Test changes beside the original, measure lift.
teleon.dev/docs/mutations emits sibling rows for each personality, keyword set or instruction pack while the baseline row stays untouched, then measures the lift each change brings.
The concise-style sibling wins by seven points with enough effective samples to trust it.
What it does
Baseline untouched
Every mutation is a sibling row; the original is the control.
Five personalities
Instruction styles and keyword sets run side by side.
Lift you can trust
Paired comparisons with shrinkage and effective sample size.
The contract
Siblings share a parent; lift is computed against the parent with paired data.
teleon_probabilistic_mutation_request/v1kish_neff_factor_lift
Endpoints
| Method | Path | Access | What it does | Status |
|---|---|---|---|---|
POST | /v1/mutations | RING-2 | Emit sibling rows for a task | Listed |
GET | /v1/mutations?parent={id} | RING-3 | Read siblings and their lift |
Listed paths appear in the service’s published files today. In the one-host plan every path here is served under api.teleon.dev/v1/mutations.
How it connects
For agents
Give your agent the read-only guide first. It reports back before any write.
Words on this page
- Holdout
- Test data kept apart from training and tuning, so scores on it show real performance.
- Idempotent
- Safe to repeat: running the same request twice leaves the same result as running it once.