The last 10% takes 90% of the work, and the last 1% takes 99% of the work and time. So sure you can get 80% of the way there, but that’s not the majority of the work. And you’re overestimating agency, we have had the internet for a long time, many choose to consume and a select few choose to produce and publish software and writing. There will always be a filter, the tools help but most people do not persist the roadblocks that come up.
Sure, I don't think everyone will do it. But it doesn't take that many to make a difference. It's more that the motivated people now *can* do it, not that everyone will.
There's also a softer version of this in which the composition of startup employees changes. Instead of most of them being mostly engineers, I suspect they'll be more balanced, where companies have one or two people who are very good at different functions: a couple marketers, a few engineers, and a few domain experts from whatever industry they're working in. The best companies could be ensembles like this, rather a bunch of engineers ramming at some problem with technology.
Sure it may decrease the activation energy for motivated people to get STARTED, but like everything, motivation dissipates and the vast vast majority of reasons why people fail is mostly due to self inflicted issues (not being able to fully execute on the roadmap, getting distracted and wanting to make more businesses vs scaling and doing the hard stuff in the current one, etc) vs competition. Ai helps as a productivity tool, but those who couldn’t do it before won’t be able to do it with ai either, they will filter out at some point and the filters only go up as ai raises the bar. I think it really helps on time and ease, but those who couldn’t before, may now be able to do it much more easily but a lot of learning and work does require discipline and the ai honestly I could even argue harm more than less given it doesn’t force your brain to think in the patterns that you would have to when you are forced to think. Those who think that “now it’s their time”, those people will be filtered out regardless. Now again we aren’t talking about someone who’s committed and explores ways to use the tool to improve their learnings and get over roadblocks, im talking about uncultured swine who produce ai music slop and think they are making music.
I have no comment or responses on how the companies will look like and be formed, that’s too speculative and doesn’t matter that much, given it’s also based on the founders and their goals and vibe. It’s like judging people, yeah they come in all shapes and sizes - it is what it is.
1. AI will create a lot of slop - not lazy, clickbait grift slop, but slop of people who are trying their hardest but aren't good at it. It's stuff that now can be made, but probably shouldn't.
2. But, viewed another way, I'd say a lot of stuff that got created pre-AI is that slop too. It can succeed, though, because not that many people can do it. Plus, those who do raise lots of money, and become very committed to the idea, so it sort of succeeds by brute force.
3. It's not that AI gives more people taste or talent; it just removes one of the former filters (technical skill; lots of money; etc) that prevented stuff from getting created. That means there will be a lot of slop, but it also means there will be more good stuff too.
Like, to make it in Hollywood, you have to be a good actor and good looking. If, for whatever reason, we suddenly developed a drug that made everyone look like a model, being good looking wouldn't matter anymore, and the quality of the acting would almost certainly go up, because that's the only thing people would get selected on. The same seems true here. To build a successful tech thing, you have to design it well and be capable of making it. If you remove the second filter, all we'll select on is the first.
My take is that it might just be the opposite extreme that will die, and the likes of YC will flourish. The other extreme end collapse is coming faster. I am referring to the bloated, monolithic, and outdated corporate machine that will die faster. They will implode internally at a faster rate. Previously, it was the fear of the startups that would disrupt them. It's still the same situation. This time, the other whammy is their unwillingness to embrace "plowing technology", using your words, to build their ideas faster. It's just not in the corporate culture to embrace the "one guy" vibe coding to build an idea. Yes, there are experts within the big corporation. However, they are too busy focusing on replying and answering stakeholders. I am sure corporate people are using ChatGPT and the like. I highly doubt that they will ever use Cursor, Bolt, and the like to accelerate their ideas. Perhaps the board of directors of these big corporations might adopt the notion of "one person unicorn". Many directors and just one CEO with hundreds or even thousands of AI Agents running in the background. Possible but unlikely.
Yeah, I don't think that's wrong, and the giant battleships that are hard to turn certainly seem less likely to respond to all of this than the smaller companies. My point is less that startups are in trouble, and more that startups that are run by the 22 year old stanford grad (which is the kind of the quintessential YC thing) lose a lot of their edge to the 40 year old coming from the big corporation that can do the startup thing, but knows a bunch of stuff that the kids don't.
The whole vibe coding phenomenon is super interesting and so incredibly in its infancy. There are examples of people leaving their api keys on production among other fantastic examples of why people are paid big bucks to write code. But the fundamentals have always been there, build quickly and test often. I think the Yc model has backed itself into a corner with this idea of prestige, and pumping money at a pitch deck. The way they talk about business is just dumb. Like just because some kid in SF built a web app and hooked it up to openai doesn’t mean you have a startup idea. It’s a project.
Yeah, I just had the thought that I think vibe coding is the kinda fun thing that shows up first, where lots of bad things like this happen and it kinda sorta gets dismissed as a toy. But it ultimately gets replaced with something more akin to "industrial coding," where the machines right the code, but in systems that are designed for that sort of thing.
You missed the whole point which is to get the attention of some people, get them to actually use the thing, talk to these customers and then iterate from there.
I assume you mean as the point of YC? Which, for sure, even if all this stuff happened, VCs and incubators can still do useful things (including being little credentialing badges). But who's best served by those useful things could change, from software engineering savants to like, corporate street smart savants.
For sure, I'm not saying that stuff isn't important. You can't build a company with Cursor. But you can experiment on an idea with it (and without a lot of engineers), which is a pretty big difference from where things were before.
AI is great for mockups and boilerplate code. It can do a lot of grunt work in just a minute or two. And all the scaling, fit and polish, security, proper signup pages, the marketing copy, localization, and real time statuses and notifications take months, if not a year to get right.
But you are correct that going from idea to clickable mockup or prototype in 48 hours to do a sanity check before you spend any more time on it is a great use of AI. If you casually played around with the idea for a few days, you avoid the sunk cost fallacy around your yak shaving.
And I think your grand insight is spot on with one little extra note. A lot of ideas for how Silicon Valley builds businesses — not apps but businesses — now revolve around rent-seeking parasitic middlemen rather than genuine innovation. If we want that innovation back, we have to make it easy to prove out original ideas and bypass the Tech Bro PayPal Mafia VC Industrial Complex…
I saw someone else (don't remember where) say that their hope was that all of the vibe-coding stuff would make those sorts of businesses less useful and unique, and would redirect a lot of venture money towards the bigger, more innovative problems. If everyone can build a new CRM in a week, then why build another new CRM? Why not try to do something bigger?
I'm not sure I'm so optimistic about it, but, yeah, maybe?
I’ve seen the Valley do too many dumb things to have much hope, but I sure hope that person is right, money will talk, and we’re going to see a major change in investment priorities for exactly that reason.
Venture capital is actually in dire straits these days and the concentration of AI startups makes for absurd levels of competition that will only increase startup failure rates.
It’s fun to witness on one hand by YC boosting AI startups is ultimately also a toxic symptom of the hype machine and lobbying that VC has become in America.
So hilarious that it’s such a brutal time to be a startup and yet there are more VC and founder related Newsletters now than ever before. I can’t get over this contraction. There’s really fundamentally a disconnect. But people in VC know the real score.
Why is it a brutal time to be a startup? I get why it's a bad time to be a VC (there are tons of them; lots of them are zombie funds that haven't gotten any sort of returns; probably lots of them are invested in overpriced AI companies; etc), but startups seem to be ok? Or is it because so many of them got overpriced and can't find obvious exits?
Yop:) we still have one advantage, the technical guys. Most of the "normal" world outside do not know anything about cursor, replit or lovable. Those tools are still unnoticed by 99.9% of human population. When they do get discovered and get better....that's when the software will become so cheap and booming as butter on bread for breakfast....
Yeah, and even then, knowing how it works a bit i'm sure makes you a lot more effective than not. It's not that technical skills don't matter, but the marginal value of it definitely seems like it's going down.
Partially agree, we will see what will be general adoption of those tools outside the tech community. It still takes some more exploration and building mentality to go and try the things to do it this way.
For sure, I'm sure there that a relatively small minority of people will ever use them. But a bigger number of people can build stuff than can today, I'd think.
To your point about the filter and barriers being removed, when everyone that wants to, can create their own product without having honed their technical knowledge, what kind of apps will people pay money for? Seems like the tools to build the app, like Cursor, will certainly thrive. What other kind of products will become indispensable in that world?
I'd think it's the same stuff as what thrives today, tbh. I don't think because more people can build a CRM or their own dropbox or whatever means people will *want* to build their own version; they'll still want to buy the best one. It's just that the best version could get built by someone else (or, there could be a few different versions that specialize in stuff, because they're cheaper to make).
Maybe sometime? It's a hot mess now, because I don't really know what I'm doing here, but there might be some useful core to it. I'm gonna keep futzing with it though, and see what happens.
This was a few months and versions of AIs ago - that last mile is getting easier to overcome I think.. iOS deployment might still be the toughest step though!
Oh nice, this is very cool. (Also, shoutout to an almost Ratatouille reference too.) And yeah, it really does seem like the "run this on the internet (or in the app store)" really is the most frustrating part of so much of this. It's not like it's more work than building the whole thing, but it definitely seems slower, or more opaque, or something.
The last 10% takes 90% of the work, and the last 1% takes 99% of the work and time. So sure you can get 80% of the way there, but that’s not the majority of the work. And you’re overestimating agency, we have had the internet for a long time, many choose to consume and a select few choose to produce and publish software and writing. There will always be a filter, the tools help but most people do not persist the roadblocks that come up.
Sure, I don't think everyone will do it. But it doesn't take that many to make a difference. It's more that the motivated people now *can* do it, not that everyone will.
There's also a softer version of this in which the composition of startup employees changes. Instead of most of them being mostly engineers, I suspect they'll be more balanced, where companies have one or two people who are very good at different functions: a couple marketers, a few engineers, and a few domain experts from whatever industry they're working in. The best companies could be ensembles like this, rather a bunch of engineers ramming at some problem with technology.
Sure it may decrease the activation energy for motivated people to get STARTED, but like everything, motivation dissipates and the vast vast majority of reasons why people fail is mostly due to self inflicted issues (not being able to fully execute on the roadmap, getting distracted and wanting to make more businesses vs scaling and doing the hard stuff in the current one, etc) vs competition. Ai helps as a productivity tool, but those who couldn’t do it before won’t be able to do it with ai either, they will filter out at some point and the filters only go up as ai raises the bar. I think it really helps on time and ease, but those who couldn’t before, may now be able to do it much more easily but a lot of learning and work does require discipline and the ai honestly I could even argue harm more than less given it doesn’t force your brain to think in the patterns that you would have to when you are forced to think. Those who think that “now it’s their time”, those people will be filtered out regardless. Now again we aren’t talking about someone who’s committed and explores ways to use the tool to improve their learnings and get over roadblocks, im talking about uncultured swine who produce ai music slop and think they are making music.
I have no comment or responses on how the companies will look like and be formed, that’s too speculative and doesn’t matter that much, given it’s also based on the founders and their goals and vibe. It’s like judging people, yeah they come in all shapes and sizes - it is what it is.
Ahh, so I think the effect is the inverse.
1. AI will create a lot of slop - not lazy, clickbait grift slop, but slop of people who are trying their hardest but aren't good at it. It's stuff that now can be made, but probably shouldn't.
2. But, viewed another way, I'd say a lot of stuff that got created pre-AI is that slop too. It can succeed, though, because not that many people can do it. Plus, those who do raise lots of money, and become very committed to the idea, so it sort of succeeds by brute force.
3. It's not that AI gives more people taste or talent; it just removes one of the former filters (technical skill; lots of money; etc) that prevented stuff from getting created. That means there will be a lot of slop, but it also means there will be more good stuff too.
Like, to make it in Hollywood, you have to be a good actor and good looking. If, for whatever reason, we suddenly developed a drug that made everyone look like a model, being good looking wouldn't matter anymore, and the quality of the acting would almost certainly go up, because that's the only thing people would get selected on. The same seems true here. To build a successful tech thing, you have to design it well and be capable of making it. If you remove the second filter, all we'll select on is the first.
I love the name stochastic pig
Great article.
My take is that it might just be the opposite extreme that will die, and the likes of YC will flourish. The other extreme end collapse is coming faster. I am referring to the bloated, monolithic, and outdated corporate machine that will die faster. They will implode internally at a faster rate. Previously, it was the fear of the startups that would disrupt them. It's still the same situation. This time, the other whammy is their unwillingness to embrace "plowing technology", using your words, to build their ideas faster. It's just not in the corporate culture to embrace the "one guy" vibe coding to build an idea. Yes, there are experts within the big corporation. However, they are too busy focusing on replying and answering stakeholders. I am sure corporate people are using ChatGPT and the like. I highly doubt that they will ever use Cursor, Bolt, and the like to accelerate their ideas. Perhaps the board of directors of these big corporations might adopt the notion of "one person unicorn". Many directors and just one CEO with hundreds or even thousands of AI Agents running in the background. Possible but unlikely.
Yeah, I don't think that's wrong, and the giant battleships that are hard to turn certainly seem less likely to respond to all of this than the smaller companies. My point is less that startups are in trouble, and more that startups that are run by the 22 year old stanford grad (which is the kind of the quintessential YC thing) lose a lot of their edge to the 40 year old coming from the big corporation that can do the startup thing, but knows a bunch of stuff that the kids don't.
The whole vibe coding phenomenon is super interesting and so incredibly in its infancy. There are examples of people leaving their api keys on production among other fantastic examples of why people are paid big bucks to write code. But the fundamentals have always been there, build quickly and test often. I think the Yc model has backed itself into a corner with this idea of prestige, and pumping money at a pitch deck. The way they talk about business is just dumb. Like just because some kid in SF built a web app and hooked it up to openai doesn’t mean you have a startup idea. It’s a project.
Yeah, I just had the thought that I think vibe coding is the kinda fun thing that shows up first, where lots of bad things like this happen and it kinda sorta gets dismissed as a toy. But it ultimately gets replaced with something more akin to "industrial coding," where the machines right the code, but in systems that are designed for that sort of thing.
You missed the whole point which is to get the attention of some people, get them to actually use the thing, talk to these customers and then iterate from there.
I assume you mean as the point of YC? Which, for sure, even if all this stuff happened, VCs and incubators can still do useful things (including being little credentialing badges). But who's best served by those useful things could change, from software engineering savants to like, corporate street smart savants.
I suspect she means all of the other aspects of company-building and product-building and customer-selling separate from raw development.
For sure, I'm not saying that stuff isn't important. You can't build a company with Cursor. But you can experiment on an idea with it (and without a lot of engineers), which is a pretty big difference from where things were before.
Agree 1000%
Two important takeaways for me.
AI is great for mockups and boilerplate code. It can do a lot of grunt work in just a minute or two. And all the scaling, fit and polish, security, proper signup pages, the marketing copy, localization, and real time statuses and notifications take months, if not a year to get right.
But you are correct that going from idea to clickable mockup or prototype in 48 hours to do a sanity check before you spend any more time on it is a great use of AI. If you casually played around with the idea for a few days, you avoid the sunk cost fallacy around your yak shaving.
And I think your grand insight is spot on with one little extra note. A lot of ideas for how Silicon Valley builds businesses — not apps but businesses — now revolve around rent-seeking parasitic middlemen rather than genuine innovation. If we want that innovation back, we have to make it easy to prove out original ideas and bypass the Tech Bro PayPal Mafia VC Industrial Complex…
I saw someone else (don't remember where) say that their hope was that all of the vibe-coding stuff would make those sorts of businesses less useful and unique, and would redirect a lot of venture money towards the bigger, more innovative problems. If everyone can build a new CRM in a week, then why build another new CRM? Why not try to do something bigger?
I'm not sure I'm so optimistic about it, but, yeah, maybe?
I’ve seen the Valley do too many dumb things to have much hope, but I sure hope that person is right, money will talk, and we’re going to see a major change in investment priorities for exactly that reason.
Ahhh venture capital hate, love it
When all YC’s startups are AI startups and you have less money to go around, it's not usually a great sign for innovation!
Fair, though it certainly seems like if you're an AI startup, there's no shortage of people willing to drown you in money.
Venture capital is actually in dire straits these days and the concentration of AI startups makes for absurd levels of competition that will only increase startup failure rates.
It’s fun to witness on one hand by YC boosting AI startups is ultimately also a toxic symptom of the hype machine and lobbying that VC has become in America.
So hilarious that it’s such a brutal time to be a startup and yet there are more VC and founder related Newsletters now than ever before. I can’t get over this contraction. There’s really fundamentally a disconnect. But people in VC know the real score.
Why is it a brutal time to be a startup? I get why it's a bad time to be a VC (there are tons of them; lots of them are zombie funds that haven't gotten any sort of returns; probably lots of them are invested in overpriced AI companies; etc), but startups seem to be ok? Or is it because so many of them got overpriced and can't find obvious exits?
Yop:) we still have one advantage, the technical guys. Most of the "normal" world outside do not know anything about cursor, replit or lovable. Those tools are still unnoticed by 99.9% of human population. When they do get discovered and get better....that's when the software will become so cheap and booming as butter on bread for breakfast....
Yeah, and even then, knowing how it works a bit i'm sure makes you a lot more effective than not. It's not that technical skills don't matter, but the marginal value of it definitely seems like it's going down.
Partially agree, we will see what will be general adoption of those tools outside the tech community. It still takes some more exploration and building mentality to go and try the things to do it this way.
For sure, I'm sure there that a relatively small minority of people will ever use them. But a bigger number of people can build stuff than can today, I'd think.
To your point about the filter and barriers being removed, when everyone that wants to, can create their own product without having honed their technical knowledge, what kind of apps will people pay money for? Seems like the tools to build the app, like Cursor, will certainly thrive. What other kind of products will become indispensable in that world?
I'd think it's the same stuff as what thrives today, tbh. I don't think because more people can build a CRM or their own dropbox or whatever means people will *want* to build their own version; they'll still want to buy the best one. It's just that the best version could get built by someone else (or, there could be a few different versions that specialize in stuff, because they're cheaper to make).
Awesome article I will need to get some ideas out of the old junk drawer and try this myself. Cheers!
It's pretty fun. And, at least for me, was, uh, revealing.
Indeed a lot of fun. I guess I should slowly write more here
well, can we try the composition assistant out? :)
Maybe sometime? It's a hot mess now, because I don't really know what I'm doing here, but there might be some useful core to it. I'm gonna keep futzing with it though, and see what happens.
I actually did this - I had an idea for app ages ago, but never came across a mobile developer who wanted to work on it with me. I work in AI (not as a developer) so was keen to see how capable they really are.. I wrote it up here - https://medium.com/@maxillis/me-my-app-and-my-ai-development-team-0b650384f71d
This was a few months and versions of AIs ago - that last mile is getting easier to overcome I think.. iOS deployment might still be the toughest step though!
Oh nice, this is very cool. (Also, shoutout to an almost Ratatouille reference too.) And yeah, it really does seem like the "run this on the internet (or in the app store)" really is the most frustrating part of so much of this. It's not like it's more work than building the whole thing, but it definitely seems slower, or more opaque, or something.
Maybe you just need tobtaje your idea and PIVOT
oh there was a pivot joke in this at some point, but it didn't make the cut https://www.youtube.com/watch?v=8w3wmQAMoxQ&t=80s
great and accurate
We're working on something special gAI ventures helping deep experts create agentic products that change the way people work.
Do they have to be agentic products though? Deep experts seem like they could also make pretty good analog saas products or whatever too.
As always Ben, spot on!