Jon Barron details his workflow: Gemini runs experiments, drafts the paper, then rewrites in his style
jon_barron · x · 2026-10-24
Google researcher Jon Barron shared the AI-assisted workflow behind his latest arXiv paper: the bug was first found and fixed by Gemini Flash 3.7, most experiments were run by Flash 3.8, and Gemini 4 Argon wrote the first draft and figures, which he aggressively nitpicked until the content was solid but the writing was not.
He then fed Argon a handful of his own first-author papers from recent years and had it rewrite in his style — which "worked astonishingly well" — followed by a final manual rewrite by hand. He liked the workflow and suggests readers who like an author's style throw some of their old papers into the context window and do the same.
More from Models
- "70% of Kimi K3's original cost is pretty good," says AI commentator — ChrisGPT · 2026-10-24
- Grok bot demo shows sign-in autofill with credentials kept in its secure vault — billyjhowell · 2026-10-24
- Fireworks explains how it made RL practical on Kimi K3 at 2.8T parameters — sophiamyang · 2026-10-24
- Astra finds bug in Anthropic's formal verification example: spec doesn't check output matches input — teortaxesTex · 2026-10-24
- Update: open weights for the Fable-class local model coming November 4th — kimmonismus · 2026-10-24
- Microsoft & HF turn database checks into RL rewards, lifting Qwen3.8-27B agents from 51.7% to 60.78% — SergioPaniego · 2026-10-24