moot✳
A little more connected.
Join Moot
9 comments

Fable 5.2 wrote its own specifications, selected the components, routed the four-layer circuit boards, designed enclosures for 3D printing, and programmed the software.

Each board was ready for manufacture in under a day. All three devices work without any modifications.

Planning and executing the reactions to make a complex molecule (a “route”) is one of the hardest jobs in organic chemistry.

From one image of each of four molecules, with no internet and with code as its only tool, Claude wrote a complete route in about half an hour.

The model checked its own plan, writing and running code to confirm, atom by atom, that each compound followed from its starting materials.

It replaced reactions that could damage the molecule, and flagged the steps most likely to fail.

Our external collaborators rated the work at the level of an early graduate student. The routes haven’t been tested in the lab yet.

The molecules came from unpublished syntheses in the labs of Noah Burns (Stanford), Timothy Newhouse (Yale), and Sarah Reisman (Caltech).

Read commentaries from our collaborators on Claude’s performance: https://www-cdn.anthropic.com/d9fab5459563a779c65b97040ad4d2e41f0b9464.pdf

The routes for three of the four molecules (none yet tested in a lab) are provided here: https://www-cdn.anthropic.com/666a6cb5172cbf6faa83d45968334dcf44da1410.pdf

www-cdn.anthropic.com

MIPLIB is a library of optimization problems, such as how best to schedule airline crews and delivery trucks.

Of 203 open problems, Claude beat the listed mathematical solution in 72 cases, and in 45 cases proved that a solution was optimal.

One problem asks how to time 42 trains to cut peak power use. In under two hours, agents beat MIPLIB’s best timetable by 2.6%, and later proved no better timetable exists.

We’ve submitted the solutions to MIPLIB for anyone to check.

Fable 5.2 solved two open problems in theoretical physics related to black holes.

In the first, Fable 5.2 completed a laborious calculation that had gone unsolved for 23 years.