I have an Obsidian vault full of ideas.

Some are good. Most are silly. A few are both.

And lately I've been shipping them, one by one, on top of an agentic setup I built for my side projects. Agents, orchestration, triggers, webhooks, the whole thing. The first one out the door was statuswall.fyi, a status page that looks like a stock market heatmap. Not a world-changing product. It just helped me set the base for everything else.

But something keeps happening, and it took me a while to notice the pattern.

I pick an idea from the vault. I get excited. I spend a weekend wiring it up. Prompts, agents, a little dashboard, a webhook that pings me when something is done.

Then I open X.

And there it is. Someone already built it. With a landing page, a waitlist, a demo video and 400 likes.

So I close the tab, archive the idea and move to the next one.

"Already done."

I've done this more times than I want to admit. And every time, it felt like the right call. Why build something that already exists?

Well, because "already exists" only means it already exists in my feed.

The Bubble

Most developers live in a bubble. I do too.

It's always been there, but AI made it way more obvious.

My timeline is people running ten agents in parallel. Claude Code here, MCP servers there, an orchestrator that talks to another orchestrator. Everyone is automating their automations. If you're not running a fleet of agents while you sleep, you feel behind.

And when you're inside that bubble, everything starts to look solved. Every idea has a repo. Every workflow has a template. Every "what if" already has a thread about it.

So you start thinking the market is full. Mine felt full for months, and I hadn't talked to a single person outside of it.

Then I close the laptop and go outside.

I live in Valle de Bravo. Small town, lots of small businesses, lots of people who work with their hands and their phones all day.

And man, the gap is huge.

Most of the people I talk to barely use AI. Some opened ChatGPT once, asked it something random and never came back. Some don't know what it is beyond "that thing that writes homework." A few are scared of it.

Nobody is running agents. Nobody knows what an MCP is. Nobody cares about orchestration.

But they all have problems. Small, boring, repetitive problems that eat their day. Stuff they do by hand every week because nobody ever showed them there's another way.

From inside the bubble, those problems look too simple to bother with. Something you'd build in an afternoon and never mention. Nobody on X would give it a like.

From their side, it's hours back every week.

Someone would pay for that.

Shoes for the Shoemaker

This is the mistake I kept making.

I was building for developers. For people who can build the same thing in an afternoon. People who already saw ten versions of it this week and will probably make an eleventh.

That's selling shoes to the shoemaker.

He doesn't need your shoes. He can make better ones. And even if he likes yours, he's not paying for them.

Meanwhile, the people walking around barefoot are right outside. They just don't hang out where I do.

The same "silly" idea I archived because someone on X already did it? The person who needs it has never seen that tweet. They never will. They don't follow AI accounts. They don't know that product exists, and if they found it, the landing page would scare them away with words like "agentic" and "workflow."

The idea isn't taken. It just hasn't reached anyone outside the bubble yet.

What I'm Trying Now

So I'm changing how I look at the vault.

Instead of asking "has anyone built this?" I'm asking "who outside my feed would actually use this?"

• Talk to people who don't care about AI. Not to sell them anything. Just to see how they work.
• Watch what they do by hand, over and over. That's where the product is.
• Build the simple version. No fancy stack. If it works on WhatsApp, even better.
• Put it in front of them and shut up. If they use it, keep going. If they don't, next idea.

Most of it won't work, probably. Still better than building another agent dashboard for people who already have three.

And I'm not saying stop building complex stuff. I love it. My agentic setup is the reason I can ship these small ideas in days instead of months.

But the setup is the tool, and the person on the other side doesn't need to see it. They just need the thing to get done.

It works, and they never have to know how.

The bubble is comfortable. Everyone gets your jokes. Everyone gets your stack.

The customers are somewhere else.

So next time you archive an idea because "someone already did it," ask yourself: did it for who?

Then go talk to someone who has never heard of it.