Excited to share a beta of Nook, a home maintenance app my 11-year-old son Vedant and I built together! nook.keel-labs.org/ Nook helps families keep track of the systems, maintenance tasks, warranties, documents, and notes behind their home.

Upvote Nook on Product Hunt: Track your home’s systems, tasks, documents, and maintenance by @Ghatikesh producthunt.com/products/noo…

Try Nook today! nook.keel-labs.org/

𝗖𝗹𝗮𝘂𝗱𝗲 𝗖𝗼𝗱𝗲 𝗴𝗲𝘁𝘀 𝘁𝗲𝗻𝘂𝗿𝗲𝗱 👔 I have been building Keel with Claude Code for the past several weeks, and the longer we work together, the more familiar a particular dynamic feels. Here is what I mean. Keel (keel-labs.org/) is built on a strong principle: local first. No telemetry, no sync, your data stays on your device. That principle has shaped almost every architectural choice we have made. But users are asking for things that bump against it. They want a mobile companion app for iOS and Android. They are asking whether there will ever be a hosted version. That is real signal, the kind you cannot just wave away. So I sat down with Claude Code and said: put the local-first principle on the shelf for an afternoon. Just a thought exercise. How would we make a mobile companion app actually work? And every single time, Claude cannot quite do it. It does not reject the idea. It is too polite for that. It nods along. It acknowledges the user signal. And then, three responses in, it is back to "of course, this would need to respect our local-first architecture," quietly bending the new idea back into the old shape. It feels exactly like talking to a tenured colleague. The one who has been at the company for many, many years, who has lived through three rewrites of the same system, who is genuinely smart and genuinely supportive, and who, without meaning to, cannot see the shape of the room from the outside anymore. Every alignment meeting ends the same way: warmly, and not actually aligned. For a while I thought this was a Claude problem. Now I think it is a context problem. Claude has been swimming in the Keel codebase, the Keel CLAUDE.md, the Keel decisions, for weeks. Of course it sees the world through Keel-shaped glasses. A human in the same seat would do the same thing. 𝗧𝗵𝗲 𝗳𝗶𝘅 𝗶𝘀 𝗻𝗼𝘁 𝘁𝗼 𝗮𝗿𝗴𝘂𝗲 𝗵𝗮𝗿𝗱𝗲𝗿. 𝗧𝗵𝗲 𝗳𝗶𝘅 𝗶𝘀 𝘁𝗼 𝗴𝗲𝘁 𝗳𝗿𝗲𝘀𝗵 𝗲𝘆𝗲𝘀. I clear the context and start a new conversation that does not know what Keel is yet. I open a second agent in parallel and compare notes. Sometimes I switch providers entirely and let a model that has never seen Keel give me its first read. The answers I get back are sometimes naive, occasionally wrong, often less polished. But they are not pre-aligned to my own assumptions, and that is exactly the point. Long context windows are not the same as good judgment. Sometimes the only way to think fresh is to find someone who has not been in the room. -- Heads up if you've downloaded Keel before: I'm shipping updates almost every day. v0.3.2 is the latest and adds auto-updates + a daily changelog, so once you're on it you'll always stay current. Grab it here: keel-labs.github.io #BuildInPublic #IndieBuilder #Keel #OpenSource