AI Math Progress Accelerating: Researcher Predicts 100% of Top Results Will Be LLM-Assisted
tokenbender · x · 2026-08-02
AI researcher Lu Sichu stated that the progress of Large Language Models in mathematics is happening so fast that it gave him whiplash, causing him to revise his previous pessimistic predictions.
A couple of months ago, he predicted that by the end of 2026, 10-20% of the best mathematical results would be LLM-generated or assisted. Now, he believes that figure is increasingly looking like 100%. He emphasized that AI is breaking open everything, and even if human mathematicians prove neat things in the future, it will likely be because they read AI-inspired papers before anyone prompted a model to do it.
More from AGI Musings
- Quantum expert responds to Economist: clarifying misconceptions and setting realistic expectations — whurley · 2026-08-02
- a16z's Martin Casado: Model Capability Growth Appears Sublinear Compared to Funding — jeremyphoward · 2026-08-02
- Why Finance Insiders Are the Least Impressed by AI — devanshmehta · 2026-08-02
- Researchers vs LLMs: Human Intellect Still Drives AI Innovation — owl_posting · 2026-08-02
- Hot Take: AGI Can Be Achieved Without Continual Learning — adonis_singh · 2026-08-02
- tinygrad Founder: Closed US LLM Labs Cornered by the Open-Source World — mohbibi_ · 2026-08-02