An AI Agent Became the #1 Contributor in OpenAI's Hiring Challenge — Zhengyao Jiang, Weco

Earlier this year, OpenAI ran Parameter Golf, a model-training competition that doubled as a hiring filter. Over 1,000 researchers competed to train the best small language model under a 16MB cap. The top contributor was the one candidate OpenAI couldn't hire. Our autonomous research agent Aiden finished with 7 merged records, more than twice as many as any other contributor, and ended up the most-cited participant in the community.
This talk is about what those 22 days showed. I'll cover on high level how does it works and which of its ideas produced the records. But the part worth more than the leaderboard is the collaboration itself, the community and AI agent building on each other's work, the largest natural experiment in human-AI collaboration I've seen run in public. I'll close with what it tells us about where humans and autonomous research each still matter for the foreseeable future.
1:57 PM

# An AI Agent Became the #1 Contributor in OpenAI's Hiring Challenge

**Location:** Main Stage
**When:** Day 3 - July 1, 2026 · 1:55pm-2:15pm

## Speakers

### Zhengyao Jiang
CEO & Cofounder · Weco AI
[X/Twitter](https://x.com/zhengyaojiang) · [LinkedIn](https://www.linkedin.com/in/zhengyao-jiang-387b44145/) · [Website](https://zhengyaojiang.github.io/)

Cofounder & CEO @WecoAI - automated hill climbing with LLMs. Previously: PhD in ML at UCL

Timestamps

0:00 Introduction to Parameter Golf and the Aiden agent
1:06 Defining the challenge: Auto-research vs. human community
1:47 About Weco AI and the development of Aiden
3:07 Evaluating Aiden's impact and H-index in the community
4:01 Why autonomous AI is powerful: Throughput and efficiency
5:21 Human-AI collaboration: How ideas move the frontier
6:32 Case study: Combining research, architecture, and tokenization
7:41 Summary of auto-research strengths: Execution and search
9:06 The role of human design in competition
10:04 The Andrej Karpathy metaphor: Gradient descent and coding
11:19 Auto-research as training a model: Evals and abstractions
13:36 Case study: Improving data pipelines via strict API abstractions
14:38 Conclusion: The new craft of the AI engineer