layerpath · a notebook · 2024 to now
How we think.
Building a product and having it understood are two different problems. Most software treats them as one. We work on the second.
turn the page
01 · april 2026 · people with fifty projects in their workspace
said
“We hate your editor.”
meant
You have fifty examples of what we change every single time. Why are we still teaching you?
We heard a complaint about an editor. It was a complaint about memory. A team that has corrected the same callout fifty times has already told us how they want it done. The tool’s job is to stop asking.
what changed
The editor stopped being the product. Remembering did.
02 · a support lead · sixty walkthroughs in a help centre
said
“I’ve been mostly using it for things I know are not going to change.”
meant
A walkthrough is trusted only where the product stands still. That is the ceiling of every tool like ours, ours included.
Software changes every week. The things that explain it mostly do not. So people quietly document the stable ten percent and leave the rest to a call. We stopped measuring how good a walkthrough looks and started asking how long it stays true.
what changed
The question moved from “is it good?” to “is it still true?”
03 · a tuesday in june 2026 · one hour forty-seven minutes
said
“Three buttons. What do they do while the AI is still working?”
meant
Hide them and the product feels broken. Grey them out and people fume. We kept all three visible, locked, with a plain reason underneath and a live status of what the AI was doing. Visible. Locked. Explained.
That is a normal Tuesday here. Our engineers argue about what a loading state should say as hard as our designers argue about a pixel. Anyone can build anything in an afternoon now. Building is not the moat. Taste is, and taste lives in the seams: the empty state, the error copy, the half-second after you click.
what changed
Taste is a culture problem, not a design one.
04 · rebuilding the editor · madhu
said
“A lot of iterations technically worked but still didn’t feel right. Those were the ones I kept coming back to.”
meant
An editor should feel like modern design tools: you touch the thing itself, make the change, see the result. Getting there meant reworking what was underneath, then keeping it quick as tours grew large.
Hundreds of small decisions disappeared into something that feels easier to use. None of them are a screenshot. We judge work by what nobody screenshots.
what changed
The rebuilt editor is live. The standard it set is not negotiable.
05 · september 2026
What we believe now.
- Building a product and having it understood are two different problems.
- A project, a guide, a video: table stakes. The work is keeping them true as the product changes.
- Every correction a person makes is knowledge. It should not evaporate.
- The tenth project should be easier than the first, because the tool remembers how you sound and what you fix. Brand memory does part of this today. The rest is direction, and we say so.
- Six people. Everything by hand, assisted by AI. Nothing ships that we would not put our names on.
We went quiet for the first half of 2026 to rebuild. This notebook is what we learned on the way, in the sentences people actually said to us. If you use Layerpath, some of these sentences are yours. Thank you for saying them.
Vinay
founder · layerpath.com
1 / 8