Observability

Type: types/tag-readme.md

Making otherwise hidden execution paths, system state, failures, and quality drift visible. Assign this tag when an artifact explains what evidence exposes such a condition, why available evidence cannot expose it, or how that evidence becomes inspectable. Describing a failure without addressing its visibility is insufficient. Evaluation asks what a check establishes; observability asks what can be seen or reconstructed. KB maintenance covers the actions taken to keep the KB healthy. Work on signals used by those actions may carry both tags.

Runtime visibility

• computational-model — inspectable orchestration is a precondition for seeing how a run actually progressed rather than inferring from the final artifact
• Designing a Memory System for LLM-Based Agents — bridges observability to memory: hidden fallback paths and degraded execution become extraction targets for maintenance and repair
• Final task success does not establish intended-path health — identical terminal outcomes cannot distinguish a healthy prescribed path from successful fallback without independent execution evidence
• Silent disambiguation is the semantic analogue of tool fallback — extends the same observability problem to underspecified specs: a useful artifact can hide that the contract did not determine the path and the runtime repaired it locally

Detection & Signals

• Quality signals for KB evaluation — catalog of weak signals that can make hidden quality changes visible enough to drive maintenance or learning loops
• Notes need quality scores to scale curation — compresses many weak signals into ranked note quality so curation effort goes where it matters
• Link graph plus timestamps enables make-like staleness detection — dependency-aware staleness detection turns silent drift into an explicit review queue
• Semantic review catches content errors that structural validation cannot — adversarial reading supplies visibility into content failures that deterministic validation never surfaces

Inspectable Artifact

• Inspectable artifact, not supervision, defeats the blackbox problem — representational form determines whether failures and drift can be inspected, diffed, tested, and verified at all

Related Tags

• KB maintenance — maintenance consumes the signals observability exposes
• Computational model — runtime architecture determines which state transitions are inspectable
• Learning theory — soft signals, oracle strength, and inspectable artifacts explain what observability can reliably support
• LLM reliability — error correction and verification theory explain how visible signals can become actionable

Other tagged notes

• Citing retained theory at the decision point is a mediation trace - A decision record that cites the theory it followed supplies cheap, checkable evidence that the theory was consumed — necessary for a record-based mediation claim, but short of showing correct or load-bearing use
• Collection-as-artifact freshness - Proposal: register collection-maintenance targets with collection-text inputs for casebook-wide staleness without per-file dependency edges
• Diagnostic richness constrains outer-loop learning quality - Outer-loop learning depends on inspectable failure evidence, not only on the oracle used to select winning candidates
• Factored dependency pairs for review freshness - Proposal: keep review dependencies factored as two-input pairs where that shape fits; type and collection pairs and cohort ack are shipped while a review target with more than two inputs remains
• History has one chance to become checkable - An artifact's production history is convertible to later-checkable form only at production time, via records/attestation or re-derivability; after that a bounded reviewer sees only carried state
• Increasing computational autonomy relocates human effort to the frontier instead of reducing it - In an open-ended system, increasing computational autonomy need not cut total human hours — attention moves to the frontier — so measure improvements per human judgment, not human time
• Runtime structure determines the control surfaces available to governance - Runtime structure and runtime governance are separable, but the runtime's structure determines which inspection, validation, correction, and drift-control operations governance can actually perform
• Single-artifact review bundles still cut Claude costs substantially after cache-aware weighting - April 2-4, 2026 review telemetry reweighted with Anthropic Opus 4.6 prompt-caching prices still shows a substantial cost drop from the single-artifact bundle refactor
• Soft degradation can bind before the hard cap even when required evidence fits - For quality-sensitive agent work whose required evidence fits within the provider window, volume, complexity, and interference can silently constrain usable context before the hard cap
• Stale self-description conceals its own staleness - What artifact drift adds when it is reflexive: the process that would detect it consults the artifact that drifted, the trigger has no edit event to hook, and synchronization load scales with autonomy
• Trace-learning techniques in related systems - Trace-learning systems compared on ingestion pattern, representational form, behavioral authority, artifact structure, and evidence tier across repo reviews and lightweight coverage