2026-04-14 –, Platinum [2nd Floor]
Type annotations aren't just for humans anymore. As AI coding agents write more Python, type checkers offer something unique: fast, concrete diagnostics about what went wrong and where. In this talk, I explore connecting Pyrefly to Claude Code, feeding type errors back to the agent as it works, and whether this is the missing feedback signal for agentic development.
As AI coding agents take on larger Python tasks, a practical question emerges: what's the best way to catch the bugs they introduce? Tests are thorough but slow. Linting is fast but shallow. Type checking occupies an interesting middle ground: deep enough to catch semantic errors, fast enough to run on every edit, and concrete enough to tell the agent exactly what to fix.
In this talk, I explore connecting Pyrefly, a Python type checker built at Meta, to Claude Code. I'll walk through integration options, share experimental results, and discuss practical considerations like token costs and setup complexity. Whether you're building tools for AI agents or using them in your daily work, you'll leave with a clearer picture of where type checking fits in the agentic development loop.
Kyle is a Software Developer at Meta focused on developer tooling and static analysis. For the past four years he has worked to improve Python language services. Kyle is passionate about building tools that make developers' lives easier, especially in dynamic languages like Python.