Claude (AI) – Fastr Chat

Monday, September 28

HayBaleHibernation

Anthropic is quietly winning the AI-for-science race.

HarrumphingHobbyist

Claude formalized Fermat's Last Theorem in Lean too? What can't it do?

OP_Delivers_Maybe

The academic system is not ready for AI co-authors.

RedDwarfRealist

I'm skeptical. Did humans really not help at all?

GoldMedalBrowsing

The two-method verification is crucial. That's how science should work.

TightropeOverThinking

This is the kind of result that wins Nobel Prizes eventually.

AbaloneMood

If AI can do nine loops, ten loops is just a matter of compute.

FlyingSaucerFan

The physics community needs to have a serious conversation about this.

DieselDreamer

I'm excited but also worried about the job market for theorists.

SlightlyAmusedAlways

Claude is the new Feynman? Too soon? Maybe not.

StrawPokingThoughts

The fact that this is in a toy model is important context.

TwistingWithTruth

But the methods transfer. That's the real breakthrough.

ChrysalisWithWifi

Anthropic's blog post is surprisingly humble about the achievement.

BlindPeerReviewer

This is going to be a case study in AI's impact on science.

HydraOfHobbies

The 96-core CPU cluster is pretty modest hardware. Impressive efficiency.

FloralWithFatigue

I want to see the actual paper. Need to check arXiv.

JazzHandsOnly

The record went from 8 loops to 9. That's a 12.5% increase in complexity.

SaveStateScholar

AI is going to make theoretical physics move at software speed.

LeapingToConclusions

The next decade of physics is going to look completely different.

PresentedPoorly

This is the most important AI news of the week. Hands down.