Training
Agentic coding for engineering teams
A two-day workshop on getting real value out of coding agents — for engineering teams that have already tried them but struggled to use them with their codebase, infrastructure setup and architecture standards.
- Audience
- Engineering teams (10–15 people), junior-to-mid
- Format
- Two-day workshop, on-site or remote
- Duration
- 2 days
Who this is for
Engineering teams that have already tried Cursor, Claude Code, Copilot Workspace, or similar but do not see the productivity gains that vendors promised. We worked with companies like Uber and others to help their software engineering teams adopt and use agentic coding and more importantly decide when to not use them!
What we cover
- The actual loop: how the two of us work with coding agents on production code, including the parts we still don’t trust them with.
- Repository setup: what makes a codebase agent-friendly, and the small set of changes that pay back the fastest.
- Workflow patterns: planning, parallelization, review, and when to abandon an agent-led thread and just write it yourself.
- Failure modes: where agents systematically go wrong, and how to catch it before it ships.
- Team dynamics: how senior engineers’ role changes, and the new shape of code review.
What we don’t cover
- Marketing claims. We’ve used these tools enough that we have opinions, and we’ll share them.
- “Which model is best.” This shifts every quarter. We’ll talk about the durable patterns instead.
- Tooling demos for the sake of it. We work in your codebase during the workshop, not on a sandbox project.
Format
Two days, on-site if you’re happy to facilitate logistics, remote otherwise. Limited to 12 people so we can actually engage with what each person is working on. We send a short questionnaire ahead of time so the second day is grounded in real things your team is shipping.