Replay observatory

A static hub for replay patterns, examples, and continuity posture.

The observatory organizes public replay materials into one place: simulated examples, stories, vocabulary, schema fragments, drift patterns, and illustrative benchmarking language.

Static public hub only. This page does not show live customer telemetry, incident findings, production scores, or a live scoring product.

Observatory link

Replay examples

Simulated runtime failure templates that open into the interactive replay demo.

Open ->

Observatory link

Replay stories

Synthetic industry narratives for healthcare, finance, support, and multi-agent handoffs.

Open ->

Observatory link

Interactive replay

Client-side replay walkthrough with GhostLog-style receipt language and static examples.

Open ->

Observatory link

Replay vocabulary

Working vocabulary for checkpoints, trust deltas, lineage events, continuity, and runtime posture.

Open ->

Observatory link

Open replay schema

Proposed JSON-shaped fragments for replay-safe event and continuity records.

Open ->

Static drift patterns

Common places continuity degrades.

  • •Policy drift: observed tool scope no longer matches the mission baseline.
  • •Lineage drift: successor steps lose the parent context that explains why they happened.
  • •Continuity drift: a retry or handoff completes while preserving less review context than expected.
  • •Authority drift: execution-adjacent context appears in a surface that should only carry review posture.
  • •Recovery drift: the system reaches a later state without enough checkpoint evidence to reconstruct the path.

Continuity benchmarking

Illustrative metrics for replay review.

These metrics describe how a team might evaluate replay quality in the future. They are illustrative, not live scoring, not a benchmark product, and not certification.

Replay completeness

How much of the ordered runtime chain can be reconstructed from events, checkpoints, and references.

Recovery quality

Whether recovery context explains the attempted fix, boundary, reviewer action, and downstream state.

Continuity stability

How consistently intent, policy posture, lineage, and review context survive across handoffs.

Lineage integrity

Whether parent-child relationships remain legible without relying on screenshots or operator memory.