Great post on the democratisation wave. But I reckon there's a constraint you're underweighting: these products can't actually tackle distribution against foundational models.
The insight is right, creation got cheap. But here's what that actually creates: throwaway software. Feature-shaped tools, not businesses. Marc's movie recommender, Tobi's MRI dashboard-brilliant, shipped in hours. But they're stuck in this weird place. Too niche to be proper products, but facing platforms like Claude or GPT that'll just absorb that capability into their main interface. Why would anyone context-switch to a separate dashboard when the model just does it for you?
The YouTube parallel breaks down because YouTube was the distribution. These builders are shipping on platforms that are increasingly the competition. Unlike the 90s indie director era you mention, there's no distribution moat to build. The core models already own all the reach.
So the democratisation part is absolutely real. But what it actually democratises isn't businesses, it's throwaway software. Used once, then becomes a feature inside the model.
Creation didn't hit the wall. Competition against models with total distribution did.
Each app, can be likened to a little blog post. The model is no longer the competitor then. You used the model, to create something. Something unique. Something that has enough of niche-ness, unique-ness, incongruence, baked in, that it can draw some attention to itself... maybe just because its a little weird. The model output is too benign, too proper, too sanitized, to have that 'color' in it. This thing has the color. This thing has the personality inside of it. And thats why this 'blog post' (read app) attracts attention. People come and experiment around with its weird attributes. Their attention is monetized. The app does not even need to have utility per se. As the article mentions, its a medium of expression now. You come to it, to play your weird. Its unbelieveable what awaits us on the other side !!
What if the app is the new blog of 2005… The creator who has leveraged YouTube is essentially an entrepreneur, he doesn’t care about distribution moat so long as he is making ad money , selling his courses or merch ( affiliate and direct )
The YouTube analogy for software really matters for new founders: this shift isn’t just about tooling, it’s about entirely new markets. When app creation moves from “something only developers do” to “something anyone with an idea and an AI copilot can try”, you unlock products designed by people who actually live the problems. Teachers, doctors, shop owners, creators, niche communities. That’s where the next wave of small but durable businesses can emerge: tiny, expressive software “posts” that evolve into long‑term products instead of short‑lived content.
The YouTube analogy works, but there's an important distinction: YouTube's long-tail value came from consumption (watching niche videos), while software's value comes from utility (solving specific problems). This matters because the "throwaway software" critique misses the point - even software used once and forgotten represents captured value that previously didn't exist. The interesting question isn't whether these micro-apps become businesses, but whether they accelerate the feedback loop between identifying problems and solving them.
Truely it is. I am 50 and wrote 0 code in my entire life. But I just managed to make an educational APP within 2 weeks with the help of Gemini 3. However, there is still an obstacle from "everyone can build" to "everyone can publish". I found it still very tricky to publish my APP in APPstore or in Google Play. Too many steps to register a so-called "developer's account". And if you want to use LLM API or user profile, you also need to setup a back-end. That process is very long and full of uncertainties. Why can't there be an platform for everyone to publish all their apps, just as easy as uploading a video in Youtube?
A speaking German learning APP. I have issues connecting Backend on my computer. Error message: WARN: failed to read file: open /Users/..../supabase/functions/gemini-proxy/index.ts: no such file or directory
Error: entrypoint path does not exist (supabase/functions/gemini-proxy/index.ts)
error running container: exit 1
Try rerunning the command with --debug to troubleshoot the error.
But Gemini cannot fix the bug. It keeps trying to recreate file index.ts and keeps failing. I don't know why. And as I am not a software engineer, neither can I. I was stuck here and had to pass the bug to my engineer friend and wait until he has time to see the problem.
Thanks for your help. I feel I am learning a huge bund of new things in the past few days. I decided to keep a work log, writing down what I did and learned everyday working with AI (and learning from them). Here is the link of my log page. https://zhouzq.substack.com/ please come by and see where I am heading. =)
Software is becoming easier to build. That part is clear now.
Cursor, Claude Code, Replit, and similar tools have changed the speed completely. One person can now go from an idea to a working prototype without waiting for a full dev team.
But building is not the whole game.
YouTube made video publishing easy, but it did not make distribution easy. Most videos still got ignored. The winners understood audience, timing, packaging, consistency, and what people actually wanted to watch.
Software will follow the same pattern.
AI will make it easier for more people to build tools, dashboards, agents, automations, and internal systems. But the hard part will still be choosing the right problem, understanding the workflow, and getting the right people to use it.
I recently built a small GTM engine. The code was not the hardest part. AI helped with that. The harder part was turning a messy GTM workflow into something useful: ICP, lead scoring, personalization, follow-ups, reporting, and decisions.
That is the real opportunity now.
The tool is not the moat anymore. The workflow understanding is.
I worked in the movie business and saw those waves crash on the shore of the business, and it did allow for some great films, but slowly, it pushed the money out of the business and on to YouTube. Yes, anyone can make anything and no, you can't make money as the take rate on YouTube is really high.
The "YouTube moment" analogy only works if you ignore how discovery actually happens on YouTube. 80% of views flow through the recommendation algorithm, not organic search or creator marketing. So the real question isn't whether software gets democratized, it's whether the new gatekeepers (whoever controls the AI coding assistants) will funnel demand toward quality or just toward what keeps users locked in their ecosystem. What makes you think this time the platform won't optimize for engagement over utility?
They wont, because the monetization model is completely different. Youtube wanted viewers to spend time, because thats linked to revenue. Foundation models are already being paid, as an enabling layer. (whether they are paid enough or not is another debate). Their payment structure is decoupled from the attention of the users. Google analogy comes to mind. Google did not want you to stay on the page. But click off, as soon as possible. They were enabling information organization, not gatekeeping it. Ofcourse the monetization models evolve. As this one will too. But this is to state that many variations are possible in this one. Not just where engagement needs to be maximized.
THIS...This is the most important post you'll write in 2026. THIS Vibe Coding movement will change everything. THIS gives the POWER to the masses. To BUILD. To CREATE. To EXECUTE what they see in their mind, FINALLY! AppSumo and ProductHunt are about to EXPLODE.
Remote Engineers will be the job title of 2026 - they show up to your business, buid a PURE CODE SaaS to does everything you need, grab a check and move one.
iPhones will build apps that do exactly what you want, just ask Siri.
Alexa will be your best friend, running your schedule, life, business.
So much will change. It's starting. And it's a good thing.
This citizen development is such a huge unlock. The more people doing the work today start building their own solutions, the better those solutions are. We're doing this with agents today - everyone on the delivery team builds agents up to MVP, and they stay "coparents" with our ops team. The rate of development and applicability of the solutions is massively better
Great analogy — with software democratized and apps becoming a form of expressive content rather than traditional products, does the VC playbook shift? Are VCs now primarily platform investors (AI infra, agents, orchestration) rather than app founders? Or is there still room to back niche app creators within these platforms?
Great post on the democratisation wave. But I reckon there's a constraint you're underweighting: these products can't actually tackle distribution against foundational models.
The insight is right, creation got cheap. But here's what that actually creates: throwaway software. Feature-shaped tools, not businesses. Marc's movie recommender, Tobi's MRI dashboard-brilliant, shipped in hours. But they're stuck in this weird place. Too niche to be proper products, but facing platforms like Claude or GPT that'll just absorb that capability into their main interface. Why would anyone context-switch to a separate dashboard when the model just does it for you?
The YouTube parallel breaks down because YouTube was the distribution. These builders are shipping on platforms that are increasingly the competition. Unlike the 90s indie director era you mention, there's no distribution moat to build. The core models already own all the reach.
So the democratisation part is absolutely real. But what it actually democratises isn't businesses, it's throwaway software. Used once, then becomes a feature inside the model.
Creation didn't hit the wall. Competition against models with total distribution did.
Each app, can be likened to a little blog post. The model is no longer the competitor then. You used the model, to create something. Something unique. Something that has enough of niche-ness, unique-ness, incongruence, baked in, that it can draw some attention to itself... maybe just because its a little weird. The model output is too benign, too proper, too sanitized, to have that 'color' in it. This thing has the color. This thing has the personality inside of it. And thats why this 'blog post' (read app) attracts attention. People come and experiment around with its weird attributes. Their attention is monetized. The app does not even need to have utility per se. As the article mentions, its a medium of expression now. You come to it, to play your weird. Its unbelieveable what awaits us on the other side !!
Great point by Jayesh, and almost completely invalidates the argument presented.
What if the app is the new blog of 2005… The creator who has leveraged YouTube is essentially an entrepreneur, he doesn’t care about distribution moat so long as he is making ad money , selling his courses or merch ( affiliate and direct )
The YouTube analogy for software really matters for new founders: this shift isn’t just about tooling, it’s about entirely new markets. When app creation moves from “something only developers do” to “something anyone with an idea and an AI copilot can try”, you unlock products designed by people who actually live the problems. Teachers, doctors, shop owners, creators, niche communities. That’s where the next wave of small but durable businesses can emerge: tiny, expressive software “posts” that evolve into long‑term products instead of short‑lived content.
The YouTube analogy works, but there's an important distinction: YouTube's long-tail value came from consumption (watching niche videos), while software's value comes from utility (solving specific problems). This matters because the "throwaway software" critique misses the point - even software used once and forgotten represents captured value that previously didn't exist. The interesting question isn't whether these micro-apps become businesses, but whether they accelerate the feedback loop between identifying problems and solving them.
Truely it is. I am 50 and wrote 0 code in my entire life. But I just managed to make an educational APP within 2 weeks with the help of Gemini 3. However, there is still an obstacle from "everyone can build" to "everyone can publish". I found it still very tricky to publish my APP in APPstore or in Google Play. Too many steps to register a so-called "developer's account". And if you want to use LLM API or user profile, you also need to setup a back-end. That process is very long and full of uncertainties. Why can't there be an platform for everyone to publish all their apps, just as easy as uploading a video in Youtube?
Just curious, what apps did you build?
happy to work with you to resolve some of your sses.
A speaking German learning APP. I have issues connecting Backend on my computer. Error message: WARN: failed to read file: open /Users/..../supabase/functions/gemini-proxy/index.ts: no such file or directory
Error: entrypoint path does not exist (supabase/functions/gemini-proxy/index.ts)
error running container: exit 1
Try rerunning the command with --debug to troubleshoot the error.
But Gemini cannot fix the bug. It keeps trying to recreate file index.ts and keeps failing. I don't know why. And as I am not a software engineer, neither can I. I was stuck here and had to pass the bug to my engineer friend and wait until he has time to see the problem.
Glad you are learning a new language and using LLMs to help build an app for it.
likely AI hallucinated or some missing dependency.
Please try with one of the following prompts to fix things
a)
"The gemini-proxy function is missing its index.ts file. Please regenerate the Supabase function folder structure and the index.ts file for me."
b) "Perhaps /gemini-proxy/index.ts is wrongly named. Please find the relevant index.ts (case insensitive) and use it"
Thanks for your help. I feel I am learning a huge bund of new things in the past few days. I decided to keep a work log, writing down what I did and learned everyday working with AI (and learning from them). Here is the link of my log page. https://zhouzq.substack.com/ please come by and see where I am heading. =)
Really enjoyed this post… that’s why we are doing what we are doing over at Replay.io (an a16z backed co)
We started by building a killer time travel code debugger, and now we’ve built a vibecoding tool wrapped around it.
Software is becoming easier to build. That part is clear now.
Cursor, Claude Code, Replit, and similar tools have changed the speed completely. One person can now go from an idea to a working prototype without waiting for a full dev team.
But building is not the whole game.
YouTube made video publishing easy, but it did not make distribution easy. Most videos still got ignored. The winners understood audience, timing, packaging, consistency, and what people actually wanted to watch.
Software will follow the same pattern.
AI will make it easier for more people to build tools, dashboards, agents, automations, and internal systems. But the hard part will still be choosing the right problem, understanding the workflow, and getting the right people to use it.
I recently built a small GTM engine. The code was not the hardest part. AI helped with that. The harder part was turning a messy GTM workflow into something useful: ICP, lead scoring, personalization, follow-ups, reporting, and decisions.
That is the real opportunity now.
The tool is not the moat anymore. The workflow understanding is.
I worked in the movie business and saw those waves crash on the shore of the business, and it did allow for some great films, but slowly, it pushed the money out of the business and on to YouTube. Yes, anyone can make anything and no, you can't make money as the take rate on YouTube is really high.
The "YouTube moment" analogy only works if you ignore how discovery actually happens on YouTube. 80% of views flow through the recommendation algorithm, not organic search or creator marketing. So the real question isn't whether software gets democratized, it's whether the new gatekeepers (whoever controls the AI coding assistants) will funnel demand toward quality or just toward what keeps users locked in their ecosystem. What makes you think this time the platform won't optimize for engagement over utility?
They wont, because the monetization model is completely different. Youtube wanted viewers to spend time, because thats linked to revenue. Foundation models are already being paid, as an enabling layer. (whether they are paid enough or not is another debate). Their payment structure is decoupled from the attention of the users. Google analogy comes to mind. Google did not want you to stay on the page. But click off, as soon as possible. They were enabling information organization, not gatekeeping it. Ofcourse the monetization models evolve. As this one will too. But this is to state that many variations are possible in this one. Not just where engagement needs to be maximized.
THIS...This is the most important post you'll write in 2026. THIS Vibe Coding movement will change everything. THIS gives the POWER to the masses. To BUILD. To CREATE. To EXECUTE what they see in their mind, FINALLY! AppSumo and ProductHunt are about to EXPLODE.
Remote Engineers will be the job title of 2026 - they show up to your business, buid a PURE CODE SaaS to does everything you need, grab a check and move one.
iPhones will build apps that do exactly what you want, just ask Siri.
Alexa will be your best friend, running your schedule, life, business.
So much will change. It's starting. And it's a good thing.
Maybe if you are a Roblox user.
This citizen development is such a huge unlock. The more people doing the work today start building their own solutions, the better those solutions are. We're doing this with agents today - everyone on the delivery team builds agents up to MVP, and they stay "coparents" with our ops team. The rate of development and applicability of the solutions is massively better
I kinda disagree. Too much of a doomsday scenario. Shared my thoughts at https://muboli.substack.com/p/agents-vs-enterprise-saas-why-disintermediation
YouTube was a significant shift in the way world evolved.
Love to see your insights on how YT as distribution medium applies to the LLM enabled software creation.
A future where kids aspire to be developers and *value creators* rather than content creators is a world I support.
Great analogy — with software democratized and apps becoming a form of expressive content rather than traditional products, does the VC playbook shift? Are VCs now primarily platform investors (AI infra, agents, orchestration) rather than app founders? Or is there still room to back niche app creators within these platforms?
When will it be time to build great stuff, like YouTube itself or even greater not just channels or videos on YouTube?