The tools had crossed a line. An idea no longer needed a five-year engineering plan and a prayer — just a laptop, a decent question, and the willingness to keep going when things got weird.
So I did what a lot of people do with a new kind of creative power: I started too many things. Marketing tools. Creative tools. Half-products with logos and no reason to exist beyond a notes app and a fever dream. Some lasted a week. Some lasted a day. Some lasted until the architecture showed up wearing steel-toe boots.
First lesson: starting is easier now. Finishing is still very real.
What vibe coding actually is
You describe what you want, iterate with AI, and build by feel as much as by spec. It's fast, it's fun, and it can make you feel briefly omnipotent.
It does not remove product judgment. It does not make a messy idea elegant, or tell you what deserves to exist. AI gives you leverage. It does not give you taste.
The reward wasn't the product. It was learning to build.
I thought the prize would be shipping something impressive. The real prize was learning how to work with the tools: how to ask better questions, how to break an idea into pieces small enough to survive contact with reality, and how to notice when the model was confidently wrong — or when I was, which is ruder, frankly.
I learned "just one small change" is a cursed phrase. I learned every feature has a secret basement, and that engineering is architecture, tradeoffs, patience, and naming things. Sometimes the syllabus was one bug wearing a fake mustache for six hours.
Software humbled me properly
My background is marketing and digital — small e-commerce projects, digital products, brand experiments, the usual internet tinkering. Building software gave me a new respect for developers. Before this, I understood coding was hard in theory. Now I understand it in my bones. Developers are out here holding invisible buildings together with logic and caffeine. I salute you. I fear you.
UAT was the part I didn't expect to hurt
Testing the thing like a normal person might use it was humbling. You are not a normal user. You know where the bodies are buried, which button is temporary, which flow works if you approach it politely.
Other people don't approach software politely. They use it like real life: distracted, fast, uninterested in your internal logic. That forced me to stop thinking like someone making a product and start thinking like someone living inside a workflow.
The idea kept getting smaller. That was the clue.
For a while I tried to build tools that felt big. They were interesting. They weren't urgent. Then I noticed what I actually did every day: moving constantly between apps, AI tools, tabs, files, screenshots, prompts, and half-formed ideas. AI had made me faster, but it had also changed the shape of my work — I was collecting pieces, feeding them into tools, pulling outputs back, comparing, saving, losing, finding, carrying too much context by hand.
The work had become fluid. My desktop had not.
I needed a place for the little things still active in a task. Not ready for a folder. Not important enough for a project board. Just useful, still in motion, needed soon. A link. A color. A screenshot. A prompt. A draft.
That became Tansei.
Build from your own problem (harder than it sounds)
Everyone says it. But sometimes your own problem is so familiar it turns invisible — you assume the friction is normal, that everyone works this way. When I finally named it, I felt human relief. Not pitch-deck relief. The kind where your brain says: yes, that. Every good product needs some version of that little flame — not "this could be useful," but "I want this to exist."
Curiosity beats ego
My best results didn't come from sounding smart. They came from staying curious — asking the obvious question, the slightly embarrassing one, the "explain it again, but like I'm holding a broken toaster" one. AI doesn't replace learning. It gives you more doors into it.
What I'm making now
Tansei is a simple shelf for your Mac. It holds the small pieces of work you need nearby while you keep moving — the things you don't want to lose but don't want to organize yet. Add something. Move between apps. Use it when you need it.
I don't want it to become another system to manage. I want it to feel like a small native surface at the edge of your work. Quiet. Useful. There when the work moves.
Still learning — that's the honest part
I'm not pretending to be a traditional engineer. I come from digital, marketing, front-end taste, and a long obsession with how people use tools. AI made the work possible; it didn't make it easy. There's a difference. But I care about this product in a way I didn't care about the earlier ideas. It came from a real problem in the way I work — and I have a feeling I'm not the only one.
If you're building too: pay attention to what keeps pulling you back. Not every idea deserves to become a product. That's not failure. That's taste forming.
A few questions people keep asking
Can a non-developer build a real product with AI? Yes. Real doesn't mean perfect. It means useful enough that someone besides you would try it.
Why do so many AI-built projects never ship? Starting is cheap now. Judgment, testing, and finishing are not.
How do you know which idea to keep? The one that survives contact with your actual day — not the one that sounds best in a notebook.
What is Tansei? A local shelf for your Mac: links, screenshots, prompts, files, colors, and other small pieces of active work that aren't ready for a folder yet.
This didn't start with a giant idea. It started with a small one: I was tired of carrying context by hand.
↑ Back to top