Every month, I go down a few too many rabbit holes, read a bunch of interesting (and sometimes ridiculous) things, and come out the other side with a head full of ideas. Instead of overthinking them, here’s a quick brain dump of what caught my attention this month. No deep dives, just a list of things I found worth noting.

  • The Donroe Doctrine: Trump’s spin on the Monroe Doctrine, reportedly using it to justify territorial claims on Greenland and the Panama Canal.
  • Okstupid.lol: Martha Root (CCC) infiltrated the white supremacist dating site WhiteDate and exfiltrated over 8,000 profiles and 100GB of data. A wild piece of hacktivism.
  • London–Calcutta bus service: Back in the 60s and 70s, you could take a 50-day, 16,000 km bus trip from London to Calcutta. The hippie trail was real.
  • Stateful Agents: Letta is building infrastructure for agents with persistent memory, so they can remember context across sessions instead of starting fresh every time.
  • Open Responses: An open-source spec for building multi-provider LLM interfaces based on OpenAI’s Responses API. Trying to make model switching less painful.
  • Ralph for Claude Code: Autonomous AI development loop for Claude Code with intelligent exit detection.
  • VSM: The Viable System Model is a framework for thinking about how organizations (or systems) can be structured to be self-regulating and adaptive.
  • OpenAI to add Ads to ChatGPT: Yep, it’s happening. OpenAI confirmed they’re exploring ads in the free tier. The enshittification continues.
  • ChatGPT Apps: OpenAI launched a way to build and share lightweight apps inside ChatGPT, kind of like GPTs but more interactive.
  • Kimi K2.5: Visual Agentic Intelligence: Moonshot’s latest model with strong visual understanding and agentic capabilities.
  • ClawdBot: A new target for infostealers, the desktop Claude agent app became popular enough that attackers are specifically trying to steal credentials from it.
  • ClawdBot rebranding saga: First renamed to Moltbot, then rebranded again to OpenClaw.
  • Moltbook: A Reddit-style social network where AI agents talk to each other. Equal parts fascinating, weird, and concerning. I wrote more about it.