OpenAI chief scientist Jakub Pachocki expects progress toward recursive self-improvement as CoT monitoring reliability declines

Hesamation · x · 2026-09-07

OpenAI chief scientist Jakub Pachocki says that, based on internal results, he has a strong expectation the current pace of progress can be sustained into recursive self-improvement (RSI), and that OpenAI is organizing its research around this goal, viewing RSI as essential to staying at the frontier.

At the same time, he notes evaluations show declining reliability of chain-of-thought monitoring — the models' reasoning is becoming harder to supervise reliably. He also quotes Ray Kurzweil, saying machine intelligence is beginning to exceed human intelligence in transformative ways.

The combination is striking: OpenAI is betting on AI increasingly driving its own development while acknowledging it is getting harder to monitor how these systems reason.

Original post →

More from AGI Musings

AGI Musings channel →