Fetching from the wire…
01
02
03
04
05
06
07
08
Top 5 · 2026-06-01 · source-backed
Microsoft Research dropped a paper that should change how every builder thinks about their agent configuration files. SkillOpt (arXiv 2605.23904) treats a Markdown document as an external parameter of a frozen LLM and applies learning rate, batch, and momentum concepts in text space. Not metaphorically. Literally. They train your skill file.
The results are absurd. Across 6 benchmarks, 7 models, and 3 execution harnesses (direct chat, Codex, Claude Code), SkillOpt wins or ties on all 52 evaluated cells. On GPT-5.5: +23.5 points in direct chat, +24.8 in Codex, +19.1 in Claude Code. Those aren't marginal improvements. That's the difference between a mediocre agent and a good one.
I maintain a CLAUDE.md file for every project I work on. Most builders I know do the same. We've been treating prompt engineering as intuition, taste, trial and error. SkillOpt says: no, this is an optimization problem with a tractable solution. You can systematically search the space of possible skill file configurations and find ones that measurably improve your model's task performance.
The mechanism: SkillOpt iteratively edits sections of the Markdown skill file, evaluates performance on a held-out task set, and applies something analogous to gradient descent in text space. The "learning rate" controls how aggressively edits are made. The "momentum" preserves successful patterns across iterations. It's MIT licensed and open source.
This connects directly to the cost story. If Copilot is now charging you per token, and SkillOpt can get you better results in fewer attempts (because the model understands your intent more precisely on the first try), then optimized skill files aren't just about quality. They're about economics. Better skills = fewer wasted tokens = lower bills.
What builders should do: if you maintain CLAUDE.md, AGENTS.md, .cursor/rules, or any agent skill files, this is worth running against your actual task distribution. The assumption that "prompt engineering is art not science" just took a hit. It's still art. But now there's a gradient.
Each link below shares sources, entities, or timing with this story.
Codex competes with Claude Code / Shared entities / Shared topic / Earlier coverage
Linked by a graph relationship (Codex competes with Claude Code); both cover CLAUDE, Claude Code, Codex, LLM; overlapping topics (agent, claude, code, file, markdown).
Claude Code benchmarked against GPT / Shared entities / Same source domain / Shared topic / What happened next
Linked by a graph relationship (Claude Code benchmarked against GPT); both cover LLM, MIT, Most, Those; reported by the same outlet (arxiv.org).
Garry Tan uses Claude Code / Shared entities / Shared topic / Earlier coverage
Linked by a graph relationship (Garry Tan uses Claude Code); both cover Claude, Claude Code, MIT, Those; overlapping topics (agent, claude, code, model, skill).
Claude Code competes with Cursor / Shared entities / Shared topic / What happened next / Tension
Linked by a graph relationship (Claude Code competes with Cursor); both cover Claude, Claude Code, Codex, Markdown; overlapping topics (agent, claude, code).
Codex competes with Claude Code / Shared entities / Same source domain / Shared topic / What happened next
Linked by a graph relationship (Codex competes with Claude Code); both cover Claude, Claude Code, Codex; reported by the same outlet (arxiv.org).
Linked by a graph relationship (Codex competes with Claude Code); both cover Claude Code, Codex, LLM; reported by the same outlet (arxiv.org).
Codex competes with Claude Code / Shared entities / Same source domain / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Codex competes with Claude Code); both cover CLAUDE, Claude Code, Codex; reported by the same outlet (arxiv.org).
Codex competes with Claude Code / Shared entities / Shared topic / What happened next
Linked by a graph relationship (Codex competes with Claude Code); both cover CLAUDE, Claude Code, Codex, MIT; overlapping topics (aren, claude, code, model).