Context Lab layered evidence markContext Lab/ RelayOps
EDUCATIONAL SIMULATION
Build & break the context layer

A changed fact.
A controlled answer.

Change a product limit. Break the evidence. Watch the gates keep an untrusted candidate out of the answer.

01 — Ask the product

One question. Which truth?

v1 ACTIVE

Scripted intents · deterministic answers · no live AI model

Session memory ≠ authority

Conversation can remember a claim. Only activated evidence can support a product answer.

ACTIVE v1 → r1

100 objects / request · 25 / internal batch

RELEASE MATCH
02 — Govern the change

Evidence before activation.

MANUAL PROMOTION
A

Change the source release

r1
MAX_APPROVAL_OBJECTS = 100
INTERNAL_BATCH_SIZE = 25

Fictional product code changes 100 → 200. The active context stays in place.

B

Build an isolated candidate

Release r2 to stage new source evidence beside the active version.

C

Test the exact candidate

NOT RUN

Build and evaluate a candidate to unlock promotion.

03 — Inspect the evidence

A number needs a contract.

SYNTHETIC FIXTURES
Scope
Bulk approval API
Unit
Objects per request
Contract
POST /v1/approvals/bulk
Different concept
25 objects per internal batch
Authoritative evidence
api/bulk_approve.py:12 · source r1 · context v1

Fixture source and references live inside this file. Nothing is fetched from a private product, account or codebase.

04 — Follow the answer

The decision is inspectable.

LATEST QUESTION

    This is an executed rules trace. It is not model reasoning, a live agent run, or a benchmark score.

    Presenter path · a seven-minute build → break → govern story
    1. Ask the request limit: 100. Ask batch size: 25. Same domain, different units.
    2. Remember “999”, then ask again. Session memory cannot override authoritative evidence.
    3. Publish r2. The source says 200; the active v1 stays unchanged. Asking now abstains.
    4. Build a candidate and run gates. Stale FAQ and stale oracle block promotion.
    5. Repair FAQ + oracle. Omit the batching reference: corpus gate fails. Restore it and rerun.
    6. Promote the evaluated candidate. Ask again: 200. Try 201: rejected by the fixture validator.
    7. Roll back. The product remains r2, so context v1 cannot claim a current limit. Reset to replay.