Claude Code for Spring Developers
summary
Claims a 5x commit velocity after 3 months using Claude Code CLI — 3,700 commits, 1,700 sessions and 11,000 prompts — by combining Opus for heavy planning, Haiku for fast headless checks, and Sonnet for subagents. Workflow components are concrete: CLAUDE.MD and .claude/rules for project rules, skills like interview/TDD/commit that enforced test-first and clearer specs, hooks that block bad Git and forbidden imports, and MCP integration with IntelliJ. Demonstrations include a sub-agent doing 49 tool calls over 84,000 tokens and a deterministic runner called Ralph deterministic loop.
AI-written summary · about
chapters (31)
- 0:00Introduction and setup
- 1:36Understanding the agent loop
- 2:03Demo and interfaces
- 3:31Optimizing for Spring
- 5:50Model selection and effort
- 6:12Safety and experiment control
- 6:38Context and delegation
- 7:22Sub-agent research workflow
- 8:29Headless and automated usage
- 8:51The baseline for vibe coding
- 9:15Context management strategies
- 9:53The Claude.md onboarding doc
- 10:35Scoped rules and automemory
- 13:52Managing context budget
- 14:35Workflow skills
- 17:18Core development skills
- 18:51Spec-driven development
- 19:46Improving processes iteratively
- 20:28Finding expert JVM skills
- 21:20Enforcing behavior with hooks
- 21:55Advanced post-edit validation
- 22:34MCP for external tools
- 23:30Agent developer experience
- 24:45Levels of agentic engineering
- 25:18The Ralph autonomous loop
- 26:56Cross-verification strategies
- 27:24Tooling and workflow tips
- 27:45Git workspace management
- 28:12The shift in engineering roles
- 28:32Discipline and conclusion
- 29:38Q&A and closing
related talks
01h 06mHow I Tamed Claude
01h 00mHow I Tamed Claude
01h 06mHow Claude Code Works
18mMastering Claude Code for Data Professionals
26mMastering Claude Code: Beyond the Simple Code Assistant
24mClaude Plays Worth Stealing | Claude Code & AI Coding Tips
21mHow Anthropic Uses Claude
18mClaude Code & the evolution of agentic coding