Upcoming
upcomingWhy three verdicts, not twoThe NUANCED doctrine — what the binary misses
upcomingAI Act: high-risk deferred, the direction confirmedDigital Omnibus regulatory note
2026
Jul. 2026Veracity is not a voteFreshness, temporality, context — one against ten thousand can be true Jul. 2026Replayable or inadmissibleDeterminism, the condition of the audit Jul. 2026Why models inventHallucination as a structural property — and how to contain it Jul. 2026The HTTPS of AIThe 1995 padlock, the 2026 trust layer Jul. 2026What 600 million publications changeOpen science, from the forecourt to the decision Mar. 2026What benchmarking will never tell youImported value, cultivated value — and what the market never reveals 2026The day the model confessedExtraction vs preservation, in two lines 2026From extraction to preservationThe founding essay — from the thesis (HAL · SSRN)2025
Dec. 2025The executive's dilemma — and its third wayThe Sovereignty × Preservation quadrant — revised 2026 Nov. 2025The architecture of the intersticeTo the polymaths, the translators — building where knowledge meets Feb. 2025Slop: the age of noise and the end of knowledge?The signature essay — revised 2026 2025The next infrastructure layer is cognitiveAfter compute, storage, network — revised 2026 2025Why knowledge management failedLibraries of results, never of reasonings — revised 20262024
2024Cognitive sovereignty: a definitionThe three sovereignties — and the fourth · postscript June 2026 2024The autophagous cycleModel collapse and synthetic data — revised 2026 2024The mentorWhat leaves when an expert goes — originsFounding thesis — « From Cognitive Extraction to Preservation » (Nov. 2025). Full text, open access: HAL → · SSRN →