(1/10) Today, we’re sharing new research from @OpenAI: How does CoT monitorability (no-CoT capability and CoT controllability) evolve over the model training lifecycle?
We emerged with unsurprising but sobering findings: as we scale toward more intelligent systems, models are becoming harder to monitor through their CoT. 🧵
My first 🤯 moment was with Astra. It described what it was doing so beautifully that I thought, “Wow, it really understood!”
Then I saw the actual result… and it didn't match reasoning at all.
that scared me a little.
Nov 11, 2026 · 7:03 PM UTC
1
193