Inside Coding Agents
v0.2 preview
EXPERIMENT · CONTROLLED · PASSED

Reference Harness replay and branch probe

Can append-only events deterministically rebuild session state and create a branch from an explicit sequence boundary?

reference-session-replay-v12 repetitions5 metrics
Evidence boundary

This is a controlled experiment. It supports only the listed subjects, fixture, configuration, and outputs; it does not generalize to native product behavior or cross-agent ranking.

STATUSpassed
SUCCESSFUL2/2
FIXTUREd369431d683d
PROMPTbafdff395bb0
SUBJECTS + CONTROLS

Which variables are fixed?

Both repetitions will replay five events, fork at sequence two with three inherited events, and share one normalized Trace fingerprint.

SDK

Reference Harness

reference-harness-2026-08-10-controlled

Model
scripted-model-v1
Provider
local-fixture
Comparability
not-comparable
Inspect snapshot →
CONTROLS

Run manifest

network
false
temperature
0
writes
false
sandbox
lesson-owned-temporary-fixture
RUNS + METRICS

Are repeated runs consistent?

python3 labs/runner.py reference-session-replay-v1 --check
RUN 01passed

reference-session-replay-v1-run-001

  • event-order-validpass
  • observations-matchpass
  • final-answer-matchpass
  • fixture-unchangedpass
  • deterministic-replaypass
acc0c48a153bfe73fd3048e034fb7c72cf8ac09eda32c61a71ec8d5f3eff0106Replay 7 events →
RUN 02passed

reference-session-replay-v1-run-002

  • event-order-validpass
  • observations-matchpass
  • final-answer-matchpass
  • fixture-unchangedpass
  • deterministic-replaypass
acc0c48a153bfe73fd3048e034fb7c72cf8ac09eda32c61a71ec8d5f3eff0106Replay 7 events →
SUCCESS CRITERIA
  1. Both repetitions replay five events and record the declared fork and inherited-event counts.
  2. Both repetitions return the expected answer and share one normalized Trace fingerprint.
LIMITATIONS
  • The event log is in-memory teaching data rather than a crash-tested production store.
  • The probe does not cover schema migrations, concurrent writers, or corrupted logs.
  • The Reference Harness result is not Native vendor evidence.
GENERATED REPORT

Reference Harness replay and branch probe

  • Status: PASSED
  • Mode: controlled
  • Experiment: reference-session-replay-v1

Question

Can append-only events deterministically rebuild session state and create a branch from an explicit sequence boundary?

Reproduction

python3 labs/runner.py reference-session-replay-v1 --check

The command rebuilds the expected artifacts in memory and exits non-zero if a committed trace, result, fixture digest, prompt digest, or report differs.

Result

2 of 2 repetitions passed all declared metrics.

  • reference-session-replay-v1-run-001: passed, 7 events, fingerprint acc0c48a153bfe73fd3048e034fb7c72cf8ac09eda32c61a71ec8d5f3eff0106
  • reference-session-replay-v1-run-002: passed, 7 events, fingerprint acc0c48a153bfe73fd3048e034fb7c72cf8ac09eda32c61a71ec8d5f3eff0106

Metrics

  • event-order-valid: pass
  • observations-match: pass
  • final-answer-match: pass
  • fixture-unchanged: pass
  • deterministic-replay: pass

Input digests

  • Fixture: d369431d683da5c2cad36bb6dc6f76d429c2efbdc29171d5c8d9d4d7c3c575aa
  • Prompt: bafdff395bb00b3027f78b9eec80b803a5e79513c618cfeae2a8373c32848b8e

Evidence boundary

  • The event log is in-memory teaching data rather than a crash-tested production store.
  • The probe does not cover schema migrations, concurrent writers, or corrupted logs.
  • The Reference Harness result is not Native vendor evidence.
Back to Lab →