Your Sovereign AI Coding Agent

Why This Matters: The Night They Flipped the Switch On June 12, 2026, a frontier AI model that millions of people were using vanished for the entire planet overnight — not because it broke, but because a government letter ordered it pulled over a bug the vendor itself called minor. One day after launch. Gone, for everyone. I watched it happen in real time, mid-project. The full story — and why it’s the 1990s Crypto Wars all over again — is in The Crypto Wars, Round Two. The takeaway that matters here is simpler: if your tools live on someone else’s servers, your access is a permission, not a possession — and permissions get revoked. ...

June 13, 2026 · 11 min · Cypherpunk School

Why Your Local AI Lies to You (and the Deterministic Gate That Catches It)

The Problem: Amplified Hallucination Guide #1 in this series showed you how to run a local coding agent on your own hardware. Revocation-proof, zero per-query cost, no export-control letter touches it. What it did not cover — because it revealed itself mid-session — is the follow-on problem: the model will lie to you about whether what it built works. Not the obvious kind of lie. Not “I am unable to complete this task.” The dangerous kind: a polished, correctly-shaped, plausibly-commented artifact that is quietly broken, delivered with a confident task-complete. ...

June 13, 2026 · 14 min · Cypherpunk School

The Crypto Wars, Round Two: They Export-Controlled an AI Overnight

I found out the way you find out about most things that matter: by accident, in the middle of doing something else. A while back I’d built a small AI voice red-teaming lab — a setup to stress-test how well AI voice assistants hold up against manipulation — and I’d gone back in to revisit the project. A few prompts deep, the model threw up a bright-yellow banner: it had flagged something in my session for safety and was bumping me down to an older model, mid-task. Mildly annoying, but fine — these things have tripwires, and I figured I’d brushed one (I was using words like “exploit” and “red-team” while documenting a defensive tool; classifiers get jumpy). ...

June 12, 2026 · 7 min · Cypherpunk School