Lunch & Learn: Claude Edition
Published by O'Reilly Media, Inc.
Turn your lunch break into an AI productivity boost
Bring your lunch and join us Monday through Friday for four weeks. Over 20 sessions with instructor Kesha Williams, you’ll build one thing: an AI code reviewer that runs in your CI pipeline, reads real pull requests, and posts inline comments at the exact file and line.
Day 1 is a single API call against a code diff. By day 20, that same logic runs as a GitHub Action, loading your team’s review standards, catching real bugs, with a measured false-positive rate and a known cost per PR. Miss a day, and you’re not behind: The companion repo gets a tagged checkpoint and a demo script after every session.
But we don’t just stop at the demo. Stick around immediately after the training for a 15-minute post-show Ask Me Anything. This is your chance to bring your specific workflows, get expert troubleshooting tips, and get your toughest technical questions answered in real time. Grab your lunch, build a new AI skill, and walk away with a technique you can apply to your work the very same day.
What you’ll learn and how you can apply it
- Prompting for structured output: Get findings as JSON your scripts can parse
- Tool use and MCP: Let Claude fetch the files it needs and read real pull requests
- Context engineering: Feed the right context and cache it to cut costs on large repos
- Skills and CLAUDE.md: Package your review standards so Claude applies them every run
- Claude Code in CI: Run it headless, scope its permissions, post inline PR comments
- Agent architecture and evaluation: Split generation from review, then measure precision and cost per PR
This live event is for you because...
- You write and review code for a living and want Claude doing the first pass on every pull request.
- You’ve used Claude in a chat window and want to know how it gets into a pipeline.
- You’re the architect who decides where agents belong, what they’re allowed to touch, and what they cost.
- You’ve seen automated review tools get ignored because nobody trusts the output, and you want to build one that doesn't.
Prerequisites
- Come with your questions
- Have a pen and paper handy to capture notes, insights, and inspiration
Course Setup
- Sessions are demo-first. You can attend all 20 with nothing installed. Everything shown live lands in the companion repo with a tagged checkpoint and a demo script after each session, so setup is for replaying the builds on your own time.
To run the demos yourself, you'll need:
- An Anthropic API key with a small spending cap set. One key covers the whole series, from the Day 1 API call to Claude Code in Week 3. The cap keeps your replay costs to a few dollars.
- A recent Python 3 install. All code is short, readable Python.
- A GitHub account and a personal access token. You'll use them from Week 2, when the reviewer starts reading real pull requests.
- Claude Code installed by Day 11. The native installer takes one command and has no dependencies. Instructions are in the repo's SETUP.md.
- jq on your path by Day 12, for pulling structured findings out of the JSON output.
- Optional for Week 4: fork the sample target repo to run the CI workflow against your own copy.
Can't install locally? The companion repo includes a devcontainer. Open it in GitHub Codespaces and replay any day's demo from your browser with zero setup.
Schedule
The time frames are only estimates and may vary according to how the class is progressing.
October 7, 2026 - November 3, 2026 9:00am PT | 12:00pm ET
- Interactive demo, discussion, and Q&A (45 minutes)
Your Host
Kesha Williams
Kesha Williams is an enterprise architect and AI consultant with over three decades of experience designing, building, and operating production software systems. She is the founder and managing partner of Keysoft, where she helps organizations safely adopt and operationalize AI, including autonomous agent systems. Recognized as an AWS AI Hero, Kesha focuses on helping engineering and security teams deploy AI safely in real-world environments by applying practical architectural and operational patterns. She brings a practitioner’s perspective shaped by hands-on experience working directly with teams deploying agent-based systems in production.
In addition to her consulting work, Kesha is a globally recognized AI educator and advisor. She has developed AI training programs and advised organizations on safely integrating AI into production systems, helping practitioners move beyond experimentation and toward reliable, operational deployment. Her work focuses on helping technical teams design, operate, and secure AI systems in real-world environments. Kesha is a frequent speaker at global technology conferences and is known for translating emerging AI capabilities into clear, actionable guidance for engineers and technical leaders.