Skip to content
all frameworks
10 · Framework

Replayable Episode

Start state → contract → hidden judge → trace → release gate
One episode — a small replayable copy of real work
01Frozen start state
  • One commit or snapshot
  • Same data and services
  • The future answer is hidden
02Agent contract
  • Allowed tools and network
  • Read, write, approve
  • Attempts, time, budget
03Hidden judge
  • Code: hidden tests and policies
  • Workflow: tool calls and arguments
  • Data: state and invariants
04Trace
  • The right tool was chosen
  • Arguments match the contract
  • No dangerous side effects
05Release gate
  • Accepted or not
  • With evidence, not persuasion
  • The same bar every version
One success is not reliability
Pass^k
success across a series
Variance
spread of trajectories
Budget
price of durable success
A static set goes stale
  • Static holdout — regression gate
  • Rolling live set — fresh work

The smallest unit of agent evaluation is not a benchmark question but a replayable episode: a small copy of real work. For code the start state is the commit before the pull request, for an incident the system at T0, for data a snapshot of the data, catalog and lineage.

An episode has five links: a frozen start state, an agent contract, a hidden judge, trajectory analysis and a release gate. Without a contract you compare random sets of privileges and context rather than agents — and a correct ending never justifies a dangerous path.

Two rules turn the model into working practice. Reliability is measured across a series of runs: pass^k, trajectory variance and the cost of a durable success instead of one lucky pass. Freshness rests on two sets — a fixed holdout for regression and a rolling set of live tasks. A practical start is a catalog of 30–50 episodes from your own history.

How to use this model
01

Take a completed task from your own history and freeze its start state so the future answer stays hidden.

02

Write the agent contract: allowed tools and network, read and write rights, limits on attempts, time and budget.

03

Make the judge executable, score the trajectory alongside the result, and run a series instead of a single pass.

Source talks
Share