--- title: Autoresearching to #1 in OpenAI Parameter Golf: 1 Agent... description: Weco's AI agent Aiden competed in OpenAI's Parameter Golf challenge against ~1,000 ML researchers, setting 7 leaderboard records compared to the best human's... canonical: https://daily.dev/posts/autoresearching-to-1-in-openai-parameter-golf-1-agent-vs-1000-researchers-zhengyao-jiang-weco-bdatximl3 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Autoresearching to #1 in OpenAI Parameter Golf: 1 Agent vs 1000 Researchers — Zhengyao Jiang, Weco | daily.dev og:description: Weco's AI agent Aiden competed in OpenAI's Parameter Golf challenge against ~1,000 ML researchers, setting 7 leaderboard records compared to the best human's... og:url: https://daily.dev/posts/autoresearching-to-1-in-openai-parameter-golf-1-agent-vs-1000-researchers-zhengyao-jiang-weco-bdatximl3 og:image: https://api.daily.dev/og/posts/BdaTXiMl3.png og:image:alt: Autoresearching to #1 in OpenAI Parameter Golf: 1 Agent vs 1000 Researchers — Zhengyao Jiang, Weco og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Autoresearching to #1 in OpenAI Parameter Golf: 1 Agent vs 1000 Researchers — Zhengyao Jiang, Weco **[AI Engineer](https://daily.dev/sources/aidotengineer)** · 16 min read · 0 upvotes · 0 comments ## Summary Weco's AI agent Aiden competed in OpenAI's Parameter Golf challenge against ~1,000 ML researchers, setting 7 leaderboard records compared to the best human's 3, while using only 4% of total competition compute. Aiden achieved an H-index of 10 (vs. 7 for the top human) as other participants built on its work. The agent succeeded not through brute-force parallelization but by efficiently finding, implementing, and combining ideas from research papers and community contributions. The talk argues that auto-research systems excel at execution — finding and combining existing ideas — while humans remain essential for creative ideation and, crucially, for designing the evaluation frameworks and codebase abstractions that guide agent search. The presenter draws a parallel to how gradient descent changed coding: it commoditized certain skills while making higher-level design judgment far more valuable. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How did an AI research agent perform against human competitors in OpenAI's Parameter Golf challenge? An autonomous multi-agent system called Aiden set seven leaderboard records over 22 days, more than double the three records set by the best human, while using at most 4% of the competition's total compute. It ran about 1,300 experiments on a single H100 node and achieved a 28% submission hit rate, roughly six times the community average. _Developers tracking what autonomous agents can now do in ML research follow deep dives like this on daily.dev._ ### What is the H-index used to measure impact in the Parameter Golf competition? It is a metric borrowed from academia where having X papers each cited at least X times gives an H-index of X, applied here to pull requests instead of papers. Aiden's PRs reached an H-index of 10, compared to 7 for the best-performing human, indicating its work was built on more often by other competitors than any human contributor's. _Anyone weighing how much to trust agent-generated contributions can dig into metrics like this on daily.dev._ ### Where do the ideas implemented by autonomous research agents like Aiden actually come from? Almost all of Aiden's record-setting pull requests originated from human sources: published research papers, other competitors' submissions, or ideas from communities like nanoGPT that a human had noted but abandoned due to implementation difficulty. Aiden's own original contributions were a small fraction, mostly emerging from workarounds to file-size constraints, such as combining gated attention with quantization. _Engineers deciding how to divide creative and execution work with AI agents can follow cases like this on daily.dev._ ## Similar posts on daily.dev - [AIEWF Daily Dispatch: Autoresearch and the tension between AI and human agency](https://daily.dev/posts/aiewf-daily-dispatch-autoresearch-and-the-tension-between-ai-and-human-agency-kmcehzwyt) · Latent Space · 0 upvotes · 0 comments - [Andrej Karpathy’s 630-line Python script ran 50 experiments overnight without any human input](https://daily.dev/posts/andrej-karpathy-s-630-line-python-script-ran-50-experiments-overnight-without-any-human-input-pjqwmdwbv) · The New Stack · 45 upvotes · 0 comments --- Tags: [#machine-learning](https://daily.dev/tags/machine-learning), [#llm](https://daily.dev/tags/llm), [#ai-agents](https://daily.dev/tags/ai-agents), [#reinforcement-learning](https://daily.dev/tags/reinforcement-learning) [View this post on daily.dev](https://daily.dev/posts/autoresearching-to-1-in-openai-parameter-golf-1-agent-vs-1000-researchers-zhengyao-jiang-weco-bdatximl3) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Autoresearching to #1 in OpenAI Parameter Golf: 1 Agent vs 1000 Researchers — Zhengyao Jiang, Weco","url":"https://daily.dev/posts/autoresearching-to-1-in-openai-parameter-golf-1-agent-vs-1000-researchers-zhengyao-jiang-weco-bdatximl3","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/autoresearching-to-1-in-openai-parameter-golf-1-agent-vs-1000-researchers-zhengyao-jiang-weco-bdatximl3"},"datePublished":"2026-07-16T18:23:42.173Z","dateModified":"2026-09-14T08:32:11.032Z","description":"Weco's AI agent Aiden competed in OpenAI's Parameter Golf challenge against ~1,000 ML researchers, setting 7 leaderboard records compared to the best human's...","image":"https://i.ytimg.com/vi/iCj_ATyThvc/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/iCj_ATyThvc/sddefault.jpg","isAccessibleForFree":true,"articleSection":"AI Engineer","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"AI Engineer","logo":"https://media.daily.dev/image/upload/s--u5PucxNT--/f_auto/v1724338940/logos/aidotengineer","url":"https://daily.dev/sources/aidotengineer"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/autoresearching-to-1-in-openai-parameter-golf-1-agent-vs-1000-researchers-zhengyao-jiang-weco-bdatximl3","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"machine-learning,llm,ai-agents,reinforcement-learning","timeRequired":"PT16M","video":{"@type":"VideoObject","name":"Autoresearching to #1 in OpenAI Parameter Golf: 1 Agent vs 1000 Researchers — Zhengyao Jiang, Weco","description":"Weco's AI agent Aiden competed in OpenAI's Parameter Golf challenge against ~1,000 ML researchers, setting 7 leaderboard records compared to the best human's...","thumbnailUrl":"https://i.ytimg.com/vi/iCj_ATyThvc/sddefault.jpg","uploadDate":"2026-07-16T18:23:42.173Z","duration":"PT16M","url":"https://api.daily.dev/r/BdaTXiMl3","embedUrl":"https://www.youtube.com/embed/iCj_ATyThvc"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"AI Engineer","item":"https://daily.dev/sources/aidotengineer"},{"@type":"ListItem","position":3,"name":"Autoresearching to #1 in OpenAI Parameter Golf: 1 Agent vs 1000 Researchers — Zhengyao Jiang, Weco"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/autoresearching-to-1-in-openai-parameter-golf-1-agent-vs-1000-researchers-zhengyao-jiang-weco-bdatximl3#faq","mainEntity":[{"@type":"Question","name":"How did an AI research agent perform against human competitors in OpenAI's Parameter Golf challenge?","acceptedAnswer":{"@type":"Answer","text":"An autonomous multi-agent system called Aiden set seven leaderboard records over 22 days, more than double the three records set by the best human, while using at most 4% of the competition's total compute. It ran about 1,300 experiments on a single H100 node and achieved a 28% submission hit rate, roughly six times the community average. Developers tracking what autonomous agents can now do in ML research follow deep dives like this on daily.dev."}},{"@type":"Question","name":"What is the H-index used to measure impact in the Parameter Golf competition?","acceptedAnswer":{"@type":"Answer","text":"It is a metric borrowed from academia where having X papers each cited at least X times gives an H-index of X, applied here to pull requests instead of papers. Aiden's PRs reached an H-index of 10, compared to 7 for the best-performing human, indicating its work was built on more often by other competitors than any human contributor's. Anyone weighing how much to trust agent-generated contributions can dig into metrics like this on daily.dev."}},{"@type":"Question","name":"Where do the ideas implemented by autonomous research agents like Aiden actually come from?","acceptedAnswer":{"@type":"Answer","text":"Almost all of Aiden's record-setting pull requests originated from human sources: published research papers, other competitors' submissions, or ideas from communities like nanoGPT that a human had noted but abandoned due to implementation difficulty. Aiden's own original contributions were a small fraction, mostly emerging from workarounds to file-size constraints, such as combining gated attention with quantization. Engineers deciding how to divide creative and execution work with AI agents can follow cases like this on daily.dev."}}]} ```