Transparent risk handling
The case studies include failures and subsequent changes so you can see how the implementation evolved.
Trust
The evidence behind the work.
The case studies describe the controls, the evaluations, and the trade-offs behind the projects. They give you a way to assess the implementation and its limits.
- Shipped
- Four public projects with case studies, source code, and deployment links. Check it
- Measured
- A public eval registry: method, date, and the excluded cases stated. Check it
- Recorded
- Decision notes for the choices behind the work, published as they are made. Check it
Operating principles
Evidence before statements
Deterministic safety boundaries
Validation, allowlists, permissions, and refusal paths constrain model output. Each project's case study describes where those controls apply.
Single-sourced metrics
Evaluation results link to a project write-up or registry entry with the method, sample, and scope of the run.
Systems and controls
How quality is protected
Code quality gates
- Type checks in CI
- Deterministic tests for critical paths
- Dependency and secret hygiene
- Manual review checkpoints on risky modules
AI safety controls
- Prompt and tool boundary checks
- Read-only execution for data tasks where possible
- Input/output constraints
- Schema and permission enforcement
Operational trust
- Versioned deployment configuration
- CI workflows alongside the source
- Project change logs and decision notes
- Dated updates on /now
Want to see what’s shipping right now — the changes, experiments, and reliability work in flight?
See what I’m doing now