Test the charter before you promote it.

genesis-charter-lab · portable workflow
1Freeze the baseline charterscoped
2Run paired and held-out evaluationsscoped
3Compare evidence against the baselinescoped
4Promote or roll back explicitlyscoped
Workflow overviewRead · configure · verify

How it works

1
Freeze the baseline charter
2
Run paired and held-out evaluations
3
Compare evidence against the baseline
4
Promote or roll back explicitly

Why it beats the alternative

This product
Promoting a prompt change on one demo
Baseline
A known charter is frozen first
The comparison moves during testing
Evaluation
Paired and held-out checks
One hand-picked example
Rollback
A regression has an explicit path back
The old instructions are gone
Limit
It changes public instructions, not model weights
Model training is implied

How it actually works

Exactly what it does — and when.

1Compare candidate and baseline on paired tasks
2Use a separate held-out task before promotion
3Rollback a public charter after measured regression
In practice · A charter earns promotion

A planner owner proposes a shorter instruction set. The lab runs it beside the baseline, checks a held-out task, and promotes it only when the measured behavior holds; a regression returns to the prior version.

Use it to…

Run a paired evaluation

Measure the candidate and baseline on the same task set.

Hold out a check

Test a separate task so the candidate cannot win by memorizing the training cases.

Roll back safely

Return to the previous public charter when the measured result regresses.

The shift it creates

Without it
  • Promoting instructions after one good demo
  • No held-out check for regression
  • Public behavior changing without rollback evidence
With Genesis Charter Lab
  • Compare a proposed charter against a known baseline
  • Use a held-out check before claiming improvement
  • Roll back a public instruction when the evidence regresses

What's inside

Paired evaluations
Held-out checks
Promotion and rollback
Node.js 20+
MIT-licensed source
It changes public instructions, not model weights

Who it's for

Planner maintainers

need changes measured before promotion

Worker owners

need role instructions tested against real tasks

Teams evolving agents

need rollback when a candidate underperforms

Questions, answered

✓ Instant delivery✓ Yours to own✓ Use on every project✓ Built to ship results

Ready to put Genesis Charter Lab to work?

Instant download · one-time price · own forever · all sales final

Genesis Charter Lab