bshepp/3body-l4-log
222 MB
Brian Sheppard PRO
bshepp
AI & ML interests
Reinforcement Learning, Computer VIsion, Fundimental Mathematics, Fundimental AI architecture, AI Safety(for the AI), helping people
Recent Activity
updated a dataset about 24 hours ago
bshepp/rl-surrogate-dft-gap published a dataset about 24 hours ago
bshepp/rl-surrogate-dft-gap upvoted an article 2 days ago
📄 Research Papers DatasetOrganizations
Posts 3
Post
152
A dead 2013 Butterfly Labs "Jalapeno" SHA-256 mining ASIC sat in a drawer for a decade. It became the excuse for a small, careful question: how much structure can a tiny, cheap model learn in SHA-256, and how would I know if I were fooling myself? (The ML runs on CPU and a HF job, not the ASIC; the dead miner is just the origin story.)
Three findings, written up honestly:
1. A sharp round-4 cliff. Round-reduced SHA-256 is ~100% distinguishable through 3 rounds, then collapses to chance at round 4 and stays there out to the full 64. Reproduced across 5 seeds.
2. A controls-gated bounded null on full SHA-256: no learnable structure above a ~0.22% resolution floor at n=4,000,000. That is a bounded null at this budget, not a claim that SHA-256 is random.
3. A "signal" in the iterated-hash dynamics that a permuted-label control unmasked as a label-prior artifact. The instrument caught its own false positive. That was the point of building the controls.
Negative results, stated with their resolution. The dataset carries the controls on every row.
Dataset: bshepp/round-reduced-sha256-learnability
Code (MIT) + full writeup: https://github.com/bshepp/bfl-asic
Three findings, written up honestly:
1. A sharp round-4 cliff. Round-reduced SHA-256 is ~100% distinguishable through 3 rounds, then collapses to chance at round 4 and stays there out to the full 64. Reproduced across 5 seeds.
2. A controls-gated bounded null on full SHA-256: no learnable structure above a ~0.22% resolution floor at n=4,000,000. That is a bounded null at this budget, not a claim that SHA-256 is random.
3. A "signal" in the iterated-hash dynamics that a permuted-label control unmasked as a label-prior artifact. The instrument caught its own false positive. That was the point of building the controls.
Negative results, stated with their resolution. The dataset carries the controls on every row.
Dataset: bshepp/round-reduced-sha256-learnability
Code (MIT) + full writeup: https://github.com/bshepp/bfl-asic