Two ideas you surfaced last round.

A workshop has big initiatives over its lifetime.

You don't just "work on the agentic lab" forever. You work on it in big phases that have clear arcs:

Each mission is a thing you'd describe to Josh in one sentence as "the current focus." A mission lasts weeks-to-months, not days. It has a beginning and an end. New missions start when something genuinely new begins.

Add the mission number right after the workshop prefix.

You proposed this shape: AL1-0.2. Breaking it down:

So the workshop prefix and the mission number become one unit: AL1. Then everything after the dash follows the old pattern. Once mission 1 wraps and you start mission 2, every new chat starts with AL2-...:

Migration 02 was already doing this, kind of.

Your tags like 02.3.2 from the migration work are basically <mission-number>.<category>.<sub>. The 02 IS a mission number — it's the second-ever big migration (after some Migration 01 in your past).

You stumbled into this pattern without naming it. The mission-layer idea just makes it explicit, gives every workshop the same shape, and adds a workshop prefix in front.

Sometimes two kinds of standing chats want the same major number.

Right now, your major 0 means "orchestration." In practice that means one kind of seat: the orchestrator, who hands out work and tracks it. AL-0.1, AL-0.2, AL-0.3... each one a successor of the last (you baton-pass when context gets full).

But you raised: what if you eventually want a SECOND kind of always-on chat in major 0? Like a separate dispatcher seat that's distinct from the orchestrator? Both belong in "orchestration." Both want major 0. But they're different chains.

If both chains number from .1, you can't tell from a tag alone whether AL-0.3 is the 3rd orchestrator or the 3rd dispatcher. You need a way to disambiguate them.

Right now you're NOT hitting this. AL-0 is just orchestrator. So this might be a "park for later" decision. But it's worth picking an answer now so you have a plan.

Four ways to split a major into seat types.

I researched this. Here are the four shapes that came up, with their tradeoffs:

AL-0a.1   orch
AL-0a.2
AL-0b.1   dispatch
AL-0b.2

AL-0.orch.1
AL-0.orch.2
AL-0.disp.1
AL-0.disp.2

AL-0.1   role: orch
AL-0.2   role: disp
AL-0.3   role: orch
AL-0.4   role: disp

AL-0.1   = orch chain
AL-1.1   = disp chain
AL-2.1   = research?

My current vote is (b) named suffix. It's self-documenting (you read AL-0.orch.5 and instantly know "5th orchestrator chat"), no separate legend needed, and the AutoGen multi-agent framework uses the same pattern, so we'd be in good company.

Letter suffix (a) is my runner-up — shorter, scannable, but needs a "what's a/b/c" registry to be useful.

You don't actually have to decide this yet.

You're not currently running two seat-types in one major. So this is theoretical. Two reasonable ways to handle that:

(1) Pick now — bake one of the 4 options into the rule so future-you doesn't have to think about it when the case shows up. The rule says "if two seat-types share a major, use [the picked option]."

(2) Park it — leave the rule silent on the question. When you eventually hit it, you'll decide on the spot. The rule mentions "if this comes up, see [this open question]."