TheoremSmith: Mass-Producing Hard Proof Problems to Train LLMs
Get picks like this daily. Today's launch numbers, the top launches, AI and tech news, and a daily opportunity pick, straight to your inbox.
Want to reach these readers? Sponsor an issue.
Making models better at formal verification will help push the boundaries of math research and significantly reduce the number of bugs in our code. Recently, training models on RL environments (tasks paired with verifiers that reward correct solutions) has driven big gains in model performance, and proofs are unusually easy to verify. Here we introduce TheoremSmith, a pipeline for generating synthetic formal verification environments. It reliably creates multi-hour, frontier-difficulty tasks and can scale throughput effectively.I'd appreciate any feedback, especially on ways models could still cheat the verifier, and on whether synthetic tasks like these are likely to help on real proofs.Code: https://github.com/abundant-ai/theoremsmith
HN · launched Nov 10, 2026 · 10 upvotes (top 40%) · 0 views
Niche: specialized algorithms and computing libraries · 230 launches, 5 in the last 7 days ▼ -64% · Explore this niche →
Category and tags
- Horizontal
- Data infrastructure
- B2B
- AI-native
- Commercial product
Visit website ↗ Source listing ↗
More like this: all 230 launches in specialized algorithms and computing libraries →
Could you build this?
No The product generates hard formal-verification proof problems and RL environments, which needs deep expertise in proof assistants, RL training, and substantial compute.
What it would actually take: A real version needs researchers fluent in formal methods (Lean, Coq, or Isabelle) to design verifiable problem generators that are hard but solvable. It also needs RL engineers to build verifier-based training environments and large compute budgets to run frontier model evaluations. This is specialized research work that vibe coding cannot substitute for.
Discussion
No comments on this launch.
Competitors
Other products that read as similar to this one : 1097 launches clear the similarity bar, closest 8 shown.
Attention rank: #441 of 1098 (itself plus its competitors, highest first, normalized so YC and Product Hunt are compared fairly).
Launched 375 days after the earliest competitor.
- Cajal: Scaling Formal Verification for Scientific Discovery · yc · 2026-02-24 · 15 upvotes · similarity 0.58
- zkGolf · hn · 2026-07-02 · 69 upvotes · similarity 0.56
- genpark-loop-invariant-induction-prover-skill · github · 2026-09-09 · 8 upvotes · similarity 0.54
- Graph-Oriented Generation · hn · 2026-03-06 · 12 upvotes · similarity 0.52
- MathmoBench · github · 2026-09-23 · 88 upvotes · similarity 0.51
- Talos · hn · 2026-06-18 · 106 upvotes · similarity 0.51
- Spivak's Calculus formalized in Lean 4 · hn · 2026-09-26 · 22 upvotes · similarity 0.51
- Ramanujan · hn · 2026-09-06 · 5 upvotes · similarity 0.51
The niche, labels, vibe code rating and competitor comparison on this page are analysis by Nicheloom. If you use them, please credit nicheloom.com. How you may use our data.