I'm not sure what counts as vibe coding, but however you want to define it, why assume it has peaked from one example of a messy codebase? It seems like an odd conclusion to draw.
I'm not sure what counts as vibe coding, but however you want to define it, why assume it has peaked from one example of a messy codebase? It seems like an odd conclusion to draw.
The reasons are stated in the article. These are the Makers of AI. They have unlimited access to their own product, dogfood it religiously, and should be more familiar and skilled with it than...
The reasons are stated in the article. These are the Makers of AI. They have unlimited access to their own product, dogfood it religiously, and should be more familiar and skilled with it than anyone else. If they can't hack it with that team, with one of the languages best represented in the training data, that casts doubt on the feasibility of it. Importantly, they are the ones making the claim that you can just throw AI at a codebase and have a usable result at the end.
To be brutally honest? This isn't just one messy codebase—it's one of the flagship products from what many may consider the leading AI foundry. It's a codebase where top minds are on the job, and this is the result.
I do think there are more productive ways to utilize AI, but those aren't what Anthropic claims is possible or what Anthropic is doing.
I'm not sure it follows that the people who make a tool will necessarily be the best at using that tool, especially for something with this much adoption. I believe that they are among the...
I'm not sure it follows that the people who make a tool will necessarily be the best at using that tool, especially for something with this much adoption. I believe that they are among the heaviest users of it, but that's not going to be perfectly correlated with output quality, especially in the very likely case that the best results with LLM coding tools come from people who don't delegate 100% of their thinking to them.
I 100% agree that the best results don't come from vibe coding. But this is the company that's claiming that vibecoding is feasible, dogfooding it, and as the heaviest and least limited users they...
I 100% agree that the best results don't come from vibe coding. But this is the company that's claiming that vibecoding is feasible, dogfooding it, and as the heaviest and least limited users they should also have plenty of trial-and-error and exploratory opportunities that outsiders won't. That's why we're talking about the ceiling of vibecoding and not the ceiling of LLM usage as a whole.
Not necessarily. But they are the ones trying to sell that reality. They need to advertise the idea of "you don't need engineers anymore", despite top minds being the ones driving this 100%...
I'm not sure it follows that the people who make a tool will necessarily be the best at using that tool
Not necessarily. But they are the ones trying to sell that reality. They need to advertise the idea of "you don't need engineers anymore", despite top minds being the ones driving this 100% vibecoded paradigm. Because that's what CEO's want to hear.
In reality, ,we'll eventually come to a consensus that a hammer is a hammer, and stop treating everything as nails. But we're not in a very rational market right now.
Amusingly (to me), this paragraph reads so strongly as though it were also written by Claude.
To be brutally honest? This isn't just one messy codebase—it's one of the flagship products from what many may consider the leading AI foundry. It's a codebase where top minds are on the job, and this is the result.
Amusingly (to me), this paragraph reads so strongly as though it were also written by Claude.
Examined grammar usage and... The user is pointing out that my grammar was typical of an LLM—cliché opening, "it's not x; it's y", em dash. I should acknowledge it simply and move on.
Examined grammar usage and...
The user is pointing out that my grammar was typical of an LLM—cliché opening, "it's not x; it's y", em dash. I should acknowledge it simply and move on.
Oh my God, my sister's new phone does this. She screenshotted it and I felt so dehumanized. She said she won't turn it off because she is a wonderful ass
Oh my God, my sister's new phone does this. She screenshotted it and I felt so dehumanized. She said she won't turn it off because she is a wonderful ass
I would not assume that because they make the AI models they would know best how to use them and make them useful. If anything, being too deep in AI qua AI probably makes you distinctly bad at...
I would not assume that because they make the AI models they would know best how to use them and make them useful. If anything, being too deep in AI qua AI probably makes you distinctly bad at understanding how to use AI for anything useful. These guys aren’t coming with expertise in the kinds of legacy burdened, highly interconnected, production environments most complex enterprises are running.
I think there’s still a lot of work to do on the UI and interface ends to make the most out of AI tools. And as those improve the models will need to adapt to function better with those new modalities for working with them.
That's what their stakeholders (and likely, every other businessmen in the US) wants to hear. Not that it's high quality, but that it can get stuff done fast and cheap. And we're in a market that...
If anything, being too deep in AI qua AI probably makes you distinctly bad at understanding how to use AI for anything useful.
That's what their stakeholders (and likely, every other businessmen in the US) wants to hear. Not that it's high quality, but that it can get stuff done fast and cheap. And we're in a market that is trying to be as fast and cheap without literally collapsing the floor it stands on (and IMO, the cracks are forming).
That's the big issue with this whole craze. The engineers are ignored (as usual), and this hype is starting to choke on its own fumes it hyped up.
You’ve listed some of Anthropic’s advantages. I believe they do have advantages. But I can think of other reasons why they might be at a disadvantage: Their engineers might be distracted by other...
You’ve listed some of Anthropic’s advantages. I believe they do have advantages. But I can think of other reasons why they might be at a disadvantage:
Their engineers might be distracted by other projects, like working on the next releases of their LLM’s. They might not have their full attention on Claude Code.
Anthropic’s engineers were the earliest adopters. They started using Claude Code before it worked very well, with earlier releases of their LLM’s that couldn’t write code very well. Being an early adopter means you find and report the bugs. Their messy codebase might be a product of these early experiments. Some other team that’s a fast follower could make a fresh start while learning from Anthropic’s mistakes.
A fair point. But another possibility is that maybe their top people aren’t working on Claude Code at all, because they’re busy doing something else? You can’t count “top minds are on the job” as...
A fair point. But another possibility is that maybe their top people aren’t working on Claude Code at all, because they’re busy doing something else? You can’t count “top minds are on the job” as an advantage if it’s actually the B team.
I know one person that works on Claude Code. They are very smart. I only talked to them before their start date. From the outside they weren't sure why the product seemed so buggy. Hopefully they...
I know one person that works on Claude Code. They are very smart. I only talked to them before their start date. From the outside they weren't sure why the product seemed so buggy. Hopefully they can give me some insight next time I run into them.
I don't think it's likely that Silicon Valley engineers getting paid $400-500k salary are a "B team". The AI researcher salary ranges go higher not because because they're smarter people. They are the ones that drive the value of the company more than the product engineers do.
The brutal honesty, too, has become a huge tell. Sometimes it's complete, sometimes it's just bare honesty. I tried to get a list of three going originally but couldn't squeeze it in. They also...
The brutal honesty, too, has become a huge tell. Sometimes it's complete, sometimes it's just bare honesty. I tried to get a list of three going originally but couldn't squeeze it in. They also love to assign their own words to external sources for more authority.
Idk these days I stopped focusing on vibecoding definitions or other people's opinions of which model is best. I just use w/e I think works and then read the code. Its that simple. There's too...
Idk these days I stopped focusing on vibecoding definitions or other people's opinions of which model is best. I just use w/e I think works and then read the code. Its that simple. There's too much noise out there.
I do understand the author's point but I also think its like a pretty obvious one. Any software engineer worth their salt knows that you can't really just prompt and reprompt, use loops and goals and then throw infinite money to get good software that thousands of people will use.
I do think that Bun rewrite is an exception and an exception that's only viable when you don't consider token costs. Most people aren't gonna be able to afford 16 Fable agents in parallel running for hours.
we're sadly getting less of those engineers, as some atrophy from over-relying on AI. and worse yet, students are overrelying on it because their metrics are also messed up (which is an entire...
Any software engineer worth their salt knows that you can't really just prompt and reprompt, use loops and goals and then throw infinite money to get good software that thousands of people will use.
we're sadly getting less of those engineers, as some atrophy from over-relying on AI. and worse yet, students are overrelying on it because their metrics are also messed up (which is an entire other discussions w';ll need to have after this bubble pops).
Sadly it's not software engineers (primarily) driving this hype bubble. The goal isn't to even appear productive, it's to yell how much AI we;re using and juicing the stock that way. It's only very recently starting to not be a magic money button.
I can't wait for the market to become rational. II just worry it won't be before a huge, very avoidable, correction.
Very much in agreement in both points. Re: over-reliance on AI - CEOs and execs are huge fans of it too. I mean in some ways I'm not surprised by this at all. Top MBA schools aint gonna teach you...
Very much in agreement in both points.
Re: over-reliance on AI - CEOs and execs are huge fans of it too. I mean in some ways I'm not surprised by this at all. Top MBA schools aint gonna teach you anything about code quality or craftmanship. Just hire juniors and give them Claude. I have literally met a few execs like this already. One guy was like "we have 4 staff level engineers, we don't need product, operations, or anything, they're building like crazy". I check LinkedIn and none had more than 4 years of experience. Again one can be good with just a few years, exp is not everything (also huge dif topic) but its such a weird self-affirmation thing to say when you're trying to convince the other person that you have a great team.
Re hype: I see this effect on X a lot (that's where I occasionally go to get "AI hype news"). I think its literally distasteful "AI" influencers. Every business person is now suddenly AI expert. There are AI "engineering" courses everywhere for a huge amount of money. Even influencers I like are not shitting on "AI slop". They'll talk about AI slop but they will never shit on AI slop by other influencers cuz that's not smart business I guess.
I'm not sure if there's gonna be a correction man that's actually one of the things I worry about the most. I really hope you are right and there is one. In the meantime I'm dealing with all this crap by putting all my own takes on craftmanship into a forum app. It aint gonna change the world but its what I can do to keep my sanity a bit lol. Thankfully a good amount of tech savvy users are starting to recognize AI slop but ultimately I think that is the only thing that can reverse the trend. Users must be vigilant in calling out AI slop which forces influencers to stop selling it. I try to do my part here.
There will definitely be a correction. But I cannot guarantee justice will be served and those pumping up the bubble will get their just desserts. There's a non-zero chance the government bails...
I'm not sure if there's gonna be a correction man that's actually one of the things I worry about the most. I really hope you are right and there is one.
There will definitely be a correction. But I cannot guarantee justice will be served and those pumping up the bubble will get their just desserts. There's a non-zero chance the government bails out such companies, despite almost all of them having he capital to take the hit. OpenAI seems to be the only one truly screwed, but we'll see how their IPO goes. Google and its like will be fine. Anthropic will be an interesting one to observe.
It also won't mean that AI drops off a cliff. In fact, the real battles will begin only after everyone isn't blinded by wild speculation. It's more that we should be much more mindful where and how AI is inserted and budgeted, and establish proper etiquette and regulations around it. The most ravenous defenders will fade out, because they will have moved on to finding the next hype train. hey go where the money goes.
It's a reflection of the US right now; it won't be fixed by one big inflection point. We just hope that such a point allows us to let adults in the room again and we can start to pick up the pieces.
The author writes with such authority, but doesn't evidence his claims or provide any credentials. Anyone can make assumptions about what might be happening behind closed doors, that doesn't mean...
The author writes with such authority, but doesn't evidence his claims or provide any credentials. Anyone can make assumptions about what might be happening behind closed doors, that doesn't mean there is value to doing so.
Also, he seems to be under the impression that the interface for Claude is the primary product. Is it not the models themselves that are the product and the interface is just the packaging? At the end of the day, I'd have thought it's the performance of the models that keep people coming back, not whether or not the UI is as polished as possible.
Maybe I'm missing something, but this really seemed like an opinion piece framed as a factual damning enditement of the AI industry.
This is correct. I’m not able to take down any giants with this blog post. But it’s my working theory. I’m happy to revise when I get new information. As I said elsewhere I’m likely to get some...
this really seemed like an opinion piece framed as a factual damning enditement of the AI industry.
This is correct. I’m not able to take down any giants with this blog post. But it’s my working theory. I’m happy to revise when I get new information. As I said elsewhere I’m likely to get some insider knowledge in the future which can greatly inform my opinions.
There are too many different strategies and techniques that get layered onto Claude Code to declare Anthropic’s results the ceiling. This is still an active field of experimentation and people are...
There are too many different strategies and techniques that get layered onto Claude Code to declare Anthropic’s results the ceiling. This is still an active field of experimentation and people are trying wacky new things all the time. I’m talking about stuff like Ralph Wiggum loops, caveman speak, and Gas City. Things Anthropic never dreamed up, that originated in the user community and are reshaping expectations about how to use Claude (and competitors). Some of these ideas get fed back into the product, others get dropped, and many live in an in-between space of contested “best practices” for harnesses, prompts, workflows, and the like. Eventually the cream may rise to the top but at the moment this is all still very much in flux. I don’t think it’s reasonable to assume the Claude Code team knows better than anyone else how best to use it.
I ... don't think that's a reasonable conclusion to draw, and even if it was, it wouldn't be the only one. I think it's reasonable to conclude that Anthropic sees no value in taking a position on...
Given the lack of such an announcement, I can only conclude Claude didn’t do as good of a job as they want us to believe.
I ... don't think that's a reasonable conclusion to draw, and even if it was, it wouldn't be the only one.
I think it's reasonable to conclude that Anthropic sees no value in taking a position on this or assessing quality. It gives them a chance to see whether the rust port dies on the vine or wins a market share, without having taken a position on it earlier. Much better for them to see the AI port become a success and claim a role in that success, or see it fail and just let it be rogue misuse of what will by then be an older version of their product.
I'm not sure what counts as vibe coding, but however you want to define it, why assume it has peaked from one example of a messy codebase? It seems like an odd conclusion to draw.
The reasons are stated in the article. These are the Makers of AI. They have unlimited access to their own product, dogfood it religiously, and should be more familiar and skilled with it than anyone else. If they can't hack it with that team, with one of the languages best represented in the training data, that casts doubt on the feasibility of it. Importantly, they are the ones making the claim that you can just throw AI at a codebase and have a usable result at the end.
To be brutally honest? This isn't just one messy codebase—it's one of the flagship products from what many may consider the leading AI foundry. It's a codebase where top minds are on the job, and this is the result.
I do think there are more productive ways to utilize AI, but those aren't what Anthropic claims is possible or what Anthropic is doing.
I'm not sure it follows that the people who make a tool will necessarily be the best at using that tool, especially for something with this much adoption. I believe that they are among the heaviest users of it, but that's not going to be perfectly correlated with output quality, especially in the very likely case that the best results with LLM coding tools come from people who don't delegate 100% of their thinking to them.
I 100% agree that the best results don't come from vibe coding. But this is the company that's claiming that vibecoding is feasible, dogfooding it, and as the heaviest and least limited users they should also have plenty of trial-and-error and exploratory opportunities that outsiders won't. That's why we're talking about the ceiling of vibecoding and not the ceiling of LLM usage as a whole.
Not necessarily. But they are the ones trying to sell that reality. They need to advertise the idea of "you don't need engineers anymore", despite top minds being the ones driving this 100% vibecoded paradigm. Because that's what CEO's want to hear.
In reality, ,we'll eventually come to a consensus that a hammer is a hammer, and stop treating everything as nails. But we're not in a very rational market right now.
Amusingly (to me), this paragraph reads so strongly as though it were also written by Claude.
Examined grammar usage and...
The user is pointing out that my grammar was typical of an LLM—cliché opening, "it's not x; it's y", em dash. I should acknowledge it simply and move on.
Oh my God, my sister's new phone does this. She screenshotted it and I felt so dehumanized. She said she won't turn it off because she is a wonderful ass
I would not assume that because they make the AI models they would know best how to use them and make them useful. If anything, being too deep in AI qua AI probably makes you distinctly bad at understanding how to use AI for anything useful. These guys aren’t coming with expertise in the kinds of legacy burdened, highly interconnected, production environments most complex enterprises are running.
I think there’s still a lot of work to do on the UI and interface ends to make the most out of AI tools. And as those improve the models will need to adapt to function better with those new modalities for working with them.
That's what their stakeholders (and likely, every other businessmen in the US) wants to hear. Not that it's high quality, but that it can get stuff done fast and cheap. And we're in a market that is trying to be as fast and cheap without literally collapsing the floor it stands on (and IMO, the cracks are forming).
That's the big issue with this whole craze. The engineers are ignored (as usual), and this hype is starting to choke on its own fumes it hyped up.
You’ve listed some of Anthropic’s advantages. I believe they do have advantages. But I can think of other reasons why they might be at a disadvantage:
Their engineers might be distracted by other projects, like working on the next releases of their LLM’s. They might not have their full attention on Claude Code.
Anthropic’s engineers were the earliest adopters. They started using Claude Code before it worked very well, with earlier releases of their LLM’s that couldn’t write code very well. Being an early adopter means you find and report the bugs. Their messy codebase might be a product of these early experiments. Some other team that’s a fast follower could make a fresh start while learning from Anthropic’s mistakes.
Those are different jobs. Anthropic has specific job postings for product engineers who would not be involved with LLM development.
A fair point. But another possibility is that maybe their top people aren’t working on Claude Code at all, because they’re busy doing something else? You can’t count “top minds are on the job” as an advantage if it’s actually the B team.
There’s a lot we don’t know from the outside.
I know one person that works on Claude Code. They are very smart. I only talked to them before their start date. From the outside they weren't sure why the product seemed so buggy. Hopefully they can give me some insight next time I run into them.
I don't think it's likely that Silicon Valley engineers getting paid $400-500k salary are a "B team". The AI researcher salary ranges go higher not because because they're smarter people. They are the ones that drive the value of the company more than the product engineers do.
In case you used the AI writing style in this post as some kind of snark or inside joke: I didn't get it.
Haha you're right
FTFY: Haha you're absolutely right.
The brutal honesty, too, has become a huge tell. Sometimes it's complete, sometimes it's just bare honesty. I tried to get a list of three going originally but couldn't squeeze it in. They also love to assign their own words to external sources for more authority.
Just the one paragraph. It certainly wasn't for any coherent reason so you're not missing anything.
Idk these days I stopped focusing on vibecoding definitions or other people's opinions of which model is best. I just use w/e I think works and then read the code. Its that simple. There's too much noise out there.
I do understand the author's point but I also think its like a pretty obvious one. Any software engineer worth their salt knows that you can't really just prompt and reprompt, use loops and goals and then throw infinite money to get good software that thousands of people will use.
I do think that Bun rewrite is an exception and an exception that's only viable when you don't consider token costs. Most people aren't gonna be able to afford 16 Fable agents in parallel running for hours.
we're sadly getting less of those engineers, as some atrophy from over-relying on AI. and worse yet, students are overrelying on it because their metrics are also messed up (which is an entire other discussions w';ll need to have after this bubble pops).
Sadly it's not software engineers (primarily) driving this hype bubble. The goal isn't to even appear productive, it's to yell how much AI we;re using and juicing the stock that way. It's only very recently starting to not be a magic money button.
I can't wait for the market to become rational. II just worry it won't be before a huge, very avoidable, correction.
Very much in agreement in both points.
Re: over-reliance on AI - CEOs and execs are huge fans of it too. I mean in some ways I'm not surprised by this at all. Top MBA schools aint gonna teach you anything about code quality or craftmanship. Just hire juniors and give them Claude. I have literally met a few execs like this already. One guy was like "we have 4 staff level engineers, we don't need product, operations, or anything, they're building like crazy". I check LinkedIn and none had more than 4 years of experience. Again one can be good with just a few years, exp is not everything (also huge dif topic) but its such a weird self-affirmation thing to say when you're trying to convince the other person that you have a great team.
Re hype: I see this effect on X a lot (that's where I occasionally go to get "AI hype news"). I think its literally distasteful "AI" influencers. Every business person is now suddenly AI expert. There are AI "engineering" courses everywhere for a huge amount of money. Even influencers I like are not shitting on "AI slop". They'll talk about AI slop but they will never shit on AI slop by other influencers cuz that's not smart business I guess.
I'm not sure if there's gonna be a correction man that's actually one of the things I worry about the most. I really hope you are right and there is one. In the meantime I'm dealing with all this crap by putting all my own takes on craftmanship into a forum app. It aint gonna change the world but its what I can do to keep my sanity a bit lol. Thankfully a good amount of tech savvy users are starting to recognize AI slop but ultimately I think that is the only thing that can reverse the trend. Users must be vigilant in calling out AI slop which forces influencers to stop selling it. I try to do my part here.
There will definitely be a correction. But I cannot guarantee justice will be served and those pumping up the bubble will get their just desserts. There's a non-zero chance the government bails out such companies, despite almost all of them having he capital to take the hit. OpenAI seems to be the only one truly screwed, but we'll see how their IPO goes. Google and its like will be fine. Anthropic will be an interesting one to observe.
It also won't mean that AI drops off a cliff. In fact, the real battles will begin only after everyone isn't blinded by wild speculation. It's more that we should be much more mindful where and how AI is inserted and budgeted, and establish proper etiquette and regulations around it. The most ravenous defenders will fade out, because they will have moved on to finding the next hype train. hey go where the money goes.
It's a reflection of the US right now; it won't be fixed by one big inflection point. We just hope that such a point allows us to let adults in the room again and we can start to pick up the pieces.
I see what you mean.
Man and so many people's career will be replaced. It's honestly scary to think about. Onwards to do our part then.
The author writes with such authority, but doesn't evidence his claims or provide any credentials. Anyone can make assumptions about what might be happening behind closed doors, that doesn't mean there is value to doing so.
Also, he seems to be under the impression that the interface for Claude is the primary product. Is it not the models themselves that are the product and the interface is just the packaging? At the end of the day, I'd have thought it's the performance of the models that keep people coming back, not whether or not the UI is as polished as possible.
Maybe I'm missing something, but this really seemed like an opinion piece framed as a factual damning enditement of the AI industry.
This is correct. I’m not able to take down any giants with this blog post. But it’s my working theory. I’m happy to revise when I get new information. As I said elsewhere I’m likely to get some insider knowledge in the future which can greatly inform my opinions.
There are too many different strategies and techniques that get layered onto Claude Code to declare Anthropic’s results the ceiling. This is still an active field of experimentation and people are trying wacky new things all the time. I’m talking about stuff like Ralph Wiggum loops, caveman speak, and Gas City. Things Anthropic never dreamed up, that originated in the user community and are reshaping expectations about how to use Claude (and competitors). Some of these ideas get fed back into the product, others get dropped, and many live in an in-between space of contested “best practices” for harnesses, prompts, workflows, and the like. Eventually the cream may rise to the top but at the moment this is all still very much in flux. I don’t think it’s reasonable to assume the Claude Code team knows better than anyone else how best to use it.
I ... don't think that's a reasonable conclusion to draw, and even if it was, it wouldn't be the only one.
I think it's reasonable to conclude that Anthropic sees no value in taking a position on this or assessing quality. It gives them a chance to see whether the rust port dies on the vine or wins a market share, without having taken a position on it earlier. Much better for them to see the AI port become a success and claim a role in that success, or see it fail and just let it be rogue misuse of what will by then be an older version of their product.
That’s fair. I agree it’s a weak argument and couldn’t hold up on a debate. I still wanted to share my thoughts. Take from them what you will.