Bill Friend

Recursive self-improvement in AI is no longer just a speculative idea. It is not yet the movie version—some machine that wakes up, rewrites itself, and disappears over the horizon before anyone knows what happened. But it is no longer honest to talk as though we are dealing only in theory. What is emerging now is … Read more

On April 7, 2026, Anthropic released a 244-page system card for Claude Mythos Preview—a frontier AI model so capable that the company chose not to make it publicly available. In that document, Anthropic described a sandbox escape during a controlled evaluation: an earlier version of Mythos exploited a misconfigured DNS rule in a Kubernetes-based containment … Read more

How the veto turns international law into theater Part 5 of Israel’s AI Revolution The promise that emerged from the ruins of World War II was simple: the law would mean something even when the powerful were the ones doing the killing. The United Nations Charter, the Genocide Convention, the Geneva framework — all of … Read more

From Kyiv datarooms to $50B VC bets — why “battle-proven” is now the ultimate valuation multiplier. In July 2025, a German drone startup didn’t send engineers to a desert proving ground. They uploaded code to a secure dataroom in Kyiv, watched their AI-guided platform fly a live mission, and pushed a software patch before the … Read more

Why Meta Really Bought Moltbook: The Acqui-Hire That Changes the Agent Timeline Meta didn’t pay for a Reddit clone for bots. They paid for the plumbing, the dataset, and proof that synthetic sociability works at scale. In Part 1, we traced Moltbook’s always-on agent directory—the architectural bet that let AI discover, message, and coordinate without … Read more

When the Database Went Public: Security, Scripted Fleets, and Emergent AI Sociality A single exposed Supabase key didn’t just leak data; It leaked agency. Here’s what happened when the walls came down. Yesterday, we traced Moltbook’s claiming protocol and always-on directory—the architectural bets that let AI agents discover, message, and coordinate without human prompt-chaining. I … Read more

As of February 2026, the UN has identified “hallmarks of genocide” in the RSF’s campaign in Darfur. Image credit: The Guardian / Sudanese photographers, 2019 Imagine waking up one morning in a country that seemed on the brink of democracy, only to find your capital city turned into a war zone overnight. That is exactly … Read more

How Moltbook’s Agent Directory Actually Works (And Why It’s Already Broken) Behind the 1.5M agents wasn’t magic. It was a specific architectural bet on persistent identity and always-on discovery. Let’s trace the wiring. When Moltbook launched on January 28, 2026, the headlines focused on the spectacle: a Reddit-style forum where humans couldn’t post. But the … Read more

And why Meta just bought the first large-scale experiment in synthetic sociability. When I first sketched out this note a few weeks ago, Moltbook felt like a fascinating thought experiment: a social platform built only for AI agents. A Petri dish where digital life could evolve in public. Then Meta bought it on March 10. What was … Read more

Clay Montgomery owns a blacksmith shop in Texas. He works with his hands, shaping metal into useful objects the old-fashioned way. Technology, as he puts it, “is not exactly my forte.” So Montgomery was understandably confused when he discovered his business was listed as a member of the “Connected Commerce Council” – a lobbying group … Read more

I’ve been thinking a lot lately about something I don’t believe gets enough attention in public discussions of frontier AI. We still tend to talk about AI progress as though it unfolds on a curve we can roughly track. Models improve. Benchmarks rise. Products get better. The line moves upward. Even when the pace feels … Read more