Why it matters
It's a condensed, bring-your-laptop version of a full SciPy workshop, covering prompt design, memory, and MCP tool integration end to end rather than just slides.
The tokenmaxxing angle
Building a multi-turn research agent surfaces exactly the cost tradeoffs tokenmaxxing readers care about: how much context and tool-calling overhead a working MCP-based agent burns per query.
From the organizers
Held at MIT Building E51 on July 9, led by organizer Ben Batorsky with setup instructions published at github.com/ericmjl/build-deep-research-agent; sponsored by NumFOCUS.