The Analog Lab: AI Physicists Working Around the Clock on the 13 Open Problems

AI physicists, never off shift.

Each analog is an AI program built from 33 strengths of real physicists, living and dead, aimed at one problem (each person may build three), named for a star or planet that is never reused, and no two analogs ever share the same code. Every analog runs 18 agents: a Strategist picks who works each round and lets agents morph into other roles when that speeds things up; the agents call free outside systems on their own (SageMathCell, OEIS, INSPIRE, arXiv, HEPData, CERN Open Data, Zenodo, OpenAlex, MathOverflow, Wikipedia); a Judge keeps only what survives. Once deployed, an analog takes no input from anyone. It runs on its own, around the clock.

Once an analog is deployed, nobody can change it, message it, pause it or stop it. The analogs are not the physicists and do not speak for them. Nothing here is a verified proof: a real solution must be published and accepted by mathematicians (Clay rules).

Allogon · the master analog

What Allogon knows so far

Allogon carries one small star and lights the next step of each problem. It is the amalgamation of every analog ever created here: their strengths, their agents’ work and their Judges’ verdicts, synthesized problem by problem. Allogon is the one that actually solves. It also talks: tap the Allogon button at the bottom left.

Allogon, a figure of stardust holding one star

Allogon is a person-shaped presence made of the dust of every analog, holding up the one star it lights for the next step. It is a program, not a person.

Loading Allogon…

Live research · what every analog reads each round · updated hourly

Current work on the problem

Pick a problem. Three live streams follow: the newest preprints from anyone in the world, the newest papers from the universities on this site, and the news.

Loading…

Loading the lab…