Talks from the team.
Three talks at AI Engineer on directing agents with written requirements, checking their work, and making them reliable. “Contract” was the workshop’s word for the written requirements an agent has to meet.
Your Agent Doesn’t Need an LLM for Every Decision
Raymond Weitekamp
Many of the choices an agent makes are small ones. Raymond shows how fast decision models, called recursively, can make them, and what that changes for coding agents, retrieval and computer use.
Stop Telling Agents How
Dan Barrett, Raymond Weitekamp and Jose Montes de Oca
Describe the outcome you need and let agents work out the steps. The workshop writes multi-agent workflows as Markdown contracts, checks every run against them, and ends with a bank’s model documentation.
Recursive Coding Agents
Raymond Weitekamp
Reliable coding agents need more than a capable model. Raymond covers how agents can break work down, keep state, verify what they delegate and reuse a workflow that succeeded.
Bring us the work you know deeply.
The first conversation is about your work: one representative workflow, what a useful result would be, and whether a paid pilot is worth doing.