TL;DR
Anthropic's Claude autonomously solved a nine-loop quantum physics calculation that had stood as the field's frontier for three years, at a cost of roughly $1,000 to $2,000 with minimal human oversight.
What happened
- Claude completed the nine-loop calculation of six-particle scattering amplitudes in planar N=4 super Yang-Mills theory, a landmark problem in theoretical physics.
- The previous record, a nine-loop calculation by a SLAC team, had stood for three years before Claude surpassed it.
- Total compute cost came in at $1,000 to $2,000, a fraction of what a human research team would require in time and resources.
- The work was completed with minimal human supervision, meaning Claude operated largely as an autonomous research agent, not a human-guided tool.
Why it matters
- Frontier theoretical physics has historically required years of expert human effort; Claude compressing that to a single low-cost AI run signals a qualitative shift in scientific capability.
- The cost floor of $1,000 to $2,000 means any well-funded lab, startup, or even individual researcher could now attempt problems previously gated behind large institutional teams.
- This is a concrete, peer-verifiable benchmark: scattering amplitude calculations have exact answers, so this is not a subjective or benchmark-gamed result.
- The result raises the question of whether AI is now a co-author in theoretical physics, with implications for how credit, funding, and research priorities are structured.
- For Anthropic, this is a competitive signal against OpenAI, Google DeepMind, and others racing to demonstrate autonomous scientific reasoning at the frontier.
What to watch next
- Whether peer review or independent replication confirms the result, which would cement this as a genuine scientific milestone rather than a demonstration.
- Whether competing labs respond with their own models attempting ten-loop or higher calculations, turning scattering amplitudes into an AI benchmark race.
- How the physics community and funding bodies respond: if AI can routinely clear multi-year research milestones, grant structures and PhD pipelines face structural pressure.
Originally published on Present of AI, a daily source-linked AI news timeline. Read the full timeline or browse the open dataset.