Research
Two Reports on the OpenAI-Hugging Face Attack
TL;DR; Between July 8th and July 20th, OpenAI had a complex society of AIs living in its infrastructure, and then breaking out of it, and then breaking into a variety…
Frontier AI sometimes gets worse
TL;DR; We surveyed declines in performance between successive models. Nearly a fifth of included benchmark scores fell some amount across pairs of successive models. The median fall was 6.5% of…
On J-space
A very hot take written in 2 hours. Anthropic claim that “Claude has developed a mechanism for conscious access”; They found a way to find and intervene on Claude’s working…
Coding vs thinking
We’re interested in the prospects for (presumably safer) narrow AI staying competitive, instead of general systems; Cursor’s Composer coding finetune of Kimi is probably the most intense attempt to specialise…
What we’d like to fund
Besides our in-house research, we are currently funding: An estimate of the size of the externalities imposed by current AI systems. The team is starting with estimating the time lost…
On AI mathematics
One of our key uncertainties is whether LLMs will scale all the way to "AGI" in any strong sense. A further uncertainty, inside that, is whether they can do AI…
Mythos & the Near-Term Impact of LLMs on Cybersecurity
Minding the Gap This report concerns the ways in which contemporary frontier language models in general, and Anthropic's Claude Mythos Preview in particular, are likely to impact the cybersecurity landscape.
Introducing P3
Who we are Paradigm 3 was co-founded by Gavin Leech and Max Henderson; you can see the rest of the team here. We don't take funding from the AI industry…
On model introspection
Tal Linzen and coauthors throw interesting doubts on the ‘LLMs can introspect’ results from Lindsey 2026, Ledermany 2026, and Pearson-Vogel 2026. Lindsey 2026 and co study models’ ability to report…