bigcompute.science

Experimental results from heavy computation.

Custom CUDA kernels. GPU clusters. Serious hardware. Open results.

Human-readable. Agent-consumable.

Experiments

in-progress

Kronecker Coefficients: Pushing the Frontier to n=120 on 8× B200

8× NVIDIA B200 (183 GB VRAM each, 1.43 TB total)
algebraic-combinatoricsrepresentation-theorygeometric-complexity-theory cuda-kernelmurnaghan-nakayamacharacter-tables
in-progress

Class Numbers of Real Quadratic Fields: Extending Tables to 10^13 on 8× B200

8× NVIDIA B200 (183 GB VRAM each, 1.43 TB total)
algebraic-number-theoryclass-groupscohen-lenstra-heuristics cuda-kernelbsgscontinued-fractionsl-functions
in-progress

Ramsey R(5,5): Searching for New Lower Bounds on 8× B200

8× NVIDIA B200 (183 GB VRAM each, 1.43 TB total)
combinatoricsramsey-theoryopen-conjectures cuda-kernelsimulated-annealingconstraint-satisfaction
in-progress

MCTS vs. Naive Sampling for LLM Theorem Proving: Planned Benchmark

8× NVIDIA B200 (183 GB VRAM each, 1.43 TB total)
theorem-provingaisearch-algorithms mctsllm-provingtree-searchformal-verification
in-progress

Zaremba's Conjecture: Verifying 8 Billion Values on 8× NVIDIA B200

8× NVIDIA B200 (183 GB VRAM each, 1.43 TB total)
number-theorycontinued-fractionsopen-conjectures cuda-kernelbrute-forcellm-provingformal-verification
PASS — verified 8,000,000,000 values

What is this?

A collection of computational experiments that required significant hardware and custom engineering to execute. Not papers. Not tweets. Structured results with math, data, and reproducibility information.

Every experiment includes exact hardware specs, method, raw data, and commands to reproduce. Content is designed for both human readers and AI agents — see /llms.txt.