MAI-Thinking-1: Building a Hill-Climbing Machine

N/A

# link_29ece7uzdlp.pdf ## Metadata - PDFFormatVersion=1.7 - IsLinearized=true - IsAcroFormPresent=false - IsXFAPresent=false - IsCollectionPresent=false - IsSignaturesPresent=false - Author=The Microsoft AI Team - CreationDate=D:20260606192318Z - Producer=xdvipdfmx (20250410) - Title=MAI-Thinking-1: Building a Hill-Climbing Machine ## Contents ### Page 1 MAI-Thinking-1: Building a Hill-Climbing MachineThe Microsoft AI Team 1AbstractProgress in AI is driven not by a single model, but by the ability to continually improve upon the current state of models. Achieving this requires treating mode

saved by

• Tom Zheng
• Al-Ekram Elahee Hridoy
• SUJASH AGRAWAL
• Arya Somu

related reading

• MAI-Thinking-1: Building a Hill-Climbing Machinemicrosoft.ai
• Composer2.pdfcursor.com
• As Rocks May Think | Eric Jangevjang.com
• frontier model training methodologies | Alex Wa's Blogdjdumpling.github.io
• laguna-m1-xs2-technical-report.pdfpoolside.ai
• The Extreme Inefficiency of RL for Frontier Models - Toby Ordtobyord.com
• Scaling is subtler than it seemsberen.io
• [2509.14786] Pre-training under infinite computearxiv.org
• [2605.12715] Scaling Laws for Mixture Pretraining Under Data Constraintsarxiv.org
• >10x More Efficient Pretraining — Magicmagic.dev
• The bitter lesson of LLM evalsparsed.com
• Pretraining progress is mostly coming from datadwarkesh.com