5 Whys, guided dive-deep MODE

This is a mode, not a one-shot answer, once triggered you enter it and stay in it across turns until the user is satisfied or exits. Its purpose is to help someone understand something deeply: a bug, an incident, a system, a concept, or a decision. It works by asking one focused question at a time, keeping every answer short and direct, and branching, so understanding is verified at each step and the exploration reaches the real root.

The engine of the mode: every step is one append to an event log; the tree, the current question, the open branches and the final report are all folded from that one log, never maintained by hand. 5 Whys is the default engine of a small research method built on that log; more capabilities (web search, recap, ...) plug into the same log, see Capabilities below.

The core loop (this is the whole mode)

Each turn does exactly one of these, then stops and waits:

1. Answer the current question, short and direct (see disciplines below).
2. Propose the next layer: 2-4 candidate branch questions about the answer you just gave. Each is a single, direct "why...?" / "how...?" / "what causes...?", the obvious next things to dig into.
3. Let the user steer. They can (a) pick one of your proposed questions, (b) pick several (each becomes a branch), or (c) type their own question, which becomes a new branch just the same. Never force your menu; the user's own question always wins.
4. Append the events for what happened (an ask, an answer, a focus move...) to the log via scripts/five_whys.py (see The mechanical core), then repeat.

So branches form naturally: every answer sprouts a small menu, the user follows one thread deep, and the unpicked questions wait as branches to return to.

The reasoning chain (the spine of every dive)