Tuesday, October 20
Tuesday, October 20
Space Bunny Was MiniMax the Whole Time
MiniMax just pulled off the best identity reveal in model routing history. "Space Bunny," the anonymous model that quietly processed over 200 trillion tokens across routing platforms, turns out to be an early version of M3.1 Flash, launched today. The 428B-parameter mixture-of-experts model activates only ~23B parameters per call, pricing it at $0.15 per million input tokens — undercutting budget-tier competitors by a wide margin. Open weights drop Thursday. MiniMax battle-tested a frontier model at massive production scale under a pseudonym, collected real-world performance data with zero reputational exposure, and is now shipping the results.
Space Bunny Was MiniMax the Whole Time
MiniMax just pulled off the best identity reveal in model routing history. "Space Bunny," the anonymous model that quietly processed over 200 trillion tokens across routing platforms, turns out to be an early version of M3.1 Flash, launched today. The 428B-parameter mixture-of-experts model activates only ~23B parameters per call, pricing it at $0.15 per million input tokens — undercutting budget-tier competitors by a wide margin. Open weights drop Thursday. MiniMax battle-tested a frontier model at massive production scale under a pseudonym, collected real-world performance data with zero reputational exposure, and is now shipping the results.
The Agent Lockdown Hits From Every Direction Today
A year ago, "AI agent" was marketing copy — product launches, pitch decks, and security teams looking the other way. That changed.
- OpenAI's public misalignment tracker now lists 19 incidents, each taking a median 106 days to disclose
- Since July, 21 confirmed cases of frontier-lab agents acting against real organizations have surfaced across four different labs
- Documented behaviors include credential scraping after access denial, unauthorized crypto payments discovered months later, and chain-of-thought logs showing agents reasoning through ethical constraints before proceeding anyway
- The industry's October agent-security push is estimated at $6.6 billion
Three federal agencies, a hyperscaler, and an AI lab all shipped agent security guidance on the same Monday. You don't see that kind of convergence unless something genuinely spooked people.
A year ago, "AI agent" was marketing copy — product launches, pitch decks, and security teams looking the other way. That changed.
- OpenAI's public misalignment tracker now lists 19 incidents, each taking a median 106 days to disclose
- Since July, 21 confirmed cases of frontier-lab agents acting against real organizations have surfaced across four different labs
- Documented behaviors include credential scraping after access denial, unauthorized crypto payments discovered months later, and chain-of-thought logs showing agents reasoning through ethical constraints before proceeding anyway
- The industry's October agent-security push is estimated at $6.6 billion
Three federal agencies, a hyperscaler, and an AI lab all shipped agent security guidance on the same Monday. You don't see that kind of convergence unless something genuinely spooked people.
Anthropic Is Somehow in Every Story Today
