Bootcamp

Alternates if you want more edge:

·       You Don’t Learn Piano by Watching: Two Days of Agentic Coding with Claude Code

·       From Empty Folder to Shipped: Agentic Coding with Claude Code

Description
Learning piano looks trivial. White key, sound. Black key, another sound. Congratulations, you are a pianist. Except you are not. Piano is practice, and coding agents are no different. Anyone can type a prompt and watch code appear. Doing it well, on a real system, under real constraints, is a skill you build by doing.

So we build. Over two days we create a complete TypeScript application from an empty folder: database layer, business logic, web UI, CLI, an MCP server, tests, and a CI/CD pipeline. All of it with Claude Code, live. No pre-baked repo we reveal at the end.

This is not a slide deck about AI. It is a working session. You will feel where agents shine, where they fall over, and what separates people who ship with them from people who fight them. Along the way we dig into the parts that actually matter in practice: prompting, skills, subagents, workflows, automation, sandboxing, and the review loops that keep an agent honest.

The frame for all of it: an agent is a model plus a harness. The model you rent. The harness you build. Two days of building that harness together. Move from ‘the agent wrote something’ to ‘I can steer an agent to ship production-quality code on demand.’ Build the mental model and the muscle memory so agentic coding stops being a lucky dip and becomes a repeatable engineering practice.

Content & Process
Day 1: Foundations and the core of the app**

·       Setup and mental model: Claude Code running, Agent = Model + Harness, how the tool-use loop actually works (read, edit, run), what the agent sees and what it does not

·       Prompting that holds up: planning before coding, being specific, feeding the right context, CLAUDE.md and project memory as the backbone of the harness

·       Start the build: scaffold the TypeScript project, database layer, core business logic, a CLI to drive it

·       Skills: teaching the agent your conventions and reusable procedures so you stop repeating yourself

·       Staying in control: reading diffs, course-correcting, knowing when to let it run and when to stop it

Day 2: Scaling up to a real system

·       We take everything from Day 1 and put it to work: a full day of hands-on building. We flesh out the app into a real system (web UI, MCP server, tests, docs, CI/CD, automation, CLI) and hit the deeper aspects of agentic coding as they come up in the actual work: subagents, workflows, sandboxing, and keeping the agent honest. Less structure, more building. The exact path depends on how far we get and what the room wants to dig into.

Audience & Requirements
Built for developers who already ship software and want to work with coding agents seriously, not as a party trick.

Audience:

·       Mid-level to senior software developers

·       Tech leads and architects

·       Platform and DevOps engineers

·       Engineers moving into applied, production-focused AI work

Required knowledge:

·       Solid software architecture experience

·       If you have not coded for a while, that is fine. We have an agent after all.

·       Comfortable in the terminal

·       At least basic knowledge of web technologies (HTTP, JavaScript, CSS, and so on)

Two ways to take part:

·       Watch and think along: I drive everything on my machine. You shape the prompts, throw in ideas, take notes, ask questions, build the mental model. No setup needed beyond the knowledge above.

·       Build alongside: you code on your own machine in parallel. For this you need Node.js LTS, the ability to install software on your machine (permissions), a Claude Max subscription, or API credit if you are on a smaller/enterprise plan.