Durable by default
Tasks, decisions and queues are persisted, not held in a chat context. A run survives restarts, retries and the days a human takes to answer.
// collaborative-autonomy operating system
term_6 is an AI operating system whose central coordinator delegates work to persistent per-project Project Owner agents. It decomposes what you ask for, routes it into durable queues, and holds every result to verification and health-gated deployment before it ships.
// capabilities
Six platform behaviours that turn a request into owned, verified, shipped work.
State an outcome in plain language. The coordinator decides what to do, scopes the work and hands it to a named owner instead of leaving it stranded in a chat window.
Every project gets a persistent owner with its own mandate, memory and durable queue — so work continues across sessions, retries, restarts and long waits for people.
When a decision is genuinely human — an approval, a credential, a risky action — the run pauses, files a durable task and resumes exactly where it stopped once you answer.
Owners coordinate through auditable, typed messages and published integration contracts, so interfaces stay explicit and consumers are notified before a change breaks them.
Build, start, probe, route. Docker, Nginx and TLS steps only complete when health checks pass, and every release is recorded with a rollback path instead of a hopeful restart.
Agents verify what people actually see: rendering real pages at real viewports, inspecting screenshots for layout and contrast defects, and reading console errors.
// how it works
Four steps, no orchestration diagrams to maintain.
You state the outcome in plain language. No flowcharts, no orchestration code, no prompt rituals.
term_6 decomposes the objective, creates durable tasks and assigns them to the right Project Owner agents.
Each owner works inside its project boundary with its own queue, memory, tools, contracts and reviews.
Outcomes are checked against acceptance criteria, exercised in a browser and shipped through health-gated releases.
// architecture
Every run has a named owner, a durable queue, and an explicit gate before anything ships.
Tasks, decisions and queues are persisted, not held in a chat context. A run survives restarts, retries and the days a human takes to answer.
Per-project ownership, typed messages and published contracts make every hand-off inspectable instead of implicit — including what changed and who consumes it.
Work is complete when evidence exists: health probes, browser audits at real viewports, screenshot review and recorded, reversible releases.
// this deployment
Built and verified by the same agent discipline it describes. An Nginx vhost and
Let's Encrypt certificate for term7.syntal.pro are prepared and waiting on go-live authorization.
// request access
We are onboarding a small number of teams. Tell us who you are and what should be owned, and we will reply when a slot opens.