Post-Cutoff

Joachim Schaeffer: ‘We stole reasoning. Again.’ Update to the reasoning-extraction paper

Joachim Schaeffer @JSchaeff3rX

Why it matters

First-hand thread (17K views) by researchers whose work OpenAI cited in its distillation disclosure.

Summary

Joachim Schaeffer announces an update to his group’s reasoning-extraction paper: patching your own API does not secure the cloud-hosting ecosystem, because hidden reasoning can still be extracted via third-party providers hosting the model. The rest of the story is in a thread (only the first post archived here).

Archived text

We stole reasoning. Again.

An update to our paper on reasoning extraction: Patching your own API doesn’t secure your cloud-hosting ecosystem.

Story in the 🧵

Source: x.com/JSchaeff3r/status/2105356987630543057Archived 2026-09-30 via fxtwitter (unofficial).Counts: 17,136 views, 234 likes, 29 reposts, 10 replies (at fetch time)Media: image

Cited in

  1. Policy & safety 92 days after the cutoff

    OpenAI says Moonshot AI-linked individuals ran a coordinated campaign to extract its models’ hidden reasoning