How we teach you
to build your
own agent.

All three role paths are hands-on and tailored to your starting point. If you are an executive or process owner excited by tools like Claude Code, bring a business workflow you know well. No deep programming background is required—our instructor coaches you step by step as you turn your business rules and exceptions into tested code. If you are a contractor or consultant, you build on your Python and command-line experience to add stage rules (state machines), tool checks, diagrams, and an observer part for client workflows. If you are a project or product owner, you can start from a new idea or an existing project, and we match the coding support to your programming background. Before Day 1, we prepare your setup, select exercises for your background, and agree on one focused part of your agent to build.

About 10 to 14 days before the week, we have a short planning call and a 60–90 minute pre-week setup session outside the 40 training hours. You can train on your own Windows, Mac, or Linux computer. Together we set up your Linux coding workspace—directly on Linux, inside Windows using WSL2, in a Linux virtual machine on a Mac, or on a private remote Linux computer—and we check your AI model and coding assistant connections using access keys you enter yourself.

Our University Koans training materials, exercises, and course code are open source, so you can read, change, and keep our course code, and you keep everything you build for your own work. Hosted AI models (such as OpenAI, Claude, Gemini, Grok, and DeepSeek) and coding assistants run through their own providers under your account and are separate from our open-source course materials.

Yes. We learn the core method together on the admissions practice example using sample records, and in the later part of the week your instructor tailors the examples to a focused part of the agent or workflow you choose to make. To protect privacy, please keep real customer, employee, client, and private company records out of the training workspace. Connecting directly to live company HR, finance, or customer systems requires a separate engineering project outside this training week.

No. Bitlemmas Agentic Systems Week is professional training focused on building and testing one clear part of an agent. Passing classroom checks and repeated AI prompt tests shows how your practice system behaves on those tests. Those tests cannot guarantee that a live AI model will never make a mistake. Passing classroom checks also does not make a prototype ready for unattended production use (running live with real people and company data). You leave with a plain-English checklist of what real-world rollout still needs, such as saving data permanently, handling outside service errors safely, and completing a security review.

You train one-on-one with our instructor for 40 hours across five days, either in person in Austin or remote (ask us if you want to arrange a custom team week). Our first 40-hour one-on-one pilot with a project owner began in November 2026. Outside the 40 training hours, every learner receives a 60–90 minute pre-week setup session before Day 1. About two weeks after the final day, you join one 60-minute group clinic (a live group help session). At 30 days, you also complete one repeat check: you make one small change on your own, send in your saved check results and a short explanation, and receive one written review. Follow-up does not include ongoing support for live company systems. Talk with us to confirm dates and written pricing, or request the detailed curriculum below.