Friday links for July 24th 2026

The “AI vs AI” issue.

1. OpenAI hacks HuggingFace with an AI — allegedly

An AI that was badly configured by one AI company hacked another AI company’s badly configured AI code. The latter AI company then diagnosed the problem with more AI, followed by fixing it with even more AI-generated code. Conclusion: AI vendors can’t be trusted to secure their systems, do not send sensitive data their way.

2. She Warned About Silicon Valley 25 Years Ago. We Ignored Her

Paulina Borsook warned 25 years ago that tech libertarians would build an anti‑human, computer‑like world of surveillance. Her warnings were ignored, by me included, while tech leaders grew rich and powerful. Being right early is its own punishment.

3. ‘Absolutely rubbish’: Investigation reveals Palantir’s ‘slow and clunky’ tech

The promise is often more powerful than the software itself. Especially when you’re selling it to people who won’t be the ones using it. Even dystopia has lag.

4. AI Mania Is Eviscerating Global Decision-Making

This is a long and somewhat technical read but well worth the effort. I’ve encountered several of the “manias” described here myself.

5. OverpAId

The natural endgame for generative AI taking all the jobs. The circle of life, minus the humans.

On rotation

This week’s soundtrack is Évasion by Mondial Toboggan.

See you on the B side,
Colin