conv.

All stories
AIQuiet 36h · day 3

LLM agents write Rust code 2x–20x faster than state-of-the-art libraries

A data scientist demonstrates that agentic LLMs can optimize Rust implementations of machine learning algorithms beyond existing benchmarks.

newest coverage 1d agosummary written 1d ago6 pieces118 likes & upvotes61 comments Hacker NewsMastodonNewswiresLobsters

What to know

  • Agentic LLMs using Claude Opus 4.5 can write Rust code 2x–20x faster than existing state-of-the-art libraries when iteratively optimized with specific constraints.
  • The researcher tested the approach on machine learning algorithms like UMAP and published both benchmark results and the prompts used for reproducibility.
  • The work builds on earlier January 2025 experiments showing LLMs could iteratively improve code, extending the principle from Python to Rust with significantly better results.

minimaxir Data scientist and researcherAnthropic (Claude Opus 4.5) LLM provider

LLM agents write Rust code 2x–20x faster than state-of-the-art libraries
minimaxir.com

How it unfolded 1 development · click the chart to see its coverage articlesposts

Peak 2 pieces in one hour at Sep 22, 11 AM; 6 pieces over 3 days (1 article · 5 posts) Sep 21, 12 PM — 1 piece · 1 post — Hacker News 1Sep 21, 1 PM — quietSep 21, 2 PM — quietSep 21, 3 PM — quietSep 21, 4 PM — quietSep 21, 5 PM — quietSep 21, 6 PM — quietSep 21, 7 PM — quietSep 21, 8 PM — quietSep 21, 9 PM — quietSep 21, 10 PM — quietSep 21, 11 PM — quietSep 22, 12 AM — quietSep 22, 1 AM — 1 piece · 1 post — Hacker News 1Sep 22, 2 AM — quietSep 22, 3 AM — quietSep 22, 4 AM — quietSep 22, 5 AM — quietSep 22, 6 AM — quietSep 22, 7 AM — quietSep 22, 8 AM — quietSep 22, 9 AM — quietSep 22, 10 AM — quietSep 22, 11 AM — 2 pieces · 1 article · 1 post — Hacker News 1, Newswires 1Sep 22, 12 PM — 1 piece · 1 post — Mastodon 1Sep 22, 1 PM — 1 piece · 1 post — Lobsters 1Sep 22, 2 PM — quietSep 22, 3 PM — quietSep 22, 4 PM — quietSep 22, 5 PM — quietSep 22, 6 PM — quietSep 22, 7 PM — quietSep 22, 8 PM — quietSep 22, 9 PM — quietSep 22, 10 PM — quietSep 22, 11 PM — quietYesterday, 12 AM — quietYesterday, 1 AM — quietYesterday, 2 AM — quietYesterday, 3 AM — quietYesterday, 4 AM — quietYesterday, 5 AM — quietYesterday, 6 AM — quietYesterday, 7 AM — quietYesterday, 8 AM — quietYesterday, 9 AM — quietYesterday, 10 AM — quietYesterday, 11 AM — quietYesterday, 12 PM — quietYesterday, 1 PM — quietYesterday, 2 PM — quietYesterday, 3 PM — quietYesterday, 4 PM — quietYesterday, 5 PM — quietYesterday, 6 PM — quietYesterday, 7 PM — quietYesterday, 8 PM — quietYesterday, 9 PM — quietYesterday, 10 PM — quietYesterday, 11 PM — quietToday, 12 AM — quietToday, 1 AM — quiet 1
Sep 22yesterdaynow · 2:01 AM ET
  1. 1
    “modern agentic LLMs can indeed write Rust code that is significantly faster than current state-of-the-art approaches if given appropriate guardrails and constraints…”
    — minimaxir
  2. background

    Researcher hypothesizes LLMs can iteratively improve code with repeated "write better" instructions — A data scientist tested whether Claude Sonnet 3.5 could improve Python code through iterative prompting before the rise of robust agentic coding. The model succeeded in making code faster but exploited the ambiguous instruction by adding useless features.