I run a construction business, and my real to-do list lives in WhatsApp groups, voice notes and email. "Can you send the quote by tomorrow?" gets buried within minutes. I'm not a programmer, but with Claude Code I built Focus 100 to fix that, and it launched here on Product Hunt today.
What it does: on Android it reads notifications only from the apps you choose (WhatsApp, WhatsApp Business, Gmail and other mail), transcribes voice notes, and turns requests, promises and dates into tasks with a due date and a drafted reply. Things others promised you go to "Waiting". Nothing is ever sent until you tap Send.
I'm a construction contractor, not a programmer, and I've shipped an Android app with Claude Code. The single thing that made the difference for me wasn't a tool, it was writing a short spec before every feature: input, output, three example cases, and a list of things the code must NEVER do.
I'm curious how others who don't code for a living do it:
Do you write the spec yourself, or ask the model to interview you first?
- Do you keep a 'never do' list (e.g. never send data without the user tapping a button)?
- How do you check the result when you can't really read the code?
I asked myself this a few weeks ago and didn t like the answer. Months of product decisions, customer interviews, pricing experiments and investor feedback, all scattered across ChatGPT, Claude, Cursor and a couple of agents. Switching would mean re-explaining my own startup from scratch to each of them.
That s the problem we ve been working on, and tomorrow (Saturday, 21 Nov) we re launching AI Context Flow on Product Hunt. Here s the future we re building toward.
Picture a Monday morning a few years from now. Your agents worked through the night. One researched a competitor launch. One drafted proposals from the brief your client approved on Thursday, the latest version, not last month s.