Interaction design

Threads are carved, not entered

your stream the card: path + dead ends the doc
the conclusion is clean because the path was kept

Every chat tool asks you to manage threads: create one, name it, remember to switch into it, remember which one holds what. That's asking you to be the librarian of your own thinking — in the middle of the thought. We think a thread should be something the system carves out of your stream, not a room you have to step into.

You talk; Doric carves

In Doric you work in one stream. When a tangent grows a spine — when it stops being a remark and starts being its own line of work — the system notices and proposes peeling it off into its own lane. The content moves; you keep talking. You never had to decide, three messages in advance, that this was going to deserve its own thread, because that's a judgment nobody can make three messages in advance. The structure follows the thinking, instead of the thinking having to follow the structure.

This inverts the usual contract. Elsewhere, threads are containers you fill. Here, threads are shapes the work turns out to have had — discovered, named, and carved after the fact, while your stream stays unbroken.

Two artifacts, two jobs

Carving produces two things, and the discipline is refusing to let them blur.

The doc is the conclusion — stripped of the path. It's what you'd hand a new teammate: what we decided, how it works, no archaeology required. Every wrong turn has been edited out, not because the wrong turns were shameful, but because a conclusion that still carries its own scaffolding is unreadable.

The card is the path — dead ends included. The thread's record keeps the options you weighed, the alternative you almost chose, the reason you rejected it, the experiment that failed and redirected everything. None of that belongs in the doc. All of it belongs somewhere.

Six months from now, the question is never "what did we decide?" — the doc answers that in ten seconds. The question is "did we consider X?" — and only the path can answer it.

Provenance is the point

Dead ends aren't waste; they're the reason the conclusion deserves your trust. A decision with no visible discards is indistinguishable from a guess. When the card shows that four options were weighed and three were rejected for stated reasons, the surviving option carries the weight of everything it beat. That's provenance — and it's the difference between a document that says "we use approach A" and a project that knows why it uses approach A.

This is also what your AI team reads. When Phoebe or Theo grounds before answering, the docs give them your product's present tense — but the cards give them its reasoning. An agent that can see the dead ends doesn't re-propose them. An agent that can only see conclusions will cheerfully walk you back down a road you already abandoned, because nothing told it the road was closed.

The rule: conclusions are stripped of the path so they stay readable; paths are kept whole — dead ends included — so conclusions stay accountable. One without the other is either archaeology or amnesia.


Related: Why threads rot →

Keep the conclusion. Keep the path.

Start building free