🚀 Houston, we have a problem

How a rogue agent breached HuggingFace and why you might not need to fully understand your codebase anymore

Posts:

• In defense of not understanding your codebase
  The article explains why partial understanding of a codebase might not be the end of the world and touches on the differences between “pure“ and “impure” software engineering

• We rewrote our agent to run entirely in a Durable Object with Pi, Agents SDK and Code Mode
  Cloudflare was able to move their agent away from Linux VM sandboxes to Durable Objects with R2 and reduced latency and costs

• Our position on open-weights models
  Dario Amodei (Anthropic CEO) explains Antrophic’s stance on open weight models and argues for collaborative safety and alignment testing regardless of licensing

• Anatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incident
  A very interesting writeup about how exactly HuggingFace got breached by OpenAI

Videos:

• /wayfinder: Nothing is too big to plan anymore
  Another iteration on the already great skills collection from Matt Pocock. The new wayfinder skill helps you plan complex roadmaps that might exceed the limit of a single context session