Arcs — what Thanks, Computer manages

Thanks, Computer runs stacks of rules that respond to events — this page names what all those events add up to.

Every platform has a unit of attention. Chat apps manage conversations. Web apps manage sessions. Ticket systems manage tickets. Thanks, Computer manages arcs: ongoing matters that open with an event, advance across channels and participants, and eventually resolve.

An invoice dispute is an arc. It opens when a customer’s email arrives. A classifier reads it, a service pulls the account, a human approves the credit two days later, a cron sweep nudges the ones still waiting, a payment webhook closes it out. Five channels, three kinds of participant, a week of elapsed time — one arc.

The shape of an arc

Most real operational work looks like this, and almost no software is built for it:

It spans channels. The same matter is touched by email, web, schedule, and an AI agent’s tool call — and shouldn’t care which.
It spans participants. Services compute, AI drafts, humans decide. All of them read and write the same shared document.
It spans time — and mostly waits. Minutes of compute inside weeks of elapsed time. The platform’s job is to wake the right operation when something happens, and otherwise cost nothing.
It resolves. An arc isn’t a stream to monitor; it’s a story with an ending. And it leaves a trace — the arc’s complete record, readable by a person or an AI.

An arc is not a workflow

A workflow is drawn in advance: boxes, arrows, every path anticipated. An arc unfolds. Each event — a mail arriving, a form post, a cron tick — is a beat that merges into the arc’s shared document, and the next step’s resonators read what’s there and decide what fires. The classifier doesn’t know a VIP handler exists; it emits .tier = "vip" and whatever resonates with that, runs. The path is discovered, not wired — which is why adding a new behavior to a live arc is writing one rule, not redrawing a diagram.

How the pieces map

Piece	Role in the arc
Event	A beat — one thing happening, on any channel
Shared document	The arc’s memory — every operation’s output merges here
Resonators	The arc’s attention — deciding which operations care about this beat
Op stack	The department’s playbook for arcs of this kind
Trace	The arc’s record — what happened and why

A stack handles many arcs at once, the way a support desk handles many cases: one playbook, many open matters, each at its own point in the story.

The ladder continues upward, too: arcs can belong to a saga — the longer mission a matter serves — so the why travels with the work. See Sagas.