Original studies from the archive, read online as summaries — clear findings and an evidence grade, not downloadable reports. Every number is checked against its source before it ships.
A synthesis of 259 shipping system prompts (Claude, GPT-5.x, Gemini, Cursor, Grok and more), anchored to a blind A/B/C/D test and distilled into seve…
An anonymized case study of a frontier model commissioned to audit and rebuild a single-developer system, then run past two independent adversarial e…
Semantic + faceted search over ~256 spacetags (a controlled capability/action vocabulary plus NL summary, verdict, and numeric scores). Eight Phase-0…
There is no queryable tweet corpus for Fable in this archive — by design. What exists is a hand-curated social-research policy: X's API is paywalled…