About

Mission

To measure intelligent systems by observable behavior and reproducible evidence, not by marketing claims or model reputation, and to publish that evidence in a form others can check.

Scope

ALGEDI investigates questions that can be answered empirically: how models behave under realistic and adversarial conditions, how agents execute multi-step work against real systems, what sustained human-AI collaboration reveals, and how evaluations themselves should be built and validated.

What ALGEDI is not

Not a news outlet, a model leaderboard or a list of vendor announcements.
Not an agency offering development services. Applied engineering, audits and security engagements are outside its scope.
Not a venue for private conversations or proprietary code published without authorization.
Not a source of claims about AGI, consciousness or model internals that behavioral evidence cannot support.

Publication philosophy

A small body of work with rigorous evidence is worth more than a large catalog of unsupported claims. Every entry states how strong its evidence is, from an open research question to an independently replicated result. Work advances only when its evidence does, and corrections stay visible.

Research questions are published early, before results exist, so that the question and the standard of evidence are on record before any answer is.

Engaging with the lab

Independent researchers, engineers and research groups can propose questions, contribute to protocols, attempt replications, or report errors in published work. Replication attempts are welcome whatever their outcome. See collaboration.

The name

Algedi is the traditional name of Alpha Capricorni, a pair of stars that appear close together in the sky but lie at very different distances from Earth. Telling an apparent pair from a physical one takes measurement. The lab's mark draws on that idea.