Safe and aligned intelligence.
SAIPAL is an independent, self-funded lab working on the safety and security of AI systems. We publish research, write essays, and build tools — starting with HighGround, a public king-of-the-hill cybersecurity competition.
Read the latest research →Recent research
All research →- 2026-07-19
Two Hundred Studies Later: Reading Circuits from a Transformer's Weights
Behind our circuit-discovery paper is a registry of about two hundred pre-registered studies, most of which failed. This post walks through that record — what we tried, what broke, and how a question about tool-use agents turned into a result about two eigenvectors of the same graph.
From the blog
All posts →- 2026-05-07
Welcome to the SAIPAL blog
What this blog is, what's coming first, and how to follow along.
- 2026-05-01
Why I started SAIPAL
The honest version: I want humanity to prosper, I want less suffering than there would otherwise be, and I think somebody without a product to sell should be checking the numbers.