect. The purpose of AI is to reduce the need for human labor, and it has already been doing so. For example—though this is not an exact figure—a task that previously required 15 people might now only need 10. In no instance has the human element been completely replaced; rather, the reliance on manual labor has simply been reduced.
It's not exactly a misconception, when companies are pitching AI as a full and complete replacement for human employees. People are just reading the billboards on the side of the road.
I always found advertisements for AI to be so strange, why would you advertise your AI to the public as a danger for humanity that will also put everyone out of work? Such advertising would only appeal to sociopaths, but of course that's because it's intended to appeal to CEOs.
> though this is not an exact figure
You mean, this is an entirely made-up figure.
So what if it is? The example still stands.
A "unit of work" that required X people to complete in Y time can now be done by X/Z people in Y time, where Z is whatever efficiency you are able to get out of applying AI tooling to your business.
For some companies, Z might be less than 1 though. ;-)
So you still need skilled people, just not the same amount as before, because you have different tools available to you.
This has happened before with other advancements in industrial/technological automation. It's not a new concept.
That supposes that AI has a positive impact to efficiency. So far I see the exact contrary, at least for software engineering.
I see tons of evidence of positive impact, but I suppose everyone has their own perspective.
I did try to account for that with this line:
> For some companies, Z might be less than 1 though. ;-)
That sounds like 5 humans got replaced by AI. I don't think most people worry about whether all humans will be replaced, simply whether or not they will be replaced, or people they care about.
Or all 15 people are still employed doing the work of 22.5. Or even more people have been hired now that each person can generate 50% more value than they previously could. Or people are reallocated from the AI assisted task to another. Or some combination thereof.
Which is very short sighted. You or anyone close to you might not be replaced but it should be clear that you don't want to live in a society with 20% unemployment.
I think Spain had unemployment in that range for many years.
Do they look back on those years fondly?
You'd think we would see a large spike in unemployment if AI was reducing the number of employees needed for jobs the way these CEO talk about AI replacing people...
Look at narrative spreading -
- everyone is a programmer now
- AI will replace all jobs
market and people start panicking. Goyim notices
- Oh shit, change the narrative
- blame a few CEOs.