Research Intern, Agent Experience

The next power user on the web isn't human.

Developer Experience took twenty years to become a discipline with tooling, standards and budget behind it. Agent Experience won't get twenty years — the shift is happening now, and there's no accepted way to measure it yet. We're building the AX standard and helping teams ship agent-ready software.

We're making an ambitious effort to define the standard for AX — and that takes profound research to shape what AX actually means. Our public AX work is Agent Arena, a benchmark ranking developer tools by their agent readiness. This position is key to helping us shape it.

Experiment, measure, publish.

• Own benchmark experiments end to end — from running them across different coding agents and harnesses to publishing the methodology
• Ship the research: share your most important findings with our community of customers, partners and the broader developer ecosystem
• Challenge our thinking and shape our product with your findings — and if you're excited about building, we will gladly have you play a role in that too

You're excited about coding agents and where they excel and break.

• Want to own a conclusion and its implications, not just an analysis
• Want more than the research work — talking to the customers who use this, and challenging our methodology when it doesn't hold up

Nice to have: RL or post-training experience, work with MCP or tool-use APIs, anything you've published or open-sourced about how agents behave.

This is an internship position, but if you're looking for a full-time opportunity, we're happy to turn it into one at the end of the internship — or consider it for the right candidate from the start.

Small team, high bar.

• Make decisions fast, and revisit them when we're wrong
• Dive deep — accuracy and credibility are the product
• Challenge one another
• Stay curious and have fun

Send us your CV or GitHub, and one thing you'd change about Agent Arena and why.