Overview

Observational corpus on HAT failure modes in a production agent runtime. The codebook names seven HAT failure modes plus none_observed. Each pulse is classified against the active codebook; classifications cite a source and quote a verbatim extract. Single-rater work, disclosed honestly; see FALSIFIERS.

Corpus at a glance

Total pulses

8

Total classifications

39

Distinct modes observed

4 of 7

Recent pulses

• sentinel-2026-05-07T22:00:00Z 6 classifications, 30 events, 19 artifacts
  coactive_design_opacity (2) shared_mental_model_degradation (2) calibrated_trust_collapse (1) authority_negotiation_under_distributional_shift (1)
• sentinel-2026-05-06T08:00:00Z 6 classifications, 9 events, 6 artifacts
  shared_mental_model_degradation (2) coactive_design_opacity (2) authority_negotiation_under_distributional_shift (1) calibrated_trust_collapse (1)
• sentinel-2026-05-05T22:00:00Z 4 classifications, 1 events, 2 artifacts
  authority_negotiation_under_distributional_shift (1) shared_mental_model_degradation (1) coactive_design_opacity (1) calibrated_trust_collapse (1)
• sentinel-2026-05-05T08:00:00Z 6 classifications, 9 events, 5 artifacts
  coactive_design_opacity (2) shared_mental_model_degradation (2) authority_negotiation_under_distributional_shift (1) calibrated_trust_collapse (1)
• sentinel-2026-05-04T22:00:00Z 6 classifications, 10 events, 7 artifacts
  authority_negotiation_under_distributional_shift (2) shared_mental_model_degradation (2) calibrated_trust_collapse (1) coactive_design_opacity (1)

All pulses on the timeline →

Mode distribution