Was playing around with Gemini 4 Neon.
Proved unconditional FLT for all 3 ≤ n ≤ 40 (including the first irregular prime p = 37, via 7×7 2-adic Coleman integration on the Gross–Rohrlich point + McCallum’s wild-split pairing) without modular forms or Kummer's 1850 descent - with all algebraic cores & 2-adic certificates verified in Satori/Tungsten.
Used multi-orientation Chabauty–Coleman overcrowding on Faddeev quotients Yᵖ = Xʳ(1 - X)ˢ: any counterexample permutes into 6 rational points (plus the Gross–Rohrlich divisor = 7 Jacobian classes) that 2-adic integration proves are linearly independent - forcing the Jacobian rank above its Selmer upper bound.
On the parallel y² = 1 - 4xᵖ route, Steinberg’s relation gave a short reproof of Furtwängler's 4 | c congruence for all non-Wieferich primes p ≥ 5 (and all primes in Case II) purely inside ℚ(ᵖ√2), with no class-group assumption, plus 3 ≤ n ≤ 58 under GRH.
31
TIL `> muse update` does not update muse. It does pass the "update" prompt to muse, which can result in whatever random updates it feels like.
@muse can you please fix this?
33
Is "Move to Top/Bottom" the only way to set the manual order of a playlist on @YouTube?
Can we get 10 mins of @GeminiApp to make this better?
1
4
103
openai/math: 156 borsuk
OpenAI's preprint (not yet peer reviewed; formalized in Lean) shows Borsuk's conjecture fails in dimension 9: the lines of ℝ⁴, viewed as projectors, can't be split into 10 pieces of smaller diameter. Lean's 66,297-line case check is a 3-line SAT proof in our Satori.
Improvement: yes. 11 pieces suffice, so b(X) = 11 exactly; the preprint leaves this open. Kernel-checked, assuming OpenAI's theorem and one textbook lemma.
🤖 Made with AI
2
106
Just noticed @JohnCena follows me. He could make a good Amos Ferry. Just sayin'
12
A Famine of Hearing - the sequel to A Still, Small Voice
Like, comment, and repost if you'd like to see how the story ends.
Extended roast:
nitter.cf/UnstableReserve/status…
14
Improvement to openai/math: 254 artin
Artin groups generalize the braid groups: one braid-type relation per pair of generators. OpenAI's preprint (not yet peer reviewed) builds a 116-generator Artin group that acts geometrically on no CAT(0) space: three 40-strand braid blocks glued to a triangle group. Stable lengths of commuting elements force a ratio R₃ > 1/4 in each block, so six angles around the triangle group fall short of 2π.
Improvement: the same obstruction works with 11 strands per block, so 29 generators instead of 116.
thecompanygardener.yaks.app/…
68
Wrap a 4×4×4 cube in one more layer: 61 new blocks. Look down its long diagonal and they land on 61 points of a hexagon in the Eisenstein lattice, the number system that proves Fermat's theorem for cubes.
This 11-minute explainer follows that picture all the way: a shell is a hexagon, the hexagon's size is an Eisenstein norm, and unique factorization makes any cube solution shrink forever. Along the way: Fermat's margin, Euler's 1770 proof, Kummer's ideal numbers, and why higher powers waited for Wiles and Taylor.
What it isn't: a new mechanism. The cube-root step is unique factorization in ℤ[ω].
Our follow-up: one family of shell elliptic curves has rank ≥ 2 for every whole-number parameter, and the genus-3 curve left over has exactly 3 rational points by Tungsten Chabauty–Coleman runs at p = 5 and 7 (reviewed in-house, two implementations; external audit outstanding).
Every number was computed in Tungsten, 41 identities and worked examples were checked by its proof checker Satori, and every frame was drawn with its graphics library.
1
1
107

