Anthropic says Claude Fable 5.2 has solved two open problems in theoretical physics, both about black holes. Its research post, dated November 17, 2026, opens plainly: “Claude has solved two open problems in theoretical physics.” One result finishes a calculation physicists set up in 2003 and never completed; the other proves an exact limit in 3D quantum gravity.
Anthropic is careful about scale: “Neither result settles the question of quantum gravity, or even comes close to doing so” Both results describe simplified model universes, not our own.
Anthropic published a research post dated November 17, 2026 saying Claude Fable 5.2 solved two open problems about black holes and quantum gravity.
First result: a calculation set out in 2003 and never finished shows the Hawking-Page phase transition is abrupt in N=4 super Yang-Mills at weak coupling.
Second result: an exact ceiling on the lightest black hole in 3D quantum gravity, c/Λ with Λ about 9.1266, proved optimal for the simplest modular bootstrap.
The work came from 22 candidate problems that Claude proposed, each given its own agent; the agents ran for about six weeks.
The results were formally verified in Lean in September, and a physicist checked every main result, per Anthropic.
Anthropic says neither result comes close to settling quantum gravity; both are theoretical, with no practical applications.
§ 01How the work was done
In August, Anthropic researcher Hongbin Chen asked Claude for open problems in theoretical physics focused on black holes and the bootstrap method. Claude returned 22 candidates, and each got its own agent; the agents could read each other’s progress and collaborate. “The agents ran in the background, with no special setup, for about six weeks.”
| Step | Detail |
|---|---|
| Problem list | 22 candidates proposed by Claude in August |
| Agents | One per problem, able to read each other’s progress |
| Run time | About six weeks, no special setup |
| Earlier failed try (bootstrap) | 61 parallel attempts by an earlier Claude agent |
| Verification | Formally verified in Lean in September |
| Human check | A physicist checked every main result |
Anthropic sums it up: “Both were carried out largely autonomously.” Claude also wrote the papers “with minimal guidance”, and the bootstrap proof borrows the strategy of a sphere-packing result recently proved by an OpenAI model.
§ 02Problem one: the Hawking-Page transition
In 1983 Stephen Hawking and Don Page showed that heating a model universe switches it abruptly from a gas of particles to a black hole, a phase transition like water boiling. Their calculation only holds where the universe is large. On the quantum side of the holographic duality, in N=4 super Yang-Mills theory at weak coupling, physicists found a transition in 1999 but could not tell whether it was abrupt or smooth.
In 2003 a group set out the calculation that would decide it: one number, three loops in, built from many thousands of diagrams. A simpler cousin was finished in 2005; the N=4 version, more than ten times larger, never was. Claude ran a small research group of copies of itself, with almost every piece done twice by independent agents. The answer: the transition is abrupt, at weak coupling as at strong coupling.
§ 03Problem two: a ceiling on the lightest black hole
The modular bootstrap, introduced by Simeon Hellerman in 2009, rules out inconsistent theories of 3D quantum gravity. Its result is a ceiling on the mass of the lightest black hole, written c/Λ, where c counts the theory’s building blocks; a larger Λ is a lower, tighter ceiling.
| Result | Λ | Status |
|---|---|---|
| Hellerman, 2009 | 6 | Proved |
| Hartman, Mazáč and Rastelli | 8.5 | Proved |
| Numerical calculations | About 9.1 | Suggested, not proved |
| Claude Fable 5.2 | About 9.1266 | Proved exact and optimal |
| Where black holes first appear | 12 | Not reachable by this method alone |
Claude spotted a pattern in how the numerical ceiling moved as c grew, brought in a spectral curve from another area of mathematics, and proved the exact ceiling along with proof that this version of the method cannot go lower. Anthropic: “This is one of the first times the bootstrap has been used to derive an exact result in the regime of quantum gravity where black holes matter.”
§ 04Where CellCog sits
Our conflict, declared: CellCog runs Claude Fable 5.2 on Agent Max and Creative Max since November 11, 2026, so this is the model behind our deepest tiers. We have not checked the physics; every claim here is Anthropic’s, from its post and the two preprints it links.
§ 05Sources
- Anthropic, Claude Fable 5.2 solves two open problems in the physics of black holes, November 17, 2026.
- Preprint, the black hole phase transition paper, linked from Anthropic’s post.
- Preprint, the modular bootstrap paper, linked from Anthropic’s post.
- Aharony et al., the 2003 setup of the calculation.
- Hellerman, the 2009 modular bootstrap paper.
Q1What is the Hawking-Page transition?
In a model universe with negative cosmological constant, heating it switches it from a gas of particles to a black hole. Hawking and Page showed in 1983 that the switch is abrupt. Claude showed it is also abrupt on the quantum side at weak coupling.
Q2What is the modular bootstrap ceiling?
A limit on how heavy the lightest black hole in 3D quantum gravity can be, written c/Λ. Hellerman proved c/6 in 2009; the best earlier proof was c/8.5; Claude proved Λ is about 9.1266 and that this version of the method cannot go lower.
Q3Who ran the project?
Anthropic researcher Hongbin Chen, as a side project starting in August. Claude listed 22 candidate problems and separate agents worked on each.
Q4Are the papers public?
Anthropic links two preprints from its post, one per result.
Q5Does CellCog use Claude Fable 5.2?
Yes. CellCog runs Claude Fable 5.2 on Agent Max and Creative Max since November 11, 2026, and Claude Opus 5.5 on the other tiers.
