Extending single-minus amplitudes to gravitons

AI says some 'impossible' gravity collisions happen after all

TLDR: Researchers say a rare gravity interaction once believed impossible does occur under special motion, aided by GPT‑5.2 Pro. Comments split between skeptics calling it demo-hype and a former physicist praising human audit of AI math—making this a notable moment for science and AI.

Physics minds dropped their chalk today: a team says some gravity “collisions” long thought to be impossible actually show up under special motion, and they credit GPT‑5.2 Pro for helping derive the math. Translation: in certain alignments, tiny packets of gravity (gravitons) can interact when everyone said “nope.” The authors posted a preprint and even a model-written draft, sending the internet into a spin. For non-physicists: scattering amplitudes are just math that predicts what happens when particles bump; this result says a rare setup isn’t a dead end after all. Cue applause—and eye-rolls.

The comment war lit up fast. behnamoh slammed the AI angle: “LLMs are good for happy-path demos, that’s it,” warning models can’t handle truly new problems. micheles, a former theorist, clapped back: “this result is remarkable,” arguing the future is checking AI’s work, not hand-cranking pages of algebra. Meanwhile adampunk turned it into a meme: “That’s one heck of a next token to predict.” Fans cheered the math hack (a matrix-tree theorem trick) and symmetry talk; skeptics saw hype and cherry-picked kinematics. It’s hope vs. nope: is AI cracking real physics or just guessing fancy? Either way, the gravity of the discourse is off the charts.

Key Points

  • A new preprint shows single-minus graviton tree amplitudes can be nonzero under specific kinematics in quantum gravity.
  • Nonvanishing occurs in the half-collinear regime, where amplitudes exist as distributions supported on restricted momentum-space regions.
  • The authors derive explicit formulas based on symmetry principles and recursion relations.
  • The amplitudes realize an infinite-dimensional w-(1+∞) symmetry linked to efforts to quantize gravity.
  • GPT‑5.2 Pro assisted by extending prior gluon results to gravity using the directed matrix-tree theorem and drafting the initial paper.

Hottest takes

“LLMs are good for happy-path demos, that’s it.” — behnamoh
“That’s one heck of a next token to predict.” — adampunk
“this result is remarkable” — micheles
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.