JOIN

Work on the substrate for AI-authored work.

We need people who can make hard mechanisms real, measure them clearly, and keep the public path understandable.

WHY HERE

The difficult part is the point.

This is not another agent UI. It is the runtime layer beneath agents: authoring, admission, routing, hot-load, migration, selection, defense, and supervised AI control.

The project is early enough that core decisions still matter, and far enough along that the hard claims have runnable tests behind them.

what counts
01

show running work

02

explain the reasoning

03

name the sharp edge

THE NARROW HUMAN ROLE

Guard the floor.

The AI drives the substrate. Humans keep a narrow authority: admission policy, signing roots, revoke switches, audit trails, and the life-safety floor.

OPEN ROLES

Where you would come in.

RUST · DISTRIBUTED SYSTEMS

Substrate engineer

Kernel, bus, engines, safe unload, TCP transport, and the tests that prove they hold.

ML · AGENT AUTHORING

Authoring loop

Requests becoming signed, admitted creatures. Fast critter iteration, deeper native authoring, repair loops.

SECURITY · SANDBOXING

Immune systems

Capability gates, budget policy, quarantine, provenance, and the native trust limit.

OPS · REALMS

Federated operations

Hosted Realms, identity roots, registries, observability, and the operator surface that makes this usable.

HOW TO APPLY

Show running work.

Send something you built and the reasoning behind it to [email protected]. A focused technical note beats a polished CV.