My Year with Claude: Building Midimeria - music production analytics

Foredragsholder

Om sesjonen

As a developer and hobby music producer I've spent the past year building a music production analytics platform alongside my day job as a consultant. An equal goal has been to find the limits of agentic development, and since Claude Code launched last spring it has been my main tool. This talk is about what I've learned and what my development process looks like now.

You probably have the skills for this already. Setting up custom AI workflows is surprisingly close to regular management and software development: clear direction, good delegation, verification of results. I'll walk you from a single prompt through composable skills to autonomous loops that fill a backlog, implement features, review code, and ship PRs - using examples from my own codebase.

We will arrive with an autonomous software development setup that can run with minimal intervention and show that the difficult task is now to set a clear direction. Agents can handle a lot, but you only get good results by guiding them, and knowing when to reach for a keyword like "red-team" to trigger a security review in two words.

Along the way: where agents are brilliant, where they stumble, and what a 1986 computer music textbook predicts about their limits.

Passer for

Anyone interested in hearing an experience report on practical Claude Code use

Samtidig i programmet

1. 11:40 Bootiful Spring Boot 4 Room 1
2. 11:40 Let’s create a tiny LLM library together Room 2
3. 11:40 Talk to me Java! Room 4
4. 11:40 What RSCs can do in Next.js today Room 5
5. 11:40 Sunset as a Service: Når målet er EOL Room 6
6. 11:40 The best way to fetch multi-level hierarchies from a RDBMS using Java Room 7
7. 11:40 The 1% Problem: An Introduction to AI Security Room 8
8. 11:55 Conways lov i praksis Room 6
9. 12:20 Metodikk som tvangstrøye Room 6