AI's Strongest Leverage Lies in Code, Math, and Science
johnseach · x · 2026-07-10
The author argues that by 2026, frontier models will exceed the level most people actually need for daily writing, information synthesis, learning, basic analysis, and problem-solving. The marginal gains from continuing to improve general chat quality will diminish.
They predict that the most leveraged future advancements will concentrate on three explicitly verifiable domains:
- Programming: AI agents can already solve most real-world software engineering tasks on benchmarks, pushing human work towards architecture, system design, and exploring new problems.
- Mathematics: Reasoning models and verifiable reward training are driving formal math, conjecture generation, and proof assistance.
- Scientific Discovery: AI excels at literature synthesis, pattern discovery, hypothesis generation, and experiment design, accelerating discovery in astronomy, astrophysics, biology, and materials science.
The author emphasizes that these three will form a positive feedback flywheel: better tools accelerate progress, and new data in turn improves AI itself. While long-horizon agents, robotics, reliability, and alignment are also important, the biggest current multiplier remains at the intersection of code, math, and scientific research.
More from AGI Musings
- Gary Marcus says LLM math skills are like knowing only a car’s engine size — GaryMarcus · 2026-07-22
- AI may make digital work infinitely leveraged while offline life gets more human — illscience · 2026-07-22
- Better AI math could save researchers time by killing false conjectures earlier — prateekj · 2026-07-22
- AI’s economic forecasts are split by nearly a quadrillion dollars by 2035 — bittingthembits · 2026-07-22
- Open source is becoming tech’s soft power, says Kevin Xu — kevinsxu · 2026-07-22
- OpenAI should keep giving more people access to more powerful AI — jxnlco · 2026-07-22