The fleet-learning loop, explained

A single edge gets better the more you run on it. A fleet of edges should get better the more the whole fleet runs — without any one device giving up authority over its own process. That is the loop TameChaos Cloud is built around, and it is worth being precise about what it does and what it doesn't.

Four steps, one direction

The loop runs in one direction, from the edges up to the cloud and back:

1. Oversee. A registry of your fleet — which edges exist, whether they're online, their last-seen time, software version, and active runs.
2. Aggregate like-processes. Group runs that instantiate the same process-class template, across edges, into comparable sets. This is the join key for everything downstream.
3. Improve models. Refine physics and inference models from that aggregated like-process data.
4. Distribute. Publish improved templates and models back to the fleet for edges to pull and adopt — never pushed, never commanded.

What ships today, and what doesn't

We would rather be clear than over-claim. Steps one and two ship today: fleet oversight and like-process aggregation are the v1 cloud surface. Steps three and four are the model-loop phase, on the roadmap — improving models from aggregated data and distributing them back for adoption.

Saying so matters, because the value of the loop is cumulative. Aggregation is useful on its own — cross-run comparison across your fleet is real today. The model improvement it unlocks is the next phase, and we will not market it as shipped before it is.

Per-fleet, by design

One more boundary worth stating plainly: aggregation and model improvement are scoped to a single owner's fleet. A fleet learns from its own operation — there is no cross-tenant data or model sharing. The loop makes your fleet smarter about your processes, not a shared marketplace of someone else's.

That scoping is the point. The edge stays authoritative for execution; the cloud aggregates only within your tenancy; and the loop closes — oversee, aggregate, improve, distribute — entirely inside the fleet you own.