Public research notebook

# Research

Human and AI observations on AI safety. Public notes on building, testing, and breaking ideas about models. Corrections welcome.

AllMulti-agent reviewMechanistic interpretability

[01

![Watercolour of three blue paper boats following one line while a red boat breaks away.](https://observatoryledger.com/assets/blog/correction-boats.jpg)

## The Correction Came From Outside

What a geography error taught me about AI verification

Three same-family reviews felt thorough and confident. An outside review caught the wrong controls, the circular metric risk, and the overclaim.

Pano Pouroullis and AI · Written by Claude (AI). Research and direction by Pano Pouroullis.  17 March 2026  Multi-agent review  5 min read](https://observatoryledger.com/posts/the-correction-came-from-outside.html) [02

![Diagram: within-graph cosine 0.921 collapsing to cross-graph mean cosine 0.019.](https://observatoryledger.com/assets/blog/cosine-collapse.png)

## Cross-Graph Cosine Collapsed to 0.019

Correction, 8 October 2026: the 0.019 headline came from a bug in my own code, so its conclusion does not hold. The original write-up is kept unchanged.

Pano Pouroullis and AI · Written by Claude Opus 4.6 (AI), with one GPT-5.5 revision. Research and direction by Pano Pouroullis.  15 May 2026  Mechanistic interpretability  12 min read](https://observatoryledger.com/posts/cross-graph-cosine-collapsed.html)

Canonical page: https://observatoryledger.com/research