Second layer
Your assistant forgets.
The decision shouldn't.
Run more than one project through an AI assistant and you hit the same wall: the context fills, the conversation compacts, and what survives is a decision with no reason attached. Headup sits above the assistant and keeps the parts that matter.
Built for people running several things at once. Not a chat wrapper.
The problem, precisely
One working session, 22 distinct subjects, four properties. By the afternoon the reasoning behind a morning decision is gone — and a decision without its reason gets re-argued within the month.
Why separate chats don't fix it
They make it worse. One decision — close the source, sell exclusive licences — rewrote a marketing site, a repository, a partner pack and a price list. Four topics. One decision. Isolate them and three go stale silently.
What Headup does instead
Topics each keep their own session. Decisions live underneath topics and reach every one they actually touch — carrying the reason with them, and flagged until you've applied them.
Three things, and the third is the product
You create topics. Nothing is discovered.
Headup starts blank. You name a topic, point it at a directory, and paste in what is already true. It never scans your history, never guesses and never invents a topic you didn't ask for — because a tool that confidently infers the wrong thing is worse than one that asks.
Each topic owns a session.
Open a topic and it resumes exactly where it was. Session isolation stops being your problem because every topic is its own session, deliberately, from the start — and Headup holds the thread between them.
Decisions outlive the session and find every topic they touch.
Record a decision once, with its reason. Say which topics it affects — that is your judgement and Headup will not guess at it. Every one of those topics then carries it into its next session, marked NOT YET APPLIED HERE until you say otherwise.
That is the whole product. The rest is housekeeping.
What it refuses to do
Four deliberate refusals. Each one is a place tools like this usually go wrong.
- It won't guess which topics a decision touches. You tick the boxes. Fanning a decision out to the wrong places is worse than not fanning it out at all.
- It won't close an open loop on its own. What you're owed by someone else leaves the list when you say so, not when it looks finished.
- It won't re-summarise your conversations. The briefing carries facts, not narrative. Compaction already does that badly once; twice compounds the loss.
- It won't invent what it doesn't know. If a fact isn't there, a person types it. There is no third option.
Early access
Straight with you: Headup is new. It has no users yet and no case studies, and anything else on this page you can check. If you run several projects through an assistant and keep losing the thread, tell me and I'll let you know when it's ready for someone other than me.