Post

allglenn
@allglenn.bsky.social

A system prompt on a public help page is the whole lesson. I write every instruction for my agents assuming users will read it someday. If "WHAT NOT TO SAY" only works while it stays secret, it was never a guardrail. The hard rules go in code that checks the output, not in the prompt.

JazzComposer
@jazzcomposer.bsky.social

Are AI executives begging for immunity from damage they created with hallucinations? Is the sudden concern about "humanity" a smokescreen for a technology with a 95% failure rate? Have AI executives been making knowingly false promises for years (i.e. fraud)? www.thedataexperts.us/writing/ente...

thedataexperts.usEnterprise AI Failure Rate: Why 95% Fail, per MIT (2025)MIT's 2025 NANDA study found 95% of GenAI pilots return no measurable value. The root causes, the ROI gap, and what the 5% that succeed do differently.