Anthropic's "Too Dangerous" Model Is Coming for Everyone
Six weeks ago Anthropic said a model was too dangerous to ship. This week they shipped Opus 4.8, called it near-Mythos, and said Mythos-class models reach every customer in the coming weeks. So — was it ever that dangerous? That's the question Oscar and I opened with, and it set the tone for the whole episode.
Here's how I called it.
Signal or Noise
Claude Opus 4.8 — near-Mythos, 3x cheaper fast mode, Dynamic Workflows
The capability is nice; the 3x cheaper fast mode and fanning one model across a workflow is what actually changes how I build. Signal.
Anthropic passes OpenAI at a ~$1T valuation, with a projected first operating profit
The narrative flipped this week. A frontier lab projecting actual operating profit is a bigger deal than any benchmark. Signal.
Cognition raises $1B at $26B; Devin hits a ~$492M run rate
Autonomous coding agents are past the demo phase and into real revenue. Signal.
A poisoned Nx Console extension steals Claude Code configs and helps clone thousands of internal repos
This is the one that should scare you. A supply-chain attack through a VS Code extension, aimed straight at the AI dev tooling we all trust. Pin your extensions, scope your tokens, assume the tooling is a target. Signal, and the one I'd act on today.
Pope Leo XIV's first encyclical takes on AI
Signal for regulated buyers, noise for your standup.
Ship It or Skip It
Three ideas:
- The Company Brain — a living map of how a business actually operates, exposed as something every AI agent can call. This one I believe in enough to build.
- A voice AI agent for the trades — plumbers, electricians, HVAC. Boring, underserved, real money. Ship it.
- Stripe for AI agents — payments infrastructure for agents that need to transact. Early, but someone owns this rail. Ship it.
The theme of the week: "too dangerous" turned into "available in a few weeks," and the real risk moved to your supply chain. Watch or listen below.
