PS DEV

Session Description

As agents produce changes faster than you can review them, your attention becomes a bottleneck. Over two days you’ll build a process that lifts you out of the details without taking you out of the loop. You’ll build a working Scala 3 program that executes a “run,” a roadmap worked one “step” at a time, each step passing through distinct phases:

• Write. A model will implement the next step from a written brief, usually divided into “milestones.”
• Review. Two other models, neither of them the writer, will each get their own git worktree. They’ll read the diffs, edit code, run the tests, and report findings in a format your program can consume.
• Adjudicate. Another model will rule on the findings: what gets fixed and what doesn’t.
• Re-review and re-adjudicate. Changes made in response will face the same scrutiny.
• Plan. The planner will update the roadmap and loop to the next step – adding work you discover only by building, and reordering what’s left.

A manager agent will decide when the run needs you. As the pipeline runs, agents will record questions for you along with provisional answers; the manager will decide which ones you most need to see, and pause the pipeline for any that are important enough. It will absorb the rest and tell you where the run stands. When a run finishes, you can submit the work as a pull request and take that through your usual process, including AI and human reviews and resolving conflicts if you run multiple pipelines in parallel.

In addition to your quality pipeline, you will build an optimization loop that asks an AI for a change, measures the result against an objective, and feeds that back so the next iteration can improve. Your loop’s task will be to find the best algorithm for a given job, and the best Scala collections to run it on.

You’ll write both programs in Scala, and in this workshop they’ll target it as well: you’ll pick a project from a list we’ll bring, and your pipeline will drive agents writing Scala. That pairing is the point. A pipeline could build software in any language – the reason to aim this one at Scala is the type system, which helps an agentic loop find problems sooner and iterate faster.

The workshop will run as a series of tasks that build toward a complete quality pipeline and the optimization loop. Each task will start with a short talk from one of us (Bill or Frank); then you’ll build that piece with AI, after which we’ll look at what was built and discuss what worked. You’ll build your own pipeline, but you won’t be working alone: we’ll circulate throughout the hands-on sessions to help you past the bumps, and we’ll encourage you to help and learn from each other as well. Our goal is to make the workshop a conversation, so that we can all learn from each other.

Who it’s for

Developers who already use AI coding agents and want more from them.

What you will leave with

• Working code. A quality pipeline and an optimization loop, in your own repo, to keep using and improving after the workshop.
• New insights into how best to build software in the era of AI coding agents.
• Experience using Scala 3 as the orchestration language for AI quality processes.

Prerequisites

Working knowledge of Scala, a laptop, and access to an AI coding agent such as Claude Code or Codex. Use your own subscription, or connect to an open-source model we’ll provide – contact us beforehand so we can set it up.