{"version":1,"type":"story","url":"https://digestai.news/story/resolution-autoformalizes-math-to-speed-ai-alignment-research","json":"https://digestai.news/story/resolution-autoformalizes-math-to-speed-ai-alignment-research.json","markdown":"https://digestai.news/story/resolution-autoformalizes-math-to-speed-ai-alignment-research.md","slug":"resolution-autoformalizes-math-to-speed-ai-alignment-research","headline":"Resolution autoformalizes math to speed AI alignment research","summary":"Resolution, an AI safety startup, published work on autoformalizing mathematical statements related to statistical learning theory. The team used agents to develop plans—one high-level and one detailed—and an orchestrator agent executed the work, producing 600k lines of Lean code in about a week. The process cost roughly **$135,000** in token spend and involved **100B input tokens**, with **300M output tokens** (90% cache reads). The team claims this lays groundwork for automated research into model generalization and learning patterns, calling it part of an ‘industrialization of mathematics’ that could accelerate AI safety work.\n\nThe effort highlights how AI agents can now handle tasks previously too expensive or time-intensive for human researchers. Resolution notes the work has already sped up other aspects of their research, such as automating parts of singular learning theory to better understand sampling. The startup’s approach demonstrates how AI tools are reshaping frontier research, potentially enabling rigorous automation of scientific discovery.","keyPoints":["Resolution autoformalized math on statistical learning theory using AI agents, producing 600k lines of Lean code in a week","Process cost ~$135k and consumed 100B input tokens, 300M output tokens, with 90% cache reads","Team says work accelerates AI alignment research by automating theory and improving model generalization understanding"],"whyItMatters":"Autoformalizing math with AI could drastically cut costs and time for frontier research, enabling startups to tackle problems previously beyond their reach. This method may help refine AI alignment by automating complex theoretical work, potentially speeding up safety advancements.","category":{"slug":"research","name":"Research","url":"https://digestai.news/category/research"},"entities":{"companies":["Resolution","Mirendil","OpenAI","Vals"],"models":["Claude Opus 5.5","Nemotron-550B-Base","GLM-5.3","DeepSeek-V4-Pro-Base","Kimi-K3","Mirendil-Base"],"people":["Jack Clark"]},"firstPublishedAt":"2026-10-26T12:31:52Z","updatedAt":"2026-10-26T12:31:52Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"Import AI","title":"Import AI 478: Math for alignment; ProgramBench; and Mirendil shows what RSI means","url":"https://jack-clark.net/2026/10/26/import-ai-478-math-for-alignment-programbench-and-mirendil-shows-what-rsi-means","publishedAt":"2026-10-26T12:31:52Z","type":"newsletter","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"Resolution autoformalizes math to speed AI alignment research\", 26 October 2026, https://digestai.news/story/resolution-autoformalizes-math-to-speed-ai-alignment-research","publisher":"Digest AI","title":"Resolution autoformalizes math to speed AI alignment research","datePublished":"2026-10-26T12:31:52Z","url":"https://digestai.news/story/resolution-autoformalizes-math-to-speed-ai-alignment-research"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}