Useful intelligence should be able to accumulate.

ORIAS is an AI research and engineering company. We work on systems where useful intelligence accumulates through interaction with complex environments — not intelligence that lives only inside a model and vanishes when the session ends, but intelligence that persists around a world, grows with evidence, and compounds.

The thesis underneath is simple to state and hard to test: useful intelligence should accumulate around a persistent world, rather than exist only inside a transient model.

Chimera is our first major programme, and the first serious test of this idea.

Chimera is persistent engineering intelligence for software.

Software is the first world because it has properties almost no complex environment offers. It can be inspected. Executed. Copied, changed, branched, replayed, tested, compared, rolled back. And often verified against exact evidence.

Today, every person and every model who touches a codebase reconstructs the same understanding from scratch: what the system does, what failed before, why it failed, and what the evidence actually supports. That understanding is built, used, and then lost — scattered across chat logs, pull requests, postmortems, and a single engineer’s memory.

Chimera keeps it, and tries to make that understanding compound.

We keep what exists, what is being built, and what is still research deliberately separate. We say “does” only when something can be demonstrated today. Some of the most important parts have not yet proven themselves.