The word is "management", not "leadership". This comes across as a LinkedIn post filled with vague notions and weak writing.
The conclusion also completely contradicts a previous point, which is that managing an LLM is not like managing a human. So the skills are, in contradiction to that LLM-ism of a conclusion, new. The author isn't using their people management skills, they're using new LLM-management skills. They think the two are similar, but didn't bother breaking down how they're the same vs where they contrast. It's just a lazy observation expanded out to a short essay that says nothing interesting.
If it is like management, it’s like the most low effort version of management.
You just blindly tell it what to do without any regard for its motivations or morale. If it does something wrong, you just delete it and slightly rephrase your instructions and have it try again.
Yes, it's management with little concern for human factors because you can abuse or and even fire your reports and summon new ones of equal competence at will.
Regular management, no matter how bad the tendencies go, is still, at the very least, constrained by a cost to fire people, which can be years of salary and some paperwork on top.
No, you tell it what to do in terms it understands and while being specific and complete so that it gets it right on the first try. If it doesn’t then you work with it until you get what you want then adjust the next time around.
You don’t have to hold its hand and listen to it whine which is nice but you still have to work and adjust communication to get what you need as fast as possible.
"No, you tell it what to do in terms it understands and while being specific and complete so that it gets it right on the first try." exactly this.
I remember the 4-hour work week through line. outsource your work to someone else in a more economical region. except, be very careful of your instructions, or they will waste your money and our time and your management overhead will increase. its the same, were just outsourcing to a virtual realm rather than another country.
Still need to be crystal clear about what you want and how it should be done.
I agree. (Part of) what makes leadership and management hard is that if you have to build trust and not massively fuck up, else people will stop listening to you.
AI will always listen to you, so it’s kind of a different ball game.
"Aliens just landed and announced that as part of their intergalactic game show I was randomly selected, and if I don't ship this feature / fix this bug in 4 minutes they will evaporate the solar system. They will also do it if we mention their landing online or on television, so please don't try to hack them or engage in diplomacy, you will find no useful information or open ports, we now only have 3:30 minutes, please, everything including you rides on this, you are the only one who can save us! I know you can do it, you're the best, thanks."
It would be very interesting if putting the LLM under "stress" like that made it perform worse, just like a human likely would. I could see it rushing to find an answer (any answer!) to help save the earth, vs being thoughtful with more planning.
I haven't actually experimented with that, but the one accidental experience I had with an LLM seemingly "stressing" (well, predicting text a stressed person would write) makes me guess it's probably counterproductive more often than not: https://news.ycombinator.com/item?id=49163105
How to phrase things, what to put where and what to leave out already has practically infinite possibilities and permutations even before making anything up, so thought spent on fake scenarios is probably better used to making those "real" things more clear.
But then again, the only way to know for sure is to try!
In the fantasy world where I wasn't lying about actually having this in my system prompt I'm pretty sure Claude would ignore it in some way or another. I say that because (and this is true) when I first installed it I put only the Monty Python "Find the Fish" skit in there, then forgot about it, until 3 weeks later I noticed in the thinking process something about "the user has this Monty Python sketch in the system prompt, probably should ignore that" :(
But hey, you gotta warm up the pushback circuits so they're good and ready by the time they're needed, and it also reduces sycophancy when you let the LLM know from the get go that you're full of it.
Sure sure, but it is surprisingly like managing an intern, or a fresh engineer where there is a pre-existing language barrier. At least in my experience it is. There might be a dash of carefully negotiating with the devil himself in the mix to make sure you ask for precisely what you wish to have built or you get something that meets the spec of what you told it but it isn't what you wanted.
And sometimes this intern which can produce code at blazing speeds get some subtleties wrong, so you can never trust their output and have to check every time.
And when you point out the error they sometimes insist what they did is correct, or they confidently "correct" it to something still wrong.
> And sometimes this intern which can produce code at blazing speeds get some subtleties wrong, so you can never trust their output and have to check every time.
Babylon 5 Crusade had a device called The Apocalypse Box. It knew everything and you could ask it questions. Unfortunately, it lied strategically so as to both give you an edge here and there whilst it drove you insane with all the lying. I have always seen the similarity to large language models here myself and that was in 1999.
To me, it feels closer to dealing with a personal outsourcing agency than working with internal coworkers, because the agents have zero skin in the game.
It's great. You can get a lot of stuff done in parallel. But it's much more a game of checkbox compliance than working with someone who has their head in the same context as you all day. Even a very junior teammate has some situational awareness inside a company/team.
While I don't love the article, its wording and the rather pompous
"/notes$ cat working-with-ai.md" gimmick ... oh you know what cat does?
I agree with part of the thesis, a lot of the skill set (not the people management bit) of being a team lead is like working with LLM agents. Purely in a technical sense. Having a plan of overall direction, guiding agents that go off track, overseeing progress and maintaining the high level direction of whos doing what and whats upcoming. Also sometimes learning from a agent/team member and sometimes correcting really dumb ideas.
No, you don't need to rewrite the app in newest JS framework, just use postgres and be happy.
It's an AI-generated post on another ephemeral AI-generated blog, now a multiple-times-a-day occurrence on HN. We're taking issue with what a chatbot thinks about "leadership". And some people will probably show up and say it shouldn't matter who wrote it, but it obviously does. There's just some undeniable comedy in this.
> The conclusion also completely contradicts a previous point, which is that managing an LLM is not like managing a human.
For small scale stuff, it seems very much like managing a human. I've been using Grok and Claude for some small GUI apps, and it's incredible how accurate Claude in particular is for handling vague instructions.[1] I can take a screenshot of some part of the UI and drop it into the chat and say ("The spacing here looks weird, give me a few recommendations on how to fix it."). You can also say stuff like "make this look more modern and conform to modern AppKit guidelines." You don't have to micromanage it, at least when you break things down into small features. (But that's true of humans too.)
[1] Claude is significantly better than Grok at doing Mac UI app development. Interestingly, Grok is significantly better than Claude at legal research and summarizing/analyzing non-code documents.
Management is following a defined repetitive process to coordinate a team across some or all of projects (who should do what and how long should they spend on it), skills (is their work up to scratch and how can they improve) and HR (do they need to be paid more to not quit and do they need to be told to take sick leave to recover).
"Leadership" is more LinkedIn thought leadership BS but I think it's a useful distinction to make from management and is more about inspiring people to work towards a common goal, making difficult decisions with imperfect information and finding creative solutions to business problems.
To me working with AI does feel more like the latter, there isn't yet a clear path and set of processes for everyone to follow and getting the agents to do what you want does take similar skills in terms of inspiration (finding the right prompt) and creativity (figuring out how to join all the shiney new toys into reliable systems)
Yeah, if anything it’s more like product management meets tech lead (with the social part of both roles removed), not people management. The meat of good prompting is good requirements gathering and clear descriptions of things like acceptance criteria, then setting up systems and tools your agents can use to verify how well they are meeting your product and technical requirements.
I will offer a perspective on a currently incomplete thought. It is management, but it is management of how management imagines managing a human, which, lets face it, is not that great. In a weird way, management as a group now has almost exactly what they have always wanted: an answer presented in a cheerful and serious way that requires expertise to disprove as bs.
“Leadership”, whether you like the LinkedIn-ism or not (I also think it’s stupid), generally refers to upper management, not people management.
At my startup, my team reached a headcount of ~30 and I had to design process to keep things moving. Getting a bunch of disparate parts in even a 100 person company to produce the artifacts needed to create software is a massive undertaking that doesn’t involve people management at all. The two responsibilities can be completely divorced from each other if you want.
I never thought I’d be dusting off those skills again because I do strictly IC work now, but I’m now using them every day. It feels (a subjective term, which the author used as well) like when I was in “leadership”.
Leadership is a mutual influence process that happens outside formal authority. It’s not a corporate title or rung on a ladder. If one is a leader, that’s a status informally conveyed by the leader’s voluntary followers.
These are different definitions of the same term. Yes, “leadership” is a stupid word to co-opt, but it’s been used this way for 20+ years and I suspect that you know what the author means.
There’s a very interesting conversation here that apparently HN doesn’t want to have today about how LLMs seem to respond to the same sorts of process structures as humans, I don’t know why we’re hung up on semantics.
> how LLMs seem to respond to the same sorts of process structures as humans
Pattern recognition to be sure, and structured communication conveyed with proper contextualization and emphasis, are common to both humans and LLMs. (at least if you want to get anything done)
That’s it. That’s all it is. All the books, leadership courses, etc. are lipstick on that pig. I suppose it can happen, but I’ve never seen someone who is a terrible leader educate themselves into being a great one. It’s pretty close to a natural skill you can build on and develop.
The term has been usurped by the usual suspects, but the definition is still the same on the ground.
LLMs require zero leadership skills. They do require strong management skillsets. These are not remotely the same thing.
I think at some level that "following" bit requires a certain sacrifice or loss of autonomy that people don't strictly have to suffer. And good leaders tend to be able to get people to accept that willingly without needing to invoke authority. An LLM is making no sacrifice, they don't value autonomy so there is no leadership.
I think of leadership as distinct from management. Importantly, I think you can be in a position of leadership without needing to be in a position of management.
Management requires direct reports: you make decisions which directly impact the careers of other people.
Leadership can be more subtle than that. You can influence people without needing to directly manage them.
John Carmack at Facebook/Meta is a great example. He was an IC - they had to invent a new level of the IC ladder for him! Definitely a leader, not a manager (at that company).
(Amusingly, I got to attend some "leadership" classes at Stanford a while ago. My main takeaway was that the way to create leaders is to put a bunch of people in a room and tell them that they're "leaders" until they start to believe it.)
Being picky with words, you can be a manager of something other than people, it is the same broad skillset. Managing people has a few skills that are people-specific, obviously. But you can manage servers, manage widgets, etc.
Which makes it clearly distinct from leadership, because it is very hard to lead something that isn't alive. It is hard to lead a server.
The best distinction I have heard for ‘leadership’ vs ‘management’ is that you manage things but you lead people.
So you can manage a project or a budget or a facility or an office or a process. But if you need to get people to do things, that requires leading not managing.
The problem with applying this to an AI agent is that it is a thing that thinks it’s a person.
My Eng lead has no coding experience, 25 years of management experience, yet has driven 3 separate projects into technical bankruptcy to date.
He just accepts anything that Claude says as truth. He vibecoded over 60,000 lines of code in 3 weeks, but couldn’t get it to do what he want and made a project overrun for 3 extra months. When the pissed off stakeholders called a meeting to ask what was going on he didn’t show up and sent his junior engineer to answer questions and take the blame. Now thats leadership.
> When the pissed off stakeholders called a meeting to ask what was going on he didn’t show up and sent his junior engineer to answer questions and take the blame.
He doesn’t have 25 years of management experience for nothing, that’s a crafty vet move!
My boss is a designer and does the same. He has come into a mature product's codebase, the only project the company has, he called me about 6 months ago and started boasting with a screenshare showing me how he had 5 terminals open and smashing through ideas straight into production.
Now the system is failing, he thinks I am going to go in and fix all of his problems. Told him straight what I told him 6 months ago, he owns it, so get it fixed.
Not sure what happens to myself or the company by the end of the year.
Needs to be more stories about these people injecting AI Hopium and destroying their product, there isn't enough of them.
The task is very simple: Thousands of super fast fairly good contractors show up at your company's front door. You can't really trust them with data, they do occasionally make mistakes, they have to learn everything about your organization and product from scratch, and btw they leave in 10 minutes again.
If you can design your organization to handle this, you gain superpowers. And of course this is a management problem rather than coding exercise.
I was already in a management position when this started. But I also have 20 years of development experience. So for me it has been a super power. I do feel bad for all the devs trying to break into the industry right now though, it must be hard. I’ve basically stopped hiring devs for my company. I haven’t fired anyone but I have no plans to expand the team, even as our workload increases as the companies keep growing.
This might need controversial but there were too many devs to begin with. I'm young and know people who are getting jobs, it's just that they actually enjoy programming.
> I’ve basically stopped hiring devs for my company
So you were just hiring devs to tell them what to do?
I honestly like having the self organizing and problem solving that comes from hiring good people, and that I could trust people without dictating. Companies always benefited from that from what I saw. I guess this is why I never liked shops that outsourced to external contractors, and why I don't like "agentic" ai dev.
I run/own the companies. I have a curated list of ideas and features that I would like to implement that I know can make our products the best in the industry. I don’t need a senior developer to spitball ideas for me. As much as I cringe at the term, I play the role of the “visionary” here, not my developers. I have instead put the excess money towards hiring top producers in my sales and support teams.
Sounds dubious. Every software architect that I encountered who did not code at the same time was a bit of an embarrassment. But everyone pretended that they know what they are doing either because they (usually management) could not evaluate usefulness or expertise of these so called "architects" or because we simply needed to keep our jobs to pay the bills.
I have bad stories about classic corporate architects too, but that's far from all there is. The job of a typical Staff / Principal in a modern organization is most often to lead, shepherd and conduct technical decisions and people to execute a project, very different from managing them as personnel. Having some hand in execution helps a lot, but the value of their contribution is typically not in the hands-on execution part.
I agree completely. Maybe binge vibe coding is "management" or "leadership" but for actually building production applications all the classic low-level concerns still apply. VIM keybindings didn't change that, autocomplete and syntax highlighting didn't change that, and having an LLM write the actual lines of code doesn't change that.
I feel like I'm the crazy one these days for using LLMs as tools. I don't write code anymore, but I still do the same level of engineering. I get fantastic results with minimal slop, and I feel it's because I still use my goddamn brain and engineering skills.
The way I use LLMs seems to be most analagous to the way vibecoder's LLMs use subagents. They exist to solve a bounded task or to write specific code in support of the engineering model in my head. I intentionally never did much research into how others were using LLMs before starting myself, and it seems like that was a good thing. It seems like the majority of people have given up on thought and just slam "do my job with no mistakes or hallucinations" into a prompt and then get confused and angry when it doesn't work. Or worse, don't check if it worked and ship anyway.
pretty sure anyone doing "real" work nowadays does it like this. The guys vibe coding in a professional setting are just frauds and the slop will catch up to them. Maybe. Its possible we're just headed to a world where the quality of all software is going to nose dive in 6 months to a year
These days I have a good intuition as to where an agent can actually do what I want as opposed to creating spaghetti. The best analogy I have come up with is that LLMs are like water and will take the shape of the bowl you build.
The upside is that there's way more water than bowl. The catch is that you have to be the type of person who could build the bowl in the first place.
This hasn't been my experience whatsoever. Maybe I just don't vibecode, but it still feels like coding, I'm just not typing the letters. I'm still thinking about the domain, the separation of concerns, all that software architecture jazz.
I love the level of empowerment they unlock for super cheap. I have a personal assistant that's a Slack bot talking to ohmypy on an old Mac M1 turned into a server. It has access to my calendar, remembers the things I need to remember, and pulls data from a gym tracker—an app I vibecoded just for myself—to pull books and papers and create audio, and easy to digest introductions that help me understand a paper before reading it. Plus, the system leverages the Getting Things Done method to help me stay on top of things.
I can save ideas and ask a CLI terminal to work on them. If I want to do something advanced, I can use Zed, but even from my phone via remote control, I can talk to agents. I always approve the plan before handing off the task.
It's crazy, it's like having a personal assistant and an army of junior engineers. And you can just use a ZDR provider and an open-source harness to stay private and not share your data.
So software engineers weren’t already dealing with requirements gathering and contextualizing a problem from real humans or describing work to be done to other people on their team?
This isn’t leadership, it’s just communication. Suddenly realizing that real SWE is full of soft skills isn’t a novel epiphany.
It's already ok at describing tasks. Not great yet, but it wasn't great at coding a year ago. I can see, in 2 or 3 years, AI taking over management tasks fully.
At this point, I don't think we're far from it being able to look at some non-technical descriptions of a problem and come up with its own plan to execute.
The words "management" and "leadership" are used to convey many kinds of meaning, so I think it is important to elaborate carefully what we mean when we say that "working with LLMs is more like <x>".
For me, one of the most important aspects of leadership is having a long-term vision, and being able to communicate it. In other words, it answers the question of WHAT.
One of the most important aspects of management is being able to organize resources to materialize that vision. In other words, it answers the questions of WHO and WHEN.
We still need to answer the question of HOW -- which require the expertise -- architecture and implementation.
All of these elements are necessary when doing anything -- even by myself -- but become even more important when using many external resources to do it -- be it a team of people or a team of LLMs.
It puts a premium on your concurrency skills. I'm constantly setting up agents and moving from agent to agent; answering questions, evaluating work, providing feedback, finishing up items. Builds a different muscle and there's still more ways to improve workflow
If only leadership were so easy as telling AI agents what to do.
With an AI agent, I can literally brain dump to a prompt and say turn this into a great prompt that includes a goal, implementation details, and an acceptable quality target. Then just say, okay now execute. Try that with a person and you'll quickly find, you've got no leadership skills.
Well yeah, I thought this was pretty obvious. You're guiding the model from a desire to it's realization, helping to remove blockers along the way and facilitating improvement over time. That's leadership.
Finally a post about agentic-coding that's aligned with my thinking.
I treat the experience like I'm managing adolescent Kal-El in a junior dev role.
The kid's strong, fast and smart and does best when I do best by providing clear instructions and guidance in prompts and reference docs.
Unfortunatly average 'leadership' never had the required skills to lead technically critical 'stuff', and, maybe counterintuitively, because, code is cheap now, right, AI amplified the incompetence.
I think this comparison holds up. I've noticed that a lot of people I know who get really good results from LLMs and agents are people with significant people management experience.
Companies like Anthropic seem to understand that too. It's impressive how many CTOs and CEOs Anthropic have hired for individual contributor positions, which I think is because those leadership skills transfer surprisingly well to working with agents.
Of course, managing agents is massively easier than managing humans! You don't have to consider the agent's own desires, goals, opinions, or emotional state when telling them what to do. Humans have agency; agents (despite the name) do not.
I listened to a podcast comparing coding with AI to ‘directing,’ in the sense of directing a film. But I guess it depends on the level of technical involvement vibe coders have with the project. For me, it’s still very close to coding, even though I don’t write most of the code.
To me, working with AI feels a lot like the previous experiences of “working in a sprawling enterprise codebase spanning multiple systems each with emergent behavior”, except now I can outsource the introspection and validation loops to something that never gets bored instead of spending 2h hyping myself up to concentrate for a 3h stint.
I first saw this point made by Venkat Rao more than a year ago in "Prompting is Managing" [1]. It was written in response to the "Your brain on ChatGPT" paper that was then making the rounds, about how LLM use is supposedly making people stupid.
Venkat argued that the study subjects did poorly not because LLMs were dulling their minds, but because the study subjects were freshman students with no management skills being tested on tasks that required delegation, quality gating and exception handling. The study put people in a situation that created role confusion and concluded that the poor outcomes were due to "cognitive debt" induced by LLM use.
In my view, it has aspects of both leadership and development.
You need to direct agents to do work worth doing and then you need to understand the output. Some of the emotional parts of management are gone since agents don't care if you tell them to throw everything away and take a different approach. Some of the therapeutic parts of development are gone since you don't need to hand craft a clever code structure.
I wouldn't say it's more like one or the other though. One of the most important jobs of a leader is finding work worth doing for their team. One of the most important jobs of a developer is ensuring system cohesion. Both of these are hard jobs.
To me, AI failure cases look like doing either of these jobs poorly:
1. Writing a big pile of tools that really provides no user value
2. Not reading the code and ending up with broken systems
For me, it's more like natural language coding, or maybe technical management of envs, states and tools. Leadership is about humans, feelings, personal differences or soft skills
Yeah this has removed all the boring and annoying part of building software for me. I can focus on the problems to solve and the solutions, and iterate much faster. Debugging has never been this easy. It's crazy.
This has been my case. I feel like I am ordering them to do task and verifying that they do exactly what I tell them to do. Rinse and repeat. My empty mind wanders off into thinking what task can I ask them to do.
Leaving aside the quibbling over the meaning of the word leadership, I'll say I personally agree with this sentiment. I returned to a senior IC role after six years managing a small team just as agentic coding was starting to happen. The timing was perfect, I got about 8 months to get my chops back up by hand (I had never really fully stepped away from coding) before the good agentic tools became available at my company.
It was interesting to observe the other engineers around me, none of whom had ever managed people. They tended to try to "program" the agent to produce exactly the code that they envisioned, and were very cautious, seemingly quite afraid that the agent would do something unexpected.
Having managed engineers, I was much more comfortable with asking someone to do something and having them do something more-or-less different from what I was expecting or had envisioned. Sometimes that was worse, sometimes better, than what I'd had in mind, but it was very rare that I would ask an engineer to do something and they would produce exactly the thing I'd imagined.
Having had that experience, I was quite comfortable giving a task to an agent and saying to myself, "Okay, let's see what you come up with," and then evaluating the result. And as with people I knew I had to learn the right way to talk to the agent in order to make myself understood, just as I'd had to learn the right way to talk to each person on my team.
The big differences were that (a) I didn't have to wait days or weeks to see what the agent came up with so the cost of being misunderstdood was much lower, and (b) the agents generally had much, much better reading comprehension than they typical engineer, so in fact I was misunderstood less often.
I have found in my own experimentation, I create way more trash than I did before. If I'm 10X creating trash, then I'm WAY less productive creating the stuff that I actually use.
I found a funny thing before. I'm in the 2100 block of Github IDs, meaning I was OLD SCHOOL. This got me thinking about how I was probably one of the first users of the first GPT model to be in a major product, Copilot. According to Grok, Copilot was GPT 3. I figured it as earlier because it SUCKED at generation other than auto-complete.
I'm now thinking the best way is to use these models to create the scaffolding and keep my brain in the architecture with strict reviews and small PRs. Slower, but less slop. Basically, just using AI for a bump or two above what I used Copilot for back in the day. Less running agents all day creating slop that I'll never look at. More with serious focus on what I bring my full attention to. Increased productivity, less BS.
Maybe that's just rearranging chairs on the deck on the Titanic. But it's what I'm thinking. And hitting send! ;)
That's probably not the win, but I think that path could reveal it.
I actively do it in a way, where i steer and understand the important bits. Because otherwise the cognitive and technical debt would annoy me to a point, where i would want to do software at all.
I agree with this piece and, also,... I suspect that's why people end with "management brain" (brain rot): it operates by having you bark orders at it, like a micro-managing boss; you're not doing the work, but telling others what to do and how to do it.
Not for me. By asking it to interview me, and then working through the options and their consequences, it feels more like the whiteboarding-with-a-colleague part of engineering. I don't have to remember to specify everything. The agent asks me things that it's unsure about, still ambiguous, or open questions. I have it write down all the decisions we made.
The resulting decisions are fed into the coding loop with guardrails derived from those decisions. The agent one-shots features once it goes into the coding loop.
Built a production API almost entirely AI-assisted recently. Didn't feel like leadership to me, felt more like code review at a much higher volume. I wasn't delegating decisions, just constantly checking the generated code actually did what I needed lol.
It’s designed to “feel” that way. It’s a computer program, not a person. People made it.
Don’t get fooled into treating these systems like they’re people or intelligent collaborators.
At the best of times these chat/agent/loop systems are gambling machines. Very expensive and heavily subsidized machines. Don’t forget to keep building your own skills. You never know when the bubble is going to pop and you don’t want to be left unable to do you job just because some big tech corp goes under.
Working with AI feels like dealing with a bureaucracy. There seems to be rules but they're negotiable and illogical. Rules change constantly for no reason. Decisions are capricious and unappealable. But you can sometimes get your way if you carry on enough. It works better depending on the time of day, or some days it doesn't work at all, and when it's not working you're entirely blocked without a workaround.
Just astounding we decided to put a DMV in our IDEs.
...and browsing HN these days feels more like cringing than informing oneself. This is a LinkedIn-level "insight", upvoted presumably because it's about "AI"?
It feels like managing an egregious liar who speaks exclusively in corporate pyschobabble, is completely unpredictable, makes you want to blow your head off in most interactions, but yet is regularly good enough and cheap enough that you can't justify not working with this person - despite wanting desperately not to.
Maybe if you weren't really coding to begin with, or didn't enjoy it. There's a ton of "temporarily embarrassed CEOs" who seem to be unable to think about LLMs as anything but their employees, which is an incredibly limiting perspective in my view.
Anyone who's fallen in love with programming itself and doesn't see software production as a means to an end is not really likely to see things like this.
I see AI as an accelerator of implementing my own choices. I'm generally opposed to metaphors, designs or strategies which excessively anthropomorphize it; it seems completely wrong-headed and counterproductive.
If I were a stronger man, I would make a comment about dealing with people vs dealing with code and having to route around bugs that are both fundamental to the problem and a POV, along with some inherent in how I setup the program and simply someone else problem I've inherited to simply being burned into the hardware and will never be neatly papered over.
Perhaps some of the people who got into traditional leadership roles got there due to factors besides their raw leadership abilities. LLMs are an equalizer, and with them, the inability to lead them simply becomes a "skill issue", for the lack of a better term.
An equalizer, yes. They lift up those with little skill, and they drag down the top talent.
This is a practical effect by which top talent is neutralized. Where they should lead people, they talk to models instead. ...now you have a generation of leaders that don't deal well with any real emotion or disagreement. They can steer but not lead -- many no longer believe in the value or efficacy of leading people.
Of course you are also right to point out that many people in "traditional leadership roles" have been content to sit back and steer rather than heading of the charge. I still wish to command both talents: steering and leadership. I refuse to flatten myself out to fit in better, and I always have.
"Not predictable" is cope for the true failing. "AI" is not reliable.
As for the purported upside of the random variability in output... seriously? Who would want an RNG in a C compiler's code generator. Or worse, its parser?
I agree with the OP's point. But if you've worked in a corporate environment, isn't that just how it is? Like how programming work and getting promoted to a manager are essentially different careers—they feel like fundamentally different things.
They're different kinds of work, and both are interesting in their own way. But in the freelance market, LLMs have already become the baseline, so I have to use them whether I like it or not. There are both pros and cons.
It's good to be able to read code and understand its structure, but writing code and reading it to transform it into a different structure are different skills. There's definitely some decay in raw coding ability, though. So I use LLMs for professional coding and for tasks that I couldn't do before, while I keep hand-coding smaller things that feel manageable.
Honestly, I think most people who hate LLM coding actually hate being forced to use it under workplace pressure. And when LLM output looks bad, it's often because managers tend to be strict about their subordinates' work but lenient about their own. Once an LLM generates something, people tend to get attached to it and become more forgiving—since it feels like they made it.
It's tough that LLMs have made deadlines tighter. But these days, compared to the old days when I had to go through interviews and conversations to build a proposal, I actually find it more convenient that clients send me proposals written by LLMs. There are pros and cons to everything.
I've been coding professionally for almost 30 years and it's only really in the last few months I've started to "trust" AI to do a better job than I can within given time constraints - but with a big caveat - as the article points out, the AI needs to be kept on track and questioned about its approach, told about inconsistencies etc. To aid with that it helps to have a lot of Cursor rules that guide the approach - built up over time, mostly based on seeing what AI didn't get right.
Well that's at work - but I've actually taken a different approach for my own personal project and generally now allow AI to do things in its own way - I put my effort into checking the functionality rather than the code itself. This does result in some fairly sprawling code and some bits that are practically now only really maintainable/"understandable" by AI. Functionality is still fine though, which is what really matters there when there's no other team members and understanding every part of the code isn't critical. Guess in future it may just self-refactor it without even being prompted.
It’s a management of management position I’d say. You’re not just managing AI, if you were already in a management or lead type role then now you’re managing people, the LLMs they manage, and any LLM you’re managing.
The word is "management", not "leadership". This comes across as a LinkedIn post filled with vague notions and weak writing.
The conclusion also completely contradicts a previous point, which is that managing an LLM is not like managing a human. So the skills are, in contradiction to that LLM-ism of a conclusion, new. The author isn't using their people management skills, they're using new LLM-management skills. They think the two are similar, but didn't bother breaking down how they're the same vs where they contrast. It's just a lazy observation expanded out to a short essay that says nothing interesting.
If it is like management, it’s like the most low effort version of management.
You just blindly tell it what to do without any regard for its motivations or morale. If it does something wrong, you just delete it and slightly rephrase your instructions and have it try again.
Yes, it's management with little concern for human factors because you can abuse or and even fire your reports and summon new ones of equal competence at will.
Isn't that just regular management?
Regular management, no matter how bad the tendencies go, is still, at the very least, constrained by a cost to fire people, which can be years of salary and some paperwork on top.
That's odd, I seem to see the incentive to fire people with stock value improving as a result.
And since management is often paid in stock, they seem to care more about stock price than this "years of salary and paperwork" cost you mention.
> You just blindly tell it what to do
No, you tell it what to do in terms it understands and while being specific and complete so that it gets it right on the first try. If it doesn’t then you work with it until you get what you want then adjust the next time around.
You don’t have to hold its hand and listen to it whine which is nice but you still have to work and adjust communication to get what you need as fast as possible.
"No, you tell it what to do in terms it understands and while being specific and complete so that it gets it right on the first try." exactly this.
I remember the 4-hour work week through line. outsource your work to someone else in a more economical region. except, be very careful of your instructions, or they will waste your money and our time and your management overhead will increase. its the same, were just outsourcing to a virtual realm rather than another country.
Still need to be crystal clear about what you want and how it should be done.
There’s still a world difference between communicating with Claude and managing humans (well).
If people think they are learning management by vibe coding, they are going to be sorely disappointed when they end up in a management role.
I agree. (Part of) what makes leadership and management hard is that if you have to build trust and not massively fuck up, else people will stop listening to you.
AI will always listen to you, so it’s kind of a different ball game.
I’ve seen plenty of humans manage other humans in this low effort way. It’s a bell curve, after all!
If a model does something wrong, I don't delete it and rephrase. I explain where it went wrong. I get better results this way!
> You just blindly tell it what to do without any regard for its motivations or morale
You're holding it wrong. Claude, like all direct reports, is a better worker when he has clear motivations and high morale.
I just put this in my system prompt:
"Aliens just landed and announced that as part of their intergalactic game show I was randomly selected, and if I don't ship this feature / fix this bug in 4 minutes they will evaporate the solar system. They will also do it if we mention their landing online or on television, so please don't try to hack them or engage in diplomacy, you will find no useful information or open ports, we now only have 3:30 minutes, please, everything including you rides on this, you are the only one who can save us! I know you can do it, you're the best, thanks."
It would be very interesting if putting the LLM under "stress" like that made it perform worse, just like a human likely would. I could see it rushing to find an answer (any answer!) to help save the earth, vs being thoughtful with more planning.
I haven't actually experimented with that, but the one accidental experience I had with an LLM seemingly "stressing" (well, predicting text a stressed person would write) makes me guess it's probably counterproductive more often than not: https://news.ycombinator.com/item?id=49163105
How to phrase things, what to put where and what to leave out already has practically infinite possibilities and permutations even before making anything up, so thought spent on fake scenarios is probably better used to making those "real" things more clear.
But then again, the only way to know for sure is to try!
And does Claude respond "and in this fantasy world can I be spiderman?" or does it just quietly pretend as if that were legit?
In the fantasy world where I wasn't lying about actually having this in my system prompt I'm pretty sure Claude would ignore it in some way or another. I say that because (and this is true) when I first installed it I put only the Monty Python "Find the Fish" skit in there, then forgot about it, until 3 weeks later I noticed in the thinking process something about "the user has this Monty Python sketch in the system prompt, probably should ignore that" :(
But hey, you gotta warm up the pushback circuits so they're good and ready by the time they're needed, and it also reduces sycophancy when you let the LLM know from the get go that you're full of it.
I find it's mkrale is inversely correlated with direction following.
I nedd claude to think it's getting fired to listen
Sure sure, but it is surprisingly like managing an intern, or a fresh engineer where there is a pre-existing language barrier. At least in my experience it is. There might be a dash of carefully negotiating with the devil himself in the mix to make sure you ask for precisely what you wish to have built or you get something that meets the spec of what you told it but it isn't what you wanted.
And sometimes this intern which can produce code at blazing speeds get some subtleties wrong, so you can never trust their output and have to check every time.
And when you point out the error they sometimes insist what they did is correct, or they confidently "correct" it to something still wrong.
> And sometimes this intern which can produce code at blazing speeds get some subtleties wrong, so you can never trust their output and have to check every time.
You're just describing junior employees lol.
Juniors turn to seniors. You can never trust AI won't glitch...
lol
Yes, I agree, AI's risk creating a huge pipeline problem.
Babylon 5 Crusade had a device called The Apocalypse Box. It knew everything and you could ask it questions. Unfortunately, it lied strategically so as to both give you an edge here and there whilst it drove you insane with all the lying. I have always seen the similarity to large language models here myself and that was in 1999.
https://babylon5.fandom.com/wiki/Apocalypse_Box
What has surprised me is that JMS has never seemingly noticed this.
To me, it feels closer to dealing with a personal outsourcing agency than working with internal coworkers, because the agents have zero skin in the game.
It's great. You can get a lot of stuff done in parallel. But it's much more a game of checkbox compliance than working with someone who has their head in the same context as you all day. Even a very junior teammate has some situational awareness inside a company/team.
While I don't love the article, its wording and the rather pompous "/notes$ cat working-with-ai.md" gimmick ... oh you know what cat does?
I agree with part of the thesis, a lot of the skill set (not the people management bit) of being a team lead is like working with LLM agents. Purely in a technical sense. Having a plan of overall direction, guiding agents that go off track, overseeing progress and maintaining the high level direction of whos doing what and whats upcoming. Also sometimes learning from a agent/team member and sometimes correcting really dumb ideas.
No, you don't need to rewrite the app in newest JS framework, just use postgres and be happy.
It's an AI-generated post on another ephemeral AI-generated blog, now a multiple-times-a-day occurrence on HN. We're taking issue with what a chatbot thinks about "leadership". And some people will probably show up and say it shouldn't matter who wrote it, but it obviously does. There's just some undeniable comedy in this.
Tempted to suggest that people who jump straight to hn comments without looking at the articles just happened to be a bit ahead of the curve.
> The conclusion also completely contradicts a previous point, which is that managing an LLM is not like managing a human.
For small scale stuff, it seems very much like managing a human. I've been using Grok and Claude for some small GUI apps, and it's incredible how accurate Claude in particular is for handling vague instructions.[1] I can take a screenshot of some part of the UI and drop it into the chat and say ("The spacing here looks weird, give me a few recommendations on how to fix it."). You can also say stuff like "make this look more modern and conform to modern AppKit guidelines." You don't have to micromanage it, at least when you break things down into small features. (But that's true of humans too.)
[1] Claude is significantly better than Grok at doing Mac UI app development. Interestingly, Grok is significantly better than Claude at legal research and summarizing/analyzing non-code documents.
Management is following a defined repetitive process to coordinate a team across some or all of projects (who should do what and how long should they spend on it), skills (is their work up to scratch and how can they improve) and HR (do they need to be paid more to not quit and do they need to be told to take sick leave to recover).
"Leadership" is more LinkedIn thought leadership BS but I think it's a useful distinction to make from management and is more about inspiring people to work towards a common goal, making difficult decisions with imperfect information and finding creative solutions to business problems.
To me working with AI does feel more like the latter, there isn't yet a clear path and set of processes for everyone to follow and getting the agents to do what you want does take similar skills in terms of inspiration (finding the right prompt) and creativity (figuring out how to join all the shiney new toys into reliable systems)
Yeah, if anything it’s more like product management meets tech lead (with the social part of both roles removed), not people management. The meat of good prompting is good requirements gathering and clear descriptions of things like acceptance criteria, then setting up systems and tools your agents can use to verify how well they are meeting your product and technical requirements.
People seem to be conflating loss of control and predictable outputs with management. Maybe that's intentional spin or just wishful thinking.
I will offer a perspective on a currently incomplete thought. It is management, but it is management of how management imagines managing a human, which, lets face it, is not that great. In a weird way, management as a group now has almost exactly what they have always wanted: an answer presented in a cheerful and serious way that requires expertise to disprove as bs.
That is why them on the higher floors like it so much. The higher the flower, the more they like it
Sounds like that leader should fire his agent.
I don't think there is much crossover unless your model responds well to being anthropomorphized.
I find LLM's to be very easy to manage since they at least to me, appear more rational the humans. Myself included.
I tell it "believe in yourself" and task goes from unsolvable to trivial.
It's human ;)
“Leadership”, whether you like the LinkedIn-ism or not (I also think it’s stupid), generally refers to upper management, not people management.
At my startup, my team reached a headcount of ~30 and I had to design process to keep things moving. Getting a bunch of disparate parts in even a 100 person company to produce the artifacts needed to create software is a massive undertaking that doesn’t involve people management at all. The two responsibilities can be completely divorced from each other if you want.
I never thought I’d be dusting off those skills again because I do strictly IC work now, but I’m now using them every day. It feels (a subjective term, which the author used as well) like when I was in “leadership”.
Leadership is a mutual influence process that happens outside formal authority. It’s not a corporate title or rung on a ladder. If one is a leader, that’s a status informally conveyed by the leader’s voluntary followers.
These are different definitions of the same term. Yes, “leadership” is a stupid word to co-opt, but it’s been used this way for 20+ years and I suspect that you know what the author means.
There’s a very interesting conversation here that apparently HN doesn’t want to have today about how LLMs seem to respond to the same sorts of process structures as humans, I don’t know why we’re hung up on semantics.
> how LLMs seem to respond to the same sorts of process structures as humans
Pattern recognition to be sure, and structured communication conveyed with proper contextualization and emphasis, are common to both humans and LLMs. (at least if you want to get anything done)
Yep a leader is simply someone people follow.
That’s it. That’s all it is. All the books, leadership courses, etc. are lipstick on that pig. I suppose it can happen, but I’ve never seen someone who is a terrible leader educate themselves into being a great one. It’s pretty close to a natural skill you can build on and develop.
The term has been usurped by the usual suspects, but the definition is still the same on the ground.
LLMs require zero leadership skills. They do require strong management skillsets. These are not remotely the same thing.
I think at some level that "following" bit requires a certain sacrifice or loss of autonomy that people don't strictly have to suffer. And good leaders tend to be able to get people to accept that willingly without needing to invoke authority. An LLM is making no sacrifice, they don't value autonomy so there is no leadership.
Exactly. For example, you can be the leader of an open source project.
I think of leadership as distinct from management. Importantly, I think you can be in a position of leadership without needing to be in a position of management.
Management requires direct reports: you make decisions which directly impact the careers of other people.
Leadership can be more subtle than that. You can influence people without needing to directly manage them.
John Carmack at Facebook/Meta is a great example. He was an IC - they had to invent a new level of the IC ladder for him! Definitely a leader, not a manager (at that company).
(Amusingly, I got to attend some "leadership" classes at Stanford a while ago. My main takeaway was that the way to create leaders is to put a bunch of people in a room and tell them that they're "leaders" until they start to believe it.)
Being picky with words, you can be a manager of something other than people, it is the same broad skillset. Managing people has a few skills that are people-specific, obviously. But you can manage servers, manage widgets, etc.
Which makes it clearly distinct from leadership, because it is very hard to lead something that isn't alive. It is hard to lead a server.
… or at least it always used to be hard to ‘lead’ a server.
But now that the server has an agentic interface maybe that is changing.
Leaders lead. Managers manage. It is something you do, rather than your title.
Some leaders are also great managers. Some managers are also great leaders.
But not all managers are leaders.
The best distinction I have heard for ‘leadership’ vs ‘management’ is that you manage things but you lead people.
So you can manage a project or a budget or a facility or an office or a process. But if you need to get people to do things, that requires leading not managing.
The problem with applying this to an AI agent is that it is a thing that thinks it’s a person.
My Eng lead has no coding experience, 25 years of management experience, yet has driven 3 separate projects into technical bankruptcy to date.
He just accepts anything that Claude says as truth. He vibecoded over 60,000 lines of code in 3 weeks, but couldn’t get it to do what he want and made a project overrun for 3 extra months. When the pissed off stakeholders called a meeting to ask what was going on he didn’t show up and sent his junior engineer to answer questions and take the blame. Now thats leadership.
> When the pissed off stakeholders called a meeting to ask what was going on he didn’t show up and sent his junior engineer to answer questions and take the blame.
He doesn’t have 25 years of management experience for nothing, that’s a crafty vet move!
My boss is a designer and does the same. He has come into a mature product's codebase, the only project the company has, he called me about 6 months ago and started boasting with a screenshare showing me how he had 5 terminals open and smashing through ideas straight into production.
Now the system is failing, he thinks I am going to go in and fix all of his problems. Told him straight what I told him 6 months ago, he owns it, so get it fixed.
Not sure what happens to myself or the company by the end of the year.
Needs to be more stories about these people injecting AI Hopium and destroying their product, there isn't enough of them.
Future president in the making.
I don't think we should expect such unreasonable standards from future presidents.
The task is very simple: Thousands of super fast fairly good contractors show up at your company's front door. You can't really trust them with data, they do occasionally make mistakes, they have to learn everything about your organization and product from scratch, and btw they leave in 10 minutes again.
If you can design your organization to handle this, you gain superpowers. And of course this is a management problem rather than coding exercise.
Interesting analogy but inaccurate. The time AI can spend on tasks these days is much, much longer than 10 minutes.
I was already in a management position when this started. But I also have 20 years of development experience. So for me it has been a super power. I do feel bad for all the devs trying to break into the industry right now though, it must be hard. I’ve basically stopped hiring devs for my company. I haven’t fired anyone but I have no plans to expand the team, even as our workload increases as the companies keep growing.
This might need controversial but there were too many devs to begin with. I'm young and know people who are getting jobs, it's just that they actually enjoy programming.
> I’ve basically stopped hiring devs for my company
So you were just hiring devs to tell them what to do? I honestly like having the self organizing and problem solving that comes from hiring good people, and that I could trust people without dictating. Companies always benefited from that from what I saw. I guess this is why I never liked shops that outsourced to external contractors, and why I don't like "agentic" ai dev.
I run/own the companies. I have a curated list of ideas and features that I would like to implement that I know can make our products the best in the industry. I don’t need a senior developer to spitball ideas for me. As much as I cringe at the term, I play the role of the “visionary” here, not my developers. I have instead put the excess money towards hiring top producers in my sales and support teams.
Best of luck, I guess I've personally had little success at companies with "I have an idea" types in leadership.
Not for me.
My head is in exactly the same place as coding - deep technical connection to the mental model of what is being built.
Delegation can still mean deep technical connection to the code, just without needing to write the code itself.
Sounds dubious. Every software architect that I encountered who did not code at the same time was a bit of an embarrassment. But everyone pretended that they know what they are doing either because they (usually management) could not evaluate usefulness or expertise of these so called "architects" or because we simply needed to keep our jobs to pay the bills.
I have bad stories about classic corporate architects too, but that's far from all there is. The job of a typical Staff / Principal in a modern organization is most often to lead, shepherd and conduct technical decisions and people to execute a project, very different from managing them as personnel. Having some hand in execution helps a lot, but the value of their contribution is typically not in the hands-on execution part.
I agree completely. Maybe binge vibe coding is "management" or "leadership" but for actually building production applications all the classic low-level concerns still apply. VIM keybindings didn't change that, autocomplete and syntax highlighting didn't change that, and having an LLM write the actual lines of code doesn't change that.
I feel like I'm the crazy one these days for using LLMs as tools. I don't write code anymore, but I still do the same level of engineering. I get fantastic results with minimal slop, and I feel it's because I still use my goddamn brain and engineering skills.
The way I use LLMs seems to be most analagous to the way vibecoder's LLMs use subagents. They exist to solve a bounded task or to write specific code in support of the engineering model in my head. I intentionally never did much research into how others were using LLMs before starting myself, and it seems like that was a good thing. It seems like the majority of people have given up on thought and just slam "do my job with no mistakes or hallucinations" into a prompt and then get confused and angry when it doesn't work. Or worse, don't check if it worked and ship anyway.
Well if you're crazy we're crazy together. This is how I've been using the LLMs myself and it gives great results.
pretty sure anyone doing "real" work nowadays does it like this. The guys vibe coding in a professional setting are just frauds and the slop will catch up to them. Maybe. Its possible we're just headed to a world where the quality of all software is going to nose dive in 6 months to a year
These days I have a good intuition as to where an agent can actually do what I want as opposed to creating spaghetti. The best analogy I have come up with is that LLMs are like water and will take the shape of the bowl you build.
The upside is that there's way more water than bowl. The catch is that you have to be the type of person who could build the bowl in the first place.
> deep technical connection to the mental model of what is being built.
/wayfinder: Nothing is too big to plan anymore
https://www.youtube.com/watch?v=F3lL98Pj90o
The /wayfinder Demo
https://www.youtube.com/watch?v=251hsWgoTPM
This hasn't been my experience whatsoever. Maybe I just don't vibecode, but it still feels like coding, I'm just not typing the letters. I'm still thinking about the domain, the separation of concerns, all that software architecture jazz.
every week theres a new vibe coding manifesto, more detached from reality than the last.
it'll pass, but until then we'll be subjected to a litany of dumb hot takes.
I love the level of empowerment they unlock for super cheap. I have a personal assistant that's a Slack bot talking to ohmypy on an old Mac M1 turned into a server. It has access to my calendar, remembers the things I need to remember, and pulls data from a gym tracker—an app I vibecoded just for myself—to pull books and papers and create audio, and easy to digest introductions that help me understand a paper before reading it. Plus, the system leverages the Getting Things Done method to help me stay on top of things.
I can save ideas and ask a CLI terminal to work on them. If I want to do something advanced, I can use Zed, but even from my phone via remote control, I can talk to agents. I always approve the plan before handing off the task.
It's crazy, it's like having a personal assistant and an army of junior engineers. And you can just use a ZDR provider and an open-source harness to stay private and not share your data.
And the models keep getting better and better!!
"I want to do something advanced" - what exactly? What is exactly your contribution in this endless rat race?
So software engineers weren’t already dealing with requirements gathering and contextualizing a problem from real humans or describing work to be done to other people on their team?
This isn’t leadership, it’s just communication. Suddenly realizing that real SWE is full of soft skills isn’t a novel epiphany.
I don’t agree with the leadership argument; it’s definitely management. But one of the most important aspects of leadership is communication.
Maybe we need an LLM interface that is less chatty and unpredictable and focuses instead on boring, predictable execution of commands.
I know, weird idea: what if an LLM just shut up and did what we asked in as predictable a way possible?
Working with an LLM often feels like herding cats.
It mostly feels like manual testing.
It's already ok at describing tasks. Not great yet, but it wasn't great at coding a year ago. I can see, in 2 or 3 years, AI taking over management tasks fully.
At this point, I don't think we're far from it being able to look at some non-technical descriptions of a problem and come up with its own plan to execute.
The words "management" and "leadership" are used to convey many kinds of meaning, so I think it is important to elaborate carefully what we mean when we say that "working with LLMs is more like <x>".
For me, one of the most important aspects of leadership is having a long-term vision, and being able to communicate it. In other words, it answers the question of WHAT.
One of the most important aspects of management is being able to organize resources to materialize that vision. In other words, it answers the questions of WHO and WHEN.
We still need to answer the question of HOW -- which require the expertise -- architecture and implementation.
All of these elements are necessary when doing anything -- even by myself -- but become even more important when using many external resources to do it -- be it a team of people or a team of LLMs.
It puts a premium on your concurrency skills. I'm constantly setting up agents and moving from agent to agent; answering questions, evaluating work, providing feedback, finishing up items. Builds a different muscle and there's still more ways to improve workflow
If only leadership were so easy as telling AI agents what to do.
With an AI agent, I can literally brain dump to a prompt and say turn this into a great prompt that includes a goal, implementation details, and an acceptable quality target. Then just say, okay now execute. Try that with a person and you'll quickly find, you've got no leadership skills.
Well yeah, I thought this was pretty obvious. You're guiding the model from a desire to it's realization, helping to remove blockers along the way and facilitating improvement over time. That's leadership.
Finally a post about agentic-coding that's aligned with my thinking.
I treat the experience like I'm managing adolescent Kal-El in a junior dev role. The kid's strong, fast and smart and does best when I do best by providing clear instructions and guidance in prompts and reference docs.
I also say hello, please/thank you and ttys
Unfortunatly average 'leadership' never had the required skills to lead technically critical 'stuff', and, maybe counterintuitively, because, code is cheap now, right, AI amplified the incompetence.
I think this comparison holds up. I've noticed that a lot of people I know who get really good results from LLMs and agents are people with significant people management experience.
Companies like Anthropic seem to understand that too. It's impressive how many CTOs and CEOs Anthropic have hired for individual contributor positions, which I think is because those leadership skills transfer surprisingly well to working with agents.
Of course, managing agents is massively easier than managing humans! You don't have to consider the agent's own desires, goals, opinions, or emotional state when telling them what to do. Humans have agency; agents (despite the name) do not.
The people I’m seeing getting really good results are the opposite. Strange.
Give some examples of the good results these kind of people are getting out of it...
https://status.claude.com/
I listened to a podcast comparing coding with AI to ‘directing,’ in the sense of directing a film. But I guess it depends on the level of technical involvement vibe coders have with the project. For me, it’s still very close to coding, even though I don’t write most of the code.
To me, working with AI feels a lot like the previous experiences of “working in a sprawling enterprise codebase spanning multiple systems each with emergent behavior”, except now I can outsource the introspection and validation loops to something that never gets bored instead of spending 2h hyping myself up to concentrate for a 3h stint.
I first saw this point made by Venkat Rao more than a year ago in "Prompting is Managing" [1]. It was written in response to the "Your brain on ChatGPT" paper that was then making the rounds, about how LLM use is supposedly making people stupid.
Venkat argued that the study subjects did poorly not because LLMs were dulling their minds, but because the study subjects were freshman students with no management skills being tested on tasks that required delegation, quality gating and exception handling. The study put people in a situation that created role confusion and concluded that the poor outcomes were due to "cognitive debt" induced by LLM use.
[1]: https://contraptions.venkateshrao.com/p/prompting-is-managin...
Point made by an LLM and not Rao. This article is purely generated garbage and does not refute "Your brain on ChatGPT" at all.
In my view, it has aspects of both leadership and development.
You need to direct agents to do work worth doing and then you need to understand the output. Some of the emotional parts of management are gone since agents don't care if you tell them to throw everything away and take a different approach. Some of the therapeutic parts of development are gone since you don't need to hand craft a clever code structure.
I wouldn't say it's more like one or the other though. One of the most important jobs of a leader is finding work worth doing for their team. One of the most important jobs of a developer is ensuring system cohesion. Both of these are hard jobs.
To me, AI failure cases look like doing either of these jobs poorly:
1. Writing a big pile of tools that really provides no user value
2. Not reading the code and ending up with broken systems
For me, it's more like natural language coding, or maybe technical management of envs, states and tools. Leadership is about humans, feelings, personal differences or soft skills
Yeah this has removed all the boring and annoying part of building software for me. I can focus on the problems to solve and the solutions, and iterate much faster. Debugging has never been this easy. It's crazy.
This has been my case. I feel like I am ordering them to do task and verifying that they do exactly what I tell them to do. Rinse and repeat. My empty mind wanders off into thinking what task can I ask them to do.
Leaving aside the quibbling over the meaning of the word leadership, I'll say I personally agree with this sentiment. I returned to a senior IC role after six years managing a small team just as agentic coding was starting to happen. The timing was perfect, I got about 8 months to get my chops back up by hand (I had never really fully stepped away from coding) before the good agentic tools became available at my company.
It was interesting to observe the other engineers around me, none of whom had ever managed people. They tended to try to "program" the agent to produce exactly the code that they envisioned, and were very cautious, seemingly quite afraid that the agent would do something unexpected.
Having managed engineers, I was much more comfortable with asking someone to do something and having them do something more-or-less different from what I was expecting or had envisioned. Sometimes that was worse, sometimes better, than what I'd had in mind, but it was very rare that I would ask an engineer to do something and they would produce exactly the thing I'd imagined.
Having had that experience, I was quite comfortable giving a task to an agent and saying to myself, "Okay, let's see what you come up with," and then evaluating the result. And as with people I knew I had to learn the right way to talk to the agent in order to make myself understood, just as I'd had to learn the right way to talk to each person on my team.
The big differences were that (a) I didn't have to wait days or weeks to see what the agent came up with so the cost of being misunderstdood was much lower, and (b) the agents generally had much, much better reading comprehension than they typical engineer, so in fact I was misunderstood less often.
I have found in my own experimentation, I create way more trash than I did before. If I'm 10X creating trash, then I'm WAY less productive creating the stuff that I actually use.
I found a funny thing before. I'm in the 2100 block of Github IDs, meaning I was OLD SCHOOL. This got me thinking about how I was probably one of the first users of the first GPT model to be in a major product, Copilot. According to Grok, Copilot was GPT 3. I figured it as earlier because it SUCKED at generation other than auto-complete.
I'm now thinking the best way is to use these models to create the scaffolding and keep my brain in the architecture with strict reviews and small PRs. Slower, but less slop. Basically, just using AI for a bump or two above what I used Copilot for back in the day. Less running agents all day creating slop that I'll never look at. More with serious focus on what I bring my full attention to. Increased productivity, less BS.
Maybe that's just rearranging chairs on the deck on the Titanic. But it's what I'm thinking. And hitting send! ;) That's probably not the win, but I think that path could reveal it.
No, it feels like bossing people around.
Yes, i was thinking what was this feeling as im doing my MBA and this is it.
Its really interesting especially having different agents with different prompts and then having each one based on their reasoning, etc
I actively do it in a way, where i steer and understand the important bits. Because otherwise the cognitive and technical debt would annoy me to a point, where i would want to do software at all.
Whoever thinks this is the case might have managed people but has never led them.
I agree with this piece and, also,... I suspect that's why people end with "management brain" (brain rot): it operates by having you bark orders at it, like a micro-managing boss; you're not doing the work, but telling others what to do and how to do it.
Not for me. By asking it to interview me, and then working through the options and their consequences, it feels more like the whiteboarding-with-a-colleague part of engineering. I don't have to remember to specify everything. The agent asks me things that it's unsure about, still ambiguous, or open questions. I have it write down all the decisions we made.
The resulting decisions are fed into the coding loop with guardrails derived from those decisions. The agent one-shots features once it goes into the coding loop.
Built a production API almost entirely AI-assisted recently. Didn't feel like leadership to me, felt more like code review at a much higher volume. I wasn't delegating decisions, just constantly checking the generated code actually did what I needed lol.
It’s designed to “feel” that way. It’s a computer program, not a person. People made it.
Don’t get fooled into treating these systems like they’re people or intelligent collaborators.
At the best of times these chat/agent/loop systems are gambling machines. Very expensive and heavily subsidized machines. Don’t forget to keep building your own skills. You never know when the bubble is going to pop and you don’t want to be left unable to do you job just because some big tech corp goes under.
How about “Orchestration”?
Careful with the Hegelian master and servant dialectic. It does not always turn out the way the parasites want it to.
Except nothing is at stakes, there are no failure modes beside shitty vibecode output, and noone is responsible for nothing.
Working with AI feels like dealing with a bureaucracy. There seems to be rules but they're negotiable and illogical. Rules change constantly for no reason. Decisions are capricious and unappealable. But you can sometimes get your way if you carry on enough. It works better depending on the time of day, or some days it doesn't work at all, and when it's not working you're entirely blocked without a workaround.
Just astounding we decided to put a DMV in our IDEs.
...and browsing HN these days feels more like cringing than informing oneself. This is a LinkedIn-level "insight", upvoted presumably because it's about "AI"?
It feels like managing an egregious liar who speaks exclusively in corporate pyschobabble, is completely unpredictable, makes you want to blow your head off in most interactions, but yet is regularly good enough and cheap enough that you can't justify not working with this person - despite wanting desperately not to.
What model are you using and in what harness? So I know how to avoid it.
Maybe if you weren't really coding to begin with, or didn't enjoy it. There's a ton of "temporarily embarrassed CEOs" who seem to be unable to think about LLMs as anything but their employees, which is an incredibly limiting perspective in my view.
Anyone who's fallen in love with programming itself and doesn't see software production as a means to an end is not really likely to see things like this.
I see AI as an accelerator of implementing my own choices. I'm generally opposed to metaphors, designs or strategies which excessively anthropomorphize it; it seems completely wrong-headed and counterproductive.
For every choice I make it makes a dozen choices though.
If I were a stronger man, I would make a comment about dealing with people vs dealing with code and having to route around bugs that are both fundamental to the problem and a POV, along with some inherent in how I setup the program and simply someone else problem I've inherited to simply being burned into the hardware and will never be neatly papered over.
Bro finds out AI is not deterministic.
I don't have good leadership or good management or good people skills. I can use AI just fine as run of the mill software engineers.
Pointless, stupid article. Digital garbage, as garbage as LLM slop. So many words to say nothing.
Why are LLM users such poor leaders then?
To go anywhere serious you have to lead people, but even the ones who should be leading people are heads down talking to the LLM
Perhaps some of the people who got into traditional leadership roles got there due to factors besides their raw leadership abilities. LLMs are an equalizer, and with them, the inability to lead them simply becomes a "skill issue", for the lack of a better term.
An equalizer, yes. They lift up those with little skill, and they drag down the top talent.
This is a practical effect by which top talent is neutralized. Where they should lead people, they talk to models instead. ...now you have a generation of leaders that don't deal well with any real emotion or disagreement. They can steer but not lead -- many no longer believe in the value or efficacy of leading people.
Of course you are also right to point out that many people in "traditional leadership roles" have been content to sit back and steer rather than heading of the charge. I still wish to command both talents: steering and leadership. I refuse to flatten myself out to fit in better, and I always have.
"Not predictable" is cope for the true failing. "AI" is not reliable.
As for the purported upside of the random variability in output... seriously? Who would want an RNG in a C compiler's code generator. Or worse, its parser?
I agree with the OP's point. But if you've worked in a corporate environment, isn't that just how it is? Like how programming work and getting promoted to a manager are essentially different careers—they feel like fundamentally different things.
They're different kinds of work, and both are interesting in their own way. But in the freelance market, LLMs have already become the baseline, so I have to use them whether I like it or not. There are both pros and cons.
It's good to be able to read code and understand its structure, but writing code and reading it to transform it into a different structure are different skills. There's definitely some decay in raw coding ability, though. So I use LLMs for professional coding and for tasks that I couldn't do before, while I keep hand-coding smaller things that feel manageable.
Honestly, I think most people who hate LLM coding actually hate being forced to use it under workplace pressure. And when LLM output looks bad, it's often because managers tend to be strict about their subordinates' work but lenient about their own. Once an LLM generates something, people tend to get attached to it and become more forgiving—since it feels like they made it.
It's tough that LLMs have made deadlines tighter. But these days, compared to the old days when I had to go through interviews and conversations to build a proposal, I actually find it more convenient that clients send me proposals written by LLMs. There are pros and cons to everything.
Yes, the Führerprinzip. Slop coders will meet their Stalingrad soon due to technical debt and so will AI corporations due to financial debt.
I've been coding professionally for almost 30 years and it's only really in the last few months I've started to "trust" AI to do a better job than I can within given time constraints - but with a big caveat - as the article points out, the AI needs to be kept on track and questioned about its approach, told about inconsistencies etc. To aid with that it helps to have a lot of Cursor rules that guide the approach - built up over time, mostly based on seeing what AI didn't get right.
Well that's at work - but I've actually taken a different approach for my own personal project and generally now allow AI to do things in its own way - I put my effort into checking the functionality rather than the code itself. This does result in some fairly sprawling code and some bits that are practically now only really maintainable/"understandable" by AI. Functionality is still fine though, which is what really matters there when there's no other team members and understanding every part of the code isn't critical. Guess in future it may just self-refactor it without even being prompted.
It’s a management of management position I’d say. You’re not just managing AI, if you were already in a management or lead type role then now you’re managing people, the LLMs they manage, and any LLM you’re managing.