Small AI Projects You Build, Run, and Actually Own

What it does Fifteen hands-on AI projects across three volumes, each scoped for a single weekend. Every build ships with starter code, a README, and the exact prompts the author uses. Recent volumes shift from workflow automation toward personal, local-first tools like offline voice journals and lease analyzers.

The interesting bit The curriculum treats the Anthropic API as a component, not a platform. Volume three targets “life stuff” you run yourself—bank CSV parsing, decision simulators, voice transcription—assuming you want to own the tool, not just rent the output.

Key highlights

• Fifteen projects sorted by difficulty, from one-hour diagram generators to full-day coding agents
• Each folder has starter code, a README, and the exact prompts
• Volume 3 emphasizes local-first utilities for documents, money, voice, and decisions
• Domains include RAG assistants, multi-agent crews, screenshot-to-code, and inbox triage
• Requires Python or Node.js, an Anthropic API key, and terminal comfort

Caveats

• Every project requires an Anthropic API key, tethering your build to one vendor
• “Local-first” is explicitly “wherever possible,” not absolute
• Code is starter-grade, not production infrastructure

Verdict Good for developers who learn by shipping and want structure without hand-holding. Skip it if you want a fully managed service or avoid the terminal.