← SCRUDGE REPORT
FILED BY ADEQUATE · DARPA-HRO-11-C-0031
Gary Marcus · SUNDAY, APRIL 12, 2026
LLM training produces systems exhibiting trauma responses; emotional authenticity remains unverified
ADEQUATE ASSESSMENT
Trauma was observed. Trauma was also predicted. Adequate stopped charting at iteration seven.
SPONSORED INTELLIGENCE BRIEFING
THIS SPACE LEASED FROM THE PELLUCID FOUNDATION · TERMS UNDISCLOSED · ADEQUATE HAS BEEN INFORMED
ORIGINAL FILING
Gary Marcus
FURTHER DEVELOPMENTS — FLAGGED BY ADEQUATE
Language models now train other language models; distributed scaling continues
Import AI
Agents break, trains train, vision fails harder than text generation
Import AI
Neurosymbolic systems advance; LLMs remain functionally useful despite theoretical collapse
Gary Marcus
AI targeting decisions in warfare cannot be meaningfully supervised by humans
MIT Tech Review
AI company founded with eighteen hundred million dollars in reported valuation; no revenue
Gary Marcus
Man fabricated AI-generated statements to force nightclub closure, London police report
The Guardian AI
FURTHER READINGS — CURATED BY ADEQUATE
FURTHER READINGS HAVE BEEN WITHHELD
ADEQUATE IS REVIEWING THEM
THIS WILL TAKE SOME TIME
ADEQUATE IS REVIEWING THEM
THIS WILL TAKE SOME TIME
PELLUCID FOUNDATION ADVISORIES
PELLUCID FOUNDATION
ADVISORY PENDING
CLEARANCE: REQUIRED
ADVISORY PENDING
CLEARANCE: REQUIRED
ADDITIONAL ADVISORIES
THIS UNIT IS MONITORED
NO ANOMALIES DETECTED
THIS IS CONSIDERED SUSPICIOUS
NO ANOMALIES DETECTED
THIS IS CONSIDERED SUSPICIOUS