I can only speak for myself and will say that AI coding has gotten me more involved with personal project ideas I've let just sit for years. Mostly in that I'm able to be several times more productive than I am on my own.
I'm able to use the agent to scaffold small projects that let me work on a single library I can test in isolation. I'm not exactly "vibe coding" an entire application, but definitely using the AI to be more effective with my time.
The code was never the barrier, and that's why nothing has changed in the past years. Product and marketing are the hard part, complying to regulations is another, coding is easy unless you're working on extremely niche projects
There's certainly more weekend projects being built, but the vast majority of it is unabashed shovelware.
I see a torrent of poor ideas made a reality without enough thought put into them, designed without taste, and built without quality by people without the experience to maintain them.
Giving an audience something they never asked for is very easy when you don't have much experience interacting with them. That's where a lot of would be entrepreneurs and creators alike stumble. They forget (or don't know to) quantify the size of their obtainable market before taking action and building something.
Seeing as how LLMs just tell you what you want to hear and not what they think you need to hear I don't see this problem changing anytime soon. They might need to develop a different type of model to have it reason that way.
Everyone is talking about how many things they are building. Non-devs suddenly building... But nobody seems to call out the basic law of supply and demand.
You can be the greatest marketer but you will fail when all channels are flooded. Thinking your "taste" will save you is a false fallacy, most mainstream products suck and people still buy them. There's not an infinite demand for software.
It will eventually settle in some new market configuration. However devs shouldn't have broken their market by letting everyone in, was a stupid professional move.
Yes but back then, everybody just got bored of their Ataris and moved on.
Nobody is throwing out their phone or computer. Software will still be needed.
That said, there will be a lot of noise, with 100 choices in each category, how does one rise to the top? Is it simply the one that sticks around the longest and doesn't become abandonware?
I don’t see how this follows. If anything, ai tools let you be vastly more productive as a single dev. The limiter is and has always been the idea and your ability to market it.
> ai tools let you be vastly more productive as a single dev
Or to reframe, AI tools allow everyone to have some base-level software engineering competency and makes everyone you're competing against vastly more productive.
Generally speaking if you want to make money you need to provide a product/service within a niche. If too many people are able to compete against you, or just do whatever you're providing themselves, then you have no addressable market. It doesn't matter how productive you are.
This is why SASS valuations are crashing right now. While they might all be more productive their competition from people able to roll their own CRM and cheap alternatives from vibe-coders has increased exponentially.
Arguably any project that was previously small enough that a single dev could maintain it is now probably vibe-codeable in less than a day.
But by the same logic, people who could readily code small projects can now code large projects. It helps everyone, not just the non-programmers. There is still a large gap between what an amateur can do and what a professional can do, so the competitive advantage is still there.
It's a self-solving problem because people don't know how to market, which is 10x harder than development and which an AI cannot really help with, and that's how the real money is made.
And by the way it's SaaS, software as a service, Sass is a different tool, for CSS.
it's not base-level. it shifts the margin, at low skill base-level and your marginal level may be similar. but if you have enough experience and know the complete SLDC , having done each role during your career it, the margin is a massive wedge. if you know a niche, you now know all the adjacent niches and technolgies IF you are open minded to want to know them. that has powerful leverage.
No, the opposite is true. If you are building something good, you can do it even easier as a solo dev. The only people who will get shut down by all this is the people whose ideas were low quality, and whose only value was "I can code and you cannot."
I get the feeling that we are in the golden age of LLM dev tools, where many tools like Claude Code are very subsidized. That will go away soon, and worse. In the most recent Dwarkesh podcast with Dario, he mentioned that not all tokens should cost the same.
He gave an example where tokens giving advice to restart a computer should be cheap. On the other hand, tokens advising a pharma company on molecules should be expensive.
In the near future I can imaging CC replying:
> That's a great idea, this will solve the problem in production! However, given the number of users I see, this will be charged at 20x. Would you like me to proceed?
AI is a fantastic learning tool. Getting smarter usually leads to more success.
AI is great when you don’t have skilled co-workers to bounce ideas off of: as an indy mostly open source developer, for a few decades I would write up five to ten pages of design notes, perhaps a few diagrams like UML sequence diagrams, and then mull over my little pile of design artifacts for a while, and then start coding. Now I still create my little pile of design artifacts but will burn tokens on Gemini in research mode to critique my ideas. Very useful.
The problem with bouncing ideas off of AI is that you still need to know enough to know when something is likely a hallucination. Because unless you're double-checking it with some kind of regular cadence you're probably accepting fiction as fact. Its really easy to just trust everything these chatbots output because of the style of communication. I'll be the first admit I fall for this trap all the time.
I mean, it's his plans, so of course he'd recognize if AI is hallucinating.
For what is worth I have two different skills in claude code which are two reviewers with two distinct personalities.
At every plan I write, I have them review it and find edge cases, critique.
I don't think I've seen hallucinations since few opus versions at least.
Feedback is useful 4 times out of 5. Very useful actually. And one time is not very valuable or wrong (but it requires the two different skills to agree).
As much as people like to say, I don't think ideas are cheap. The best e.g., indie games didn't succeed just because of great execution but also great ideas and a good vision. I still think that holds true today. Happy building!
Exactly. The idea that ideas are cheap and all that matters is execution is absurd.
Imagine saying this about a great record album in music. Or any masterwork of art.
From growing up with my young brain being programmed by surrealist MTV videos, in a society driven by tiktok brains, creativity will be at an absolute premium.
Just the idea that this is a bad time for the solo developer is so uncreative that it boggles my mind but it is hardly surprising.
A huge part of implementation is how it’s done. A lot of companies made mp3 players, Apple made the iPod. Their implementation was different. It was done with more care and thought. The idea was cheap, 1,000 songs in your pocket. An implementation that connected with people and brought it to the mainstream was hard.
Someone quickly vibe coding something might fulfill the requirements of an idea. However, their implementation will likely be poor and lack the care needed to connect with people on a way that makes them want to use it.
I think understanding this has always been the key to standing out. That doesn’t change in the world of LLM, it becomes more important than ever.
There have been times in history when merely having the ability to turn an idea to software has been such a marketable skill you could fairly effortlessly get rich off it, like the early mobile app era saw some examples of that.
Those times were always pretty brief, and those markets were quickly saturated by people looking to make it rich. It was certainly not the state of anything in the years leading up to agentic coding.
Most people building with AI have zero taste and can't even read the code it outputs, meaning they are doomed from the start
The level of serious competition is surprisingly small. It mostly consists of people who could actually build without agentic LLMs but now get supercharged.
“Good at marketing” means a lot in the agentic coding era
You will do far better if you’re out there, talking to potential or current customers, understanding their needs deeply, and solving them.
It very much pays to be a solo dev, but you have to be very “heads up” and spend more time with humans. IMO it’ll difficult to get by just as a heads down pure dev.
Finally LLMs don’t know everything. Even if they knew everything, they don’t know your clients full situation. Moreover your client wants your specific n=1 opinions/experiences, not whatever average of ideas lives in the LLM.
Know your industries business well. You’ll do well coaching clients not just on the tech, but the intersection of tech and the gaming business. If you’re in a position you could critique an LLM on a topic, or offer a different perspective, that has strong market value.
If an app is to have qualiy then AI isn't nearly enough.
Directing the AI and using it to learn from is good.
Still, ideas about AI are currently ruining all sorts of markets (and being used as an excuse to lay off huge amounts of people, who then aren't going to be customers).
code quality and product quality is dropping fast. so your hand-crafted-actually-working-api-tokens-secure-in-a vault-on+backend-tests-cover-more-than-happy-path-app is raising in value actually...
It was already the case that the hardest part of a project was getting people interested enough that they invest time into trying it.
But now, the fight to get people's attention on new projects is going to get an awful lot harder, as the barrier to entry dropped,and that people who are actually good at marketing can make apps themselves.
Otherwise, as a solo dev, you can leverage AI for both coding faster, and getting better at marketing.
honestly the opposite has been true for me. I ship faster alone now than I did with a small team two years ago - the AI handles the boring scaffolding and I focus on the parts that actually matter. the real problem isn't that solo devs can't compete, it's that the bar for what counts as a product just dropped through the floor. everyone and their dog has a landing page and a waitlist now. so yeah you need marketing, you always did, but now you also need to not look like every other vibe-coded weekend project. I think the solo devs who were already good at taste and knowing what to cut will do fine. the ones who were just technically skilled but building things nobody asked for - AI doesn't fix that, it just makes it faster to build something nobody wants
Sorry to (not)disappoint, but it was _always_ like this.
If you analyze closely, every single solopreneurial success is based first and foremost on a successful marketing/sales. Whether by luck, hard work or being in the right place at the right time. The dev part and the product is never a main dish.
I can only speak for myself and will say that AI coding has gotten me more involved with personal project ideas I've let just sit for years. Mostly in that I'm able to be several times more productive than I am on my own.
I'm able to use the agent to scaffold small projects that let me work on a single library I can test in isolation. I'm not exactly "vibe coding" an entire application, but definitely using the AI to be more effective with my time.
The code was never the barrier, and that's why nothing has changed in the past years. Product and marketing are the hard part, complying to regulations is another, coding is easy unless you're working on extremely niche projects
There's certainly more weekend projects being built, but the vast majority of it is unabashed shovelware.
I see a torrent of poor ideas made a reality without enough thought put into them, designed without taste, and built without quality by people without the experience to maintain them.
Giving an audience something they never asked for is very easy when you don't have much experience interacting with them. That's where a lot of would be entrepreneurs and creators alike stumble. They forget (or don't know to) quantify the size of their obtainable market before taking action and building something.
Seeing as how LLMs just tell you what you want to hear and not what they think you need to hear I don't see this problem changing anytime soon. They might need to develop a different type of model to have it reason that way.
I'm expecting a similar situation to the Video game crash of 1983 which happened due to total market saturation with low quality products.
https://en.wikipedia.org/wiki/Video_game_crash_of_1983
Everyone is talking about how many things they are building. Non-devs suddenly building... But nobody seems to call out the basic law of supply and demand.
You can be the greatest marketer but you will fail when all channels are flooded. Thinking your "taste" will save you is a false fallacy, most mainstream products suck and people still buy them. There's not an infinite demand for software.
It will eventually settle in some new market configuration. However devs shouldn't have broken their market by letting everyone in, was a stupid professional move.
Yes but back then, everybody just got bored of their Ataris and moved on.
Nobody is throwing out their phone or computer. Software will still be needed.
That said, there will be a lot of noise, with 100 choices in each category, how does one rise to the top? Is it simply the one that sticks around the longest and doesn't become abandonware?
I don’t see how this follows. If anything, ai tools let you be vastly more productive as a single dev. The limiter is and has always been the idea and your ability to market it.
> ai tools let you be vastly more productive as a single dev
Or to reframe, AI tools allow everyone to have some base-level software engineering competency and makes everyone you're competing against vastly more productive.
Generally speaking if you want to make money you need to provide a product/service within a niche. If too many people are able to compete against you, or just do whatever you're providing themselves, then you have no addressable market. It doesn't matter how productive you are.
This is why SASS valuations are crashing right now. While they might all be more productive their competition from people able to roll their own CRM and cheap alternatives from vibe-coders has increased exponentially.
Arguably any project that was previously small enough that a single dev could maintain it is now probably vibe-codeable in less than a day.
But by the same logic, people who could readily code small projects can now code large projects. It helps everyone, not just the non-programmers. There is still a large gap between what an amateur can do and what a professional can do, so the competitive advantage is still there.
It's a self-solving problem because people don't know how to market, which is 10x harder than development and which an AI cannot really help with, and that's how the real money is made.
And by the way it's SaaS, software as a service, Sass is a different tool, for CSS.
it's not base-level. it shifts the margin, at low skill base-level and your marginal level may be similar. but if you have enough experience and know the complete SLDC , having done each role during your career it, the margin is a massive wedge. if you know a niche, you now know all the adjacent niches and technolgies IF you are open minded to want to know them. that has powerful leverage.
No, the opposite is true. If you are building something good, you can do it even easier as a solo dev. The only people who will get shut down by all this is the people whose ideas were low quality, and whose only value was "I can code and you cannot."
> you can do it even easier as a solo dev.
I get the feeling that we are in the golden age of LLM dev tools, where many tools like Claude Code are very subsidized. That will go away soon, and worse. In the most recent Dwarkesh podcast with Dario, he mentioned that not all tokens should cost the same.
He gave an example where tokens giving advice to restart a computer should be cheap. On the other hand, tokens advising a pharma company on molecules should be expensive.
In the near future I can imaging CC replying:
> That's a great idea, this will solve the problem in production! However, given the number of users I see, this will be charged at 20x. Would you like me to proceed?
Two mildly contrary points:
AI is a fantastic learning tool. Getting smarter usually leads to more success.
AI is great when you don’t have skilled co-workers to bounce ideas off of: as an indy mostly open source developer, for a few decades I would write up five to ten pages of design notes, perhaps a few diagrams like UML sequence diagrams, and then mull over my little pile of design artifacts for a while, and then start coding. Now I still create my little pile of design artifacts but will burn tokens on Gemini in research mode to critique my ideas. Very useful.
The problem with bouncing ideas off of AI is that you still need to know enough to know when something is likely a hallucination. Because unless you're double-checking it with some kind of regular cadence you're probably accepting fiction as fact. Its really easy to just trust everything these chatbots output because of the style of communication. I'll be the first admit I fall for this trap all the time.
I mean, it's his plans, so of course he'd recognize if AI is hallucinating.
For what is worth I have two different skills in claude code which are two reviewers with two distinct personalities.
At every plan I write, I have them review it and find edge cases, critique.
I don't think I've seen hallucinations since few opus versions at least.
Feedback is useful 4 times out of 5. Very useful actually. And one time is not very valuable or wrong (but it requires the two different skills to agree).
As much as people like to say, I don't think ideas are cheap. The best e.g., indie games didn't succeed just because of great execution but also great ideas and a good vision. I still think that holds true today. Happy building!
Exactly. The idea that ideas are cheap and all that matters is execution is absurd.
Imagine saying this about a great record album in music. Or any masterwork of art.
From growing up with my young brain being programmed by surrealist MTV videos, in a society driven by tiktok brains, creativity will be at an absolute premium.
Just the idea that this is a bad time for the solo developer is so uncreative that it boggles my mind but it is hardly surprising.
Ideas were “cheap” before because implementation was harder. Now I guess the coin has flipped.
A huge part of implementation is how it’s done. A lot of companies made mp3 players, Apple made the iPod. Their implementation was different. It was done with more care and thought. The idea was cheap, 1,000 songs in your pocket. An implementation that connected with people and brought it to the mainstream was hard.
Someone quickly vibe coding something might fulfill the requirements of an idea. However, their implementation will likely be poor and lack the care needed to connect with people on a way that makes them want to use it.
I think understanding this has always been the key to standing out. That doesn’t change in the world of LLM, it becomes more important than ever.
There have been times in history when merely having the ability to turn an idea to software has been such a marketable skill you could fairly effortlessly get rich off it, like the early mobile app era saw some examples of that.
Those times were always pretty brief, and those markets were quickly saturated by people looking to make it rich. It was certainly not the state of anything in the years leading up to agentic coding.
On the contrary, solo devs would produce even more output.
To the point where there's much more competition, no?
Most people building with AI have zero taste and can't even read the code it outputs, meaning they are doomed from the start
The level of serious competition is surprisingly small. It mostly consists of people who could actually build without agentic LLMs but now get supercharged.
“Good at marketing” means a lot in the agentic coding era
You will do far better if you’re out there, talking to potential or current customers, understanding their needs deeply, and solving them.
It very much pays to be a solo dev, but you have to be very “heads up” and spend more time with humans. IMO it’ll difficult to get by just as a heads down pure dev.
Finally LLMs don’t know everything. Even if they knew everything, they don’t know your clients full situation. Moreover your client wants your specific n=1 opinions/experiences, not whatever average of ideas lives in the LLM.
Know your industries business well. You’ll do well coaching clients not just on the tech, but the intersection of tech and the gaming business. If you’re in a position you could critique an LLM on a topic, or offer a different perspective, that has strong market value.
If an app is to have qualiy then AI isn't nearly enough.
Directing the AI and using it to learn from is good.
Still, ideas about AI are currently ruining all sorts of markets (and being used as an excuse to lay off huge amounts of people, who then aren't going to be customers).
code quality and product quality is dropping fast. so your hand-crafted-actually-working-api-tokens-secure-in-a vault-on+backend-tests-cover-more-than-happy-path-app is raising in value actually...
I'm not sure its a complete nail in the coffin but its definitely going to make things much harder
Lower barrier to entry = much more competition, lower prices, lower profit margins
Marketing has always been what makes or brakes a project.
You can (possibly) do more now or do new things now.
(And AI is also disrupting marketing if that makes you feel any better.)
This absolutely.
It was already the case that the hardest part of a project was getting people interested enough that they invest time into trying it.
But now, the fight to get people's attention on new projects is going to get an awful lot harder, as the barrier to entry dropped,and that people who are actually good at marketing can make apps themselves.
Otherwise, as a solo dev, you can leverage AI for both coding faster, and getting better at marketing.
honestly the opposite has been true for me. I ship faster alone now than I did with a small team two years ago - the AI handles the boring scaffolding and I focus on the parts that actually matter. the real problem isn't that solo devs can't compete, it's that the bar for what counts as a product just dropped through the floor. everyone and their dog has a landing page and a waitlist now. so yeah you need marketing, you always did, but now you also need to not look like every other vibe-coded weekend project. I think the solo devs who were already good at taste and knowing what to cut will do fine. the ones who were just technically skilled but building things nobody asked for - AI doesn't fix that, it just makes it faster to build something nobody wants
> unless you are good at marketing.
Sorry to (not)disappoint, but it was _always_ like this.
If you analyze closely, every single solopreneurial success is based first and foremost on a successful marketing/sales. Whether by luck, hard work or being in the right place at the right time. The dev part and the product is never a main dish.
Not really. It just doubles down on the concept that in software business, the code is the easy part, and that by itself, code isn't a moat.
You think that the force multiplier is going to make you less able to express your force?
I've made things with Claude I wouldn't begin to do on my own.
Okay I've not made any money out of it, surely you always needed to be good at marketing. Or did you mean advertising?
AI coding inverts the situation. Now solo entrepreneurs can get code for their projects and don't need a programmer.
Instead of programmers having to look for people with real world experiences to apply their skills to.