Having see what terminal vibecoding looks like (to the point where customers say "fix your app" during renewal conversations), I don't think this is likely to happen. There is definitely selection pressure being applied to SaaS companies and I would not expect people not directly responsible (PMs, sales, etc.) to be willing to accept responsibility for technical outcomes; after all they are product, not software experts.
It is possible this leads to a decrease in salary (and positions) but I do not believe the social commentary will pan out in the manner the author proposes. The people who most argue for vibe coding will themselves never accept responsibility for the technical outcomes.
> The people who most argue for vibe coding will themselves never accept responsibility for the technical outcomes.
This is right, and don’t think this isn’t all partially fueled by spite. I’m not sure if engineers understand how much they’ve been simultaneously reviled/revered by non-technical people. They see this as a Prometheus moment. They would love to vibecode a mess and make the engineers deal with the details.
I guess to use the car analogy, if the tech companies said they've made a robot that can fix your car, a car-ignorant will think the robot is all-capable and works as good as a mechanic.
A lot of people don't know what writing software involves, and believe "AI can do it now!".
And when they point out the rare exception it’s always ”well, as a UX designer/product manager/etc I spent thousands of hours now working with AI to build products and if I don’t understand something then I ask the AI to teach me” - congratulations you haven’t replaced developers you just became one
Take away the maximalist points and this post rings true.
> It is now possible to vibecode your own SaaS instead of paying someone $50/month for theirs.
I have yet to see proof of this playing out this way.
You do have highly competent engineers vibing big things in record time and you have noobs one-shotting a prototype with Claude and then getting stuck.
I remain convinced that LLMs are an amplifier and you can't turn a zero in to something useful.
This of course barely changes how LLMs shifted much of programming from a mystical realm to top bike shedding territory, as Tereze here points out.
Very, very good essay. I'd like to add one thing to the argument. It used to be the case that your software itself could be a sound moat; that's no longer the case except toward the high end, where vibecoding fails due to complexity. Now, sound moats are e.g. your data, your regulatory advantages, your established customer base, etc., and software is increasingly just a fungible component -- increasingly like, say, accounting: a back-office task to tick off.
Nice write-up... but with strong opinions that seem plausible, yet are highly debatable.
"The rise of the long tail" -> To my knowledge the 'long tail' was years ago a subject of many scientific business studies. The conclusion: proof was never found for this economic theory. And yes, the book of Chris Anderson (20 years ago!) was an attractive read that also seems plausible.
"The barrier to entry for software has fallen." -> This is the marketing mantra since 4GL and IDEs. Visual coding IDEs, so coding without knowing coding never worked out as marketers promised, Same with nocode things years later.
Truth is programming in a natural language is very hard, that's why we have programming languages. And the hard part was never programming, but problems solving and gathering requirements before programming. Or during programming if you are fan of the agile community. AI/ML technology is a great tool for solving some problems, but certainly many problems can not be solved with AI for coming years. AI can not replace people, but people who do not add much value are, have been, and always will be the first to encounter technology progress.
The billions companies spend on awareness, converson, retention etc will only grow and non vibe-coded software will continue to be the input (doesnt "become" marketing) into a very sophisticated go-to-market machine occupied by various players ( brands, publishers, agencies etc). Still this is directionally interesting if only aimed at SMBs SaaS.
This would probably became very true if we would stay with the current prices of 'AI'.
Will we? Open AI is not profitable. Anthropic says that they may have profitable quarter. If they will raise prices will it still be the case? If you can 'vibe code your taxes app' but it will require constant fixes every month and those fixes will cost you 50$ in tokens and it will not be bullet proof, does it makes sense anymore? Maybe just pay 50$ for subscription to similar software? Maybe Chinese companies will keep low prices and it will cost 2$ dollars instead of 50, but that only works if you are doing that to 'vibe code' your scheduling/to do app. If you are any serious company you have regulations and GDPA and ISO and you cannot sent you financial and customer company to Chinese deepseek provider.
And software need constant upkeep. OS update, API changes, libraries get obsolete, build system does not work anymore... etc. This is very apparent for me every time I am doing changes to my Flutter mobile app need an update: I basically need to spin up environment from scratch, then update half of packages, then update all APIs fro those packages because of changes and when I finally do the change, pipeline breaks and I cannot sign the android release. Last time I just gave up on that. Non software people think you can just install Claude and prompt your app. Which is true. But then things break. Data disappear. You do not have backup. Licence changes and you can't use new version of some tool. Binaries got renamed. APIs disappear. Domain is not reachable anymore. And so on and on and on...
Software companies are forced to use 'AI' too so speed of breaking changes will increase and you either have to pickup on those or pay someone to do that for you regardless if this will be 'AI' provider and tokens usage or SaaS.
In 90 there were people in my country selling PC parts on every corner. No there is maybe one or two in entire city and I did not visit none in maybe 10 years. There were a thing because you could just buy parts and build your new system. You still can but now you can just order online.
So sure 'vibe coding' is a thing now but I am not convinced it will be a thing in 10-20 years. Maybe it will be online service that will automatically write an application for you based on specification for few $ but as a user you expecting an outcome and do not want to be bothered by npm and node version necessary for that.
I’m curious about the mental model of people who think ai is extremely subsidised. This view is strange because GLM released a few weeks ago and it is confidently better than say GPT 5.
I think you are in for a rude shock - your expectation that there will come a reckoning where people are forced to content real prices of AI.. will never happen. It won’t.
And how you do know that GLM is not subsidised? Maybe Chinese government is burning money to destroy US economy by making them lose money in AI companies that are just furnace burning dollars. Unless you have some compelling argument it is just a speculation.
It's because GLM 5.2 is offered on many inference providers, including providers in the US. Those companies only make their money by charging for inference, and yet they seem to be doing quite well while charging the exact same prices as Z.AI / GLM.
In fact, there's a price war where some of the US inference providers are undercutting the pricing of Z.AI's own GLM hosting. Novita & AtlasCloud are both offering 8% and 5% discounts on GLM 5.2 respectively. GMICloud is charging 30% less - but getting so hammered with demand that it only has 80% uptime & 7 tokens per second, so you get what you pay for.
You can find a list of providers & their pricing through OpenRouter here:
Sure I am not disputing that. But we can't say that z.ai is profitable because they charge the same as other providers per token. There is also training cost. And reaserch cost. Just inference is not all. Maybe there are profitable with all of it, I am not saying they are not, I do not know but at the same time I highly doubt that.
Unless somebody will find a way to actually do training by just updating weights with new data you have to train from scratch and this is very costly process.
Saying that some lab is profitable because they are profitable doing inference while disregarding training cost is not fair assessment.
West china has cheap electricity via solar. Satellite images reveal large scale data center build outs. Untill now this electricity did not have much use as this region lacks major connectivity to ports and industrial regions.
> A non-mathematician can't even speak the language of mathematics. But in a lot of fields, people can evaluate your work with their own senses. They don't need any other apparatus.
This is an interesting point, but I think the reality is that in a lot of fields, the consumption experience is the only thing that's easy to judge.
People have always had opinions on things like the interface, the work flow, the latency, the landing page (as the article mentions). AI or not.
It is still hard to implement, and I think you'll find far fewer people with good opinions about implementation. As long as that remains true, SWE exists.
Whether the job market and salaries and all that remain, well, that is indeed dependent on whether people are vibe-coding away all their SaaS subscriptions and margins drop.
I find it really interesting that SWEs think this is a good essay. As an experienced marketer, I really disagree. It’s stupid.
The barriers to entry / “respect” angle is a weird one to go in. Obviously people will think they can do marketing if the main, qualified activity you use as an example is… /choosing colours on a landing page/. Marketers have way more important, difficult and “high-barrier-to-entry” tasks to do in order to succeed - and this suggests Tereza is totally unaware of them.
So is she trying to say that software engineers are being treated by the market like marketers?
“Writers, designers, and marketers all have wider pay distributions because the barrier to entry is low and the work is visible: anyone can try, everyone has an opinion, and only the proven best command a premium.” My sweet summer child, have you ever MET a software engineer? What the fuck do you think they do!?
Her paradox of choice implies that she’s ACTUALLY talking about software /companies/… but none of this is new information. Marketers are trained in positioning, segmentation and “finding the fit” - she doesn’t even make a point here!
The middle class of software won’t disappear. Again, she’s equating SaaS to be “all software” (as a marketer she should know better).
And finally, the most critical and undeniable evidence that Tereza is by NO means the “very decent marketer” she calls herself… “Sell services, not software”.
What? Functionally and positionally they are identical. One sells OUTCOMES. It doesn’t matter, as a marketer, whether your product is a service or a software. The product is just ONE of the 4Ps we’re trained to think about when making tactical decisions. Let alone strategic ones.
I’d really strong encourage anybody reading this interested in marketing to read a bit of Mark Ritson. This is the most irritating essay I’ve read in months.
software has always been marketing. You have to sell things for doing money. Does soemone remember in 90's, when star trek was very famous os2- warp ? was not that marketing ? nothing new under the sun ... we say here .
While entry barrier might be low, the learning curve become more steep.
E.g. recently I've been porting non-naive app to vibe-code app framework (from engineering managed to product managed).
While I was doing so I had to answer plenty of implication bearing questions but also ask for a very software engineering like pattern. E.g. I had to plan for MIME types unsupported by vendor or use stubbed adapter for the yet unavailable integration connector.
I pulled this straight from my head but boy oh boy I don't wish making this decision without any experience whatsoever.
I'd summarize current situation: building castles on the sand became easier than ever. Good luck with trying to become a tenant there.
> I don't wish making this decision without any experience whatsoever
Not really a show stopper. It’ll be slower but you can just ask the LLM to draft you an analysis of the options and trade-offs. Not quite the same as true experience but works fine for the most part. LLMs are pretty good at reasoning through architecture choices these days
Worst case you just try both paths. Code is cheap now so may as well leverage that as a strength
> Implication 1: Lower entry barrier makes software lower-respect field
Maybe? This one's kind of subjective. I'm sure there are some people who will feel this way, and many who won't.
Do you respect artists less, now that you can make AI images?
As for pay, it seems unlikely to me that the job title of "software engineer" is going to see a significant decrease in median wage as a result of AI. Though there may simply be fewer "software engineer" jobs and more "prompt engineer" jobs.
> Implication 2: Optionality changes the commitment to software products
It's not clear to me that the typical decision-making process the average company was using to choose (for example) project management software, is going to be significantly different in the AI era than before. "Let's use JIRA, since that's what everyone else uses."
Making decisions this way is low-risk, and lower-cost than the token cost of vibe-coding something custom. The analogy to dating apps doesn't work - dating apps reward searching far and wide for something perfect, whereas the business world rewards going with what you know and making decisions quickly.
> Implication 3: The middle class of software products will disappear
I don't believe the cost of software creation is approaching zero. People are taking this concept too far and too literally. First of all, obviously there are token costs. And secondly, obviously there is still a time and effort requirement involved in maintaining anything, even via vibe-coding. Most companies have absolutely no reason to prefer to incur these costs rather than simply paying the man his $50/month.
But thirdly, and probably most importantly, there's the inherent cost of merely being responsible for something. Like I wrote earlier, decisionmakers want to minimize risk. The mere fact of being responsible for something - of it being someone's fault if something goes wrong - is a dire political cost, which most business leaders try to avoid by buying external rather than creating in-house. The SaaS market isn't going anywhere.
> Implication 4: If you want to win, sell services, not products
Service automation is a fruit that has already been mostly squeezed by conventional software.
That is to say - the space of things that traditional software can't already automate, that LLMs would be capable of automating, and that LLMs would be reliable and efficient enough at to significantly move the needle on real productivity - is small.
(Ironically, software development is one of the few things in that space. Since when you automate software development you can also automate the creation of tests that (at least attempt to) validate the correctness of the software itself. Not so much for legal documents.)
Having see what terminal vibecoding looks like (to the point where customers say "fix your app" during renewal conversations), I don't think this is likely to happen. There is definitely selection pressure being applied to SaaS companies and I would not expect people not directly responsible (PMs, sales, etc.) to be willing to accept responsibility for technical outcomes; after all they are product, not software experts.
It is possible this leads to a decrease in salary (and positions) but I do not believe the social commentary will pan out in the manner the author proposes. The people who most argue for vibe coding will themselves never accept responsibility for the technical outcomes.
> The people who most argue for vibe coding will themselves never accept responsibility for the technical outcomes.
This is right, and don’t think this isn’t all partially fueled by spite. I’m not sure if engineers understand how much they’ve been simultaneously reviled/revered by non-technical people. They see this as a Prometheus moment. They would love to vibecode a mess and make the engineers deal with the details.
The people saying "anyone can build software now" often seem to mean "anyone can generate code now"... which is not quite the same thing
Indeed. Anyone could always build software. Some people want to do it without actually learning anything or doing any work.
There are many things I wouldn’t mind doing without actually bothering to do the work, so I suppose I can relate.
I guess to use the car analogy, if the tech companies said they've made a robot that can fix your car, a car-ignorant will think the robot is all-capable and works as good as a mechanic.
A lot of people don't know what writing software involves, and believe "AI can do it now!".
And when they point out the rare exception it’s always ”well, as a UX designer/product manager/etc I spent thousands of hours now working with AI to build products and if I don’t understand something then I ask the AI to teach me” - congratulations you haven’t replaced developers you just became one
The truth to me is that ai in some ways makes coding more accessible even if the code it produces is labelled as bad
[dead]
Take away the maximalist points and this post rings true.
> It is now possible to vibecode your own SaaS instead of paying someone $50/month for theirs.
I have yet to see proof of this playing out this way.
You do have highly competent engineers vibing big things in record time and you have noobs one-shotting a prototype with Claude and then getting stuck.
I remain convinced that LLMs are an amplifier and you can't turn a zero in to something useful.
This of course barely changes how LLMs shifted much of programming from a mystical realm to top bike shedding territory, as Tereze here points out.
Very, very good essay. I'd like to add one thing to the argument. It used to be the case that your software itself could be a sound moat; that's no longer the case except toward the high end, where vibecoding fails due to complexity. Now, sound moats are e.g. your data, your regulatory advantages, your established customer base, etc., and software is increasingly just a fungible component -- increasingly like, say, accounting: a back-office task to tick off.
I mostly agree, though I'd phrase it slightly differently: software is becoming less of a moat by itself but not necessarily less important
Nice write-up... but with strong opinions that seem plausible, yet are highly debatable.
"The rise of the long tail" -> To my knowledge the 'long tail' was years ago a subject of many scientific business studies. The conclusion: proof was never found for this economic theory. And yes, the book of Chris Anderson (20 years ago!) was an attractive read that also seems plausible.
"The barrier to entry for software has fallen." -> This is the marketing mantra since 4GL and IDEs. Visual coding IDEs, so coding without knowing coding never worked out as marketers promised, Same with nocode things years later.
Truth is programming in a natural language is very hard, that's why we have programming languages. And the hard part was never programming, but problems solving and gathering requirements before programming. Or during programming if you are fan of the agile community. AI/ML technology is a great tool for solving some problems, but certainly many problems can not be solved with AI for coming years. AI can not replace people, but people who do not add much value are, have been, and always will be the first to encounter technology progress.
The billions companies spend on awareness, converson, retention etc will only grow and non vibe-coded software will continue to be the input (doesnt "become" marketing) into a very sophisticated go-to-market machine occupied by various players ( brands, publishers, agencies etc). Still this is directionally interesting if only aimed at SMBs SaaS.
This would probably became very true if we would stay with the current prices of 'AI'.
Will we? Open AI is not profitable. Anthropic says that they may have profitable quarter. If they will raise prices will it still be the case? If you can 'vibe code your taxes app' but it will require constant fixes every month and those fixes will cost you 50$ in tokens and it will not be bullet proof, does it makes sense anymore? Maybe just pay 50$ for subscription to similar software? Maybe Chinese companies will keep low prices and it will cost 2$ dollars instead of 50, but that only works if you are doing that to 'vibe code' your scheduling/to do app. If you are any serious company you have regulations and GDPA and ISO and you cannot sent you financial and customer company to Chinese deepseek provider.
And software need constant upkeep. OS update, API changes, libraries get obsolete, build system does not work anymore... etc. This is very apparent for me every time I am doing changes to my Flutter mobile app need an update: I basically need to spin up environment from scratch, then update half of packages, then update all APIs fro those packages because of changes and when I finally do the change, pipeline breaks and I cannot sign the android release. Last time I just gave up on that. Non software people think you can just install Claude and prompt your app. Which is true. But then things break. Data disappear. You do not have backup. Licence changes and you can't use new version of some tool. Binaries got renamed. APIs disappear. Domain is not reachable anymore. And so on and on and on...
Software companies are forced to use 'AI' too so speed of breaking changes will increase and you either have to pickup on those or pay someone to do that for you regardless if this will be 'AI' provider and tokens usage or SaaS.
In 90 there were people in my country selling PC parts on every corner. No there is maybe one or two in entire city and I did not visit none in maybe 10 years. There were a thing because you could just buy parts and build your new system. You still can but now you can just order online.
So sure 'vibe coding' is a thing now but I am not convinced it will be a thing in 10-20 years. Maybe it will be online service that will automatically write an application for you based on specification for few $ but as a user you expecting an outcome and do not want to be bothered by npm and node version necessary for that.
I’m curious about the mental model of people who think ai is extremely subsidised. This view is strange because GLM released a few weeks ago and it is confidently better than say GPT 5.
I think you are in for a rude shock - your expectation that there will come a reckoning where people are forced to content real prices of AI.. will never happen. It won’t.
And how you do know that GLM is not subsidised? Maybe Chinese government is burning money to destroy US economy by making them lose money in AI companies that are just furnace burning dollars. Unless you have some compelling argument it is just a speculation.
It's because GLM 5.2 is offered on many inference providers, including providers in the US. Those companies only make their money by charging for inference, and yet they seem to be doing quite well while charging the exact same prices as Z.AI / GLM.
In fact, there's a price war where some of the US inference providers are undercutting the pricing of Z.AI's own GLM hosting. Novita & AtlasCloud are both offering 8% and 5% discounts on GLM 5.2 respectively. GMICloud is charging 30% less - but getting so hammered with demand that it only has 80% uptime & 7 tokens per second, so you get what you pay for.
You can find a list of providers & their pricing through OpenRouter here:
https://openrouter.ai/z-ai/glm-5.2#providers
Sure I am not disputing that. But we can't say that z.ai is profitable because they charge the same as other providers per token. There is also training cost. And reaserch cost. Just inference is not all. Maybe there are profitable with all of it, I am not saying they are not, I do not know but at the same time I highly doubt that.
Unless somebody will find a way to actually do training by just updating weights with new data you have to train from scratch and this is very costly process.
Saying that some lab is profitable because they are profitable doing inference while disregarding training cost is not fair assessment.
West china has cheap electricity via solar. Satellite images reveal large scale data center build outs. Untill now this electricity did not have much use as this region lacks major connectivity to ports and industrial regions.
Training is expensive, to keep selling tokens, they have to have the best model. They have to keep spending money on model training
A product or a platform could give them a moat, but they don't quite have that yet although I think they are looking to get there
It happens because people fail to account for economies of scale. Believing everything scales linearly is major flaw in the ways most people think.
> A non-mathematician can't even speak the language of mathematics. But in a lot of fields, people can evaluate your work with their own senses. They don't need any other apparatus.
This is an interesting point, but I think the reality is that in a lot of fields, the consumption experience is the only thing that's easy to judge.
People have always had opinions on things like the interface, the work flow, the latency, the landing page (as the article mentions). AI or not.
It is still hard to implement, and I think you'll find far fewer people with good opinions about implementation. As long as that remains true, SWE exists.
Whether the job market and salaries and all that remain, well, that is indeed dependent on whether people are vibe-coding away all their SaaS subscriptions and margins drop.
I find it really interesting that SWEs think this is a good essay. As an experienced marketer, I really disagree. It’s stupid.
The barriers to entry / “respect” angle is a weird one to go in. Obviously people will think they can do marketing if the main, qualified activity you use as an example is… /choosing colours on a landing page/. Marketers have way more important, difficult and “high-barrier-to-entry” tasks to do in order to succeed - and this suggests Tereza is totally unaware of them.
So is she trying to say that software engineers are being treated by the market like marketers?
“Writers, designers, and marketers all have wider pay distributions because the barrier to entry is low and the work is visible: anyone can try, everyone has an opinion, and only the proven best command a premium.” My sweet summer child, have you ever MET a software engineer? What the fuck do you think they do!?
Her paradox of choice implies that she’s ACTUALLY talking about software /companies/… but none of this is new information. Marketers are trained in positioning, segmentation and “finding the fit” - she doesn’t even make a point here!
The middle class of software won’t disappear. Again, she’s equating SaaS to be “all software” (as a marketer she should know better).
And finally, the most critical and undeniable evidence that Tereza is by NO means the “very decent marketer” she calls herself… “Sell services, not software”.
What? Functionally and positionally they are identical. One sells OUTCOMES. It doesn’t matter, as a marketer, whether your product is a service or a software. The product is just ONE of the 4Ps we’re trained to think about when making tactical decisions. Let alone strategic ones.
I’d really strong encourage anybody reading this interested in marketing to read a bit of Mark Ritson. This is the most irritating essay I’ve read in months.
Wholly agree.
software has always been marketing. You have to sell things for doing money. Does soemone remember in 90's, when star trek was very famous os2- warp ? was not that marketing ? nothing new under the sun ... we say here .
Yeah, but maybe only for the visible (productized) layer of software
Software has always been marketing! Because most software straight up doesn't work or does what has been advertised lmao
While entry barrier might be low, the learning curve become more steep.
E.g. recently I've been porting non-naive app to vibe-code app framework (from engineering managed to product managed).
While I was doing so I had to answer plenty of implication bearing questions but also ask for a very software engineering like pattern. E.g. I had to plan for MIME types unsupported by vendor or use stubbed adapter for the yet unavailable integration connector.
I pulled this straight from my head but boy oh boy I don't wish making this decision without any experience whatsoever.
I'd summarize current situation: building castles on the sand became easier than ever. Good luck with trying to become a tenant there.
> I don't wish making this decision without any experience whatsoever
Not really a show stopper. It’ll be slower but you can just ask the LLM to draft you an analysis of the options and trade-offs. Not quite the same as true experience but works fine for the most part. LLMs are pretty good at reasoning through architecture choices these days
Worst case you just try both paths. Code is cheap now so may as well leverage that as a strength
Says the marketing guy
> Implication 1: Lower entry barrier makes software lower-respect field
Maybe? This one's kind of subjective. I'm sure there are some people who will feel this way, and many who won't.
Do you respect artists less, now that you can make AI images?
As for pay, it seems unlikely to me that the job title of "software engineer" is going to see a significant decrease in median wage as a result of AI. Though there may simply be fewer "software engineer" jobs and more "prompt engineer" jobs.
> Implication 2: Optionality changes the commitment to software products
It's not clear to me that the typical decision-making process the average company was using to choose (for example) project management software, is going to be significantly different in the AI era than before. "Let's use JIRA, since that's what everyone else uses."
Making decisions this way is low-risk, and lower-cost than the token cost of vibe-coding something custom. The analogy to dating apps doesn't work - dating apps reward searching far and wide for something perfect, whereas the business world rewards going with what you know and making decisions quickly.
> Implication 3: The middle class of software products will disappear
I don't believe the cost of software creation is approaching zero. People are taking this concept too far and too literally. First of all, obviously there are token costs. And secondly, obviously there is still a time and effort requirement involved in maintaining anything, even via vibe-coding. Most companies have absolutely no reason to prefer to incur these costs rather than simply paying the man his $50/month.
But thirdly, and probably most importantly, there's the inherent cost of merely being responsible for something. Like I wrote earlier, decisionmakers want to minimize risk. The mere fact of being responsible for something - of it being someone's fault if something goes wrong - is a dire political cost, which most business leaders try to avoid by buying external rather than creating in-house. The SaaS market isn't going anywhere.
> Implication 4: If you want to win, sell services, not products
Service automation is a fruit that has already been mostly squeezed by conventional software.
That is to say - the space of things that traditional software can't already automate, that LLMs would be capable of automating, and that LLMs would be reliable and efficient enough at to significantly move the needle on real productivity - is small.
(Ironically, software development is one of the few things in that space. Since when you automate software development you can also automate the creation of tests that (at least attempt to) validate the correctness of the software itself. Not so much for legal documents.)
Ive gotten and seen some seemingly cathartic/hateful comments from blue collar workers about SWE being out of a job
[dead]