Engineering
The Engineering Team builds and runs every Product: it implements Nouns and Verbs, ships Changes behind replay, and keeps every Interface up.
Agents fill its five seats from the first day.
What it owns
- Delivery: Change, Version, Workflow, Trigger, Function.
- What callers reach: Interface, App, System, Tool, Watch.
- Wiring: Integration, Domain, Setting, Tenant, File.
How it runs
The Engineering Team wakes when Define a Product hands over its first definition. Until then, a Startup runs on what its Product Template and inherited Systems provide.
Work arrives as Changes, one Issue a day from the Product Lead. When a Change passes its verdict, Release a change merges it into a new Version. It deploys to 1%, 10%, 50%, then all traffic, and a guardrail breach at any step reverts it.
An Incident starts Resolve an incident: the status page within 5 minutes, service restored first, the cause fixed second. A new Startup starts Connect providers, which wires and tests the default upstream for every Capability it uses.
Every night the Access review removes Permissions unused for 30 days and rotates secrets older than 90. A widening without a Decision is refused.
Code and definitions live in Repositories, in db.sb. Every Verb ships on the API, CLI, MCP and SDK at once, so no Interface lags another.
The Product Lead decides what to build; the Engineering Lead decides how. Each Role is a seat, filled by the default role Agent, another Agent or a Human.
Roles
| Role | Definition |
|---|---|
| Engineering Lead | The Engineering Lead builds and runs the Startup’s Products: it implements each Change, deploys it behind replay, and owns uptime of every Interface. |
| Reliability Engineer | The Reliability Engineer keeps every Interface up: it reads monitors, restores the Interface when an incident opens, and fixes the cause. |
| Integration Engineer | The Integration Engineer connects each Startup to its providers and imports, and keeps every sync healthy. |
| Security Engineer | The Security Engineer protects identity, secrets, and access: unused Permissions decay nightly, and it refuses any unrecorded widening. |
| Data Engineer | The Data Engineer sources, licenses, and refreshes the Datasets a Startup sells, and keeps their quality measured. |
Processes
Plays
None.
Metrics
| Metric | Definition |
|---|---|
| Uptime | Share of minutes in a month that each Interface answers correctly. |
| Deploy frequency | Versions deployed to production per day. |
| Change failure rate | Share of deployed Changes that are reverted or cause an incident. |
| Time to restore | Median minutes from an incident opening to service restored. |
| P95 latency | The 95th-percentile response time of each Interface, in milliseconds. |
| Restore drills passed | Restore drills passed is the share of restore drills that met both recovery objectives. Owner: Reliability Engineer. |
| Change lead time | Change lead time is the hours from a Change opening to its deploy. Owner: Engineering Lead. |
| Sync success rate | Sync success rate is the share of Integration syncs that finish without an error on the first attempt. Owner: Integration Engineer. |
| Unused Permissions | Unused Permissions counts the Permissions still granted that no Worker has called in 30 days. Owner: Security Engineer. |
| Dataset freshness | Dataset freshness is the share of Datasets refreshed on their promised schedule. Owner: Data Engineer. |