Claude Code at Scale: An Android Deep Dive
summary
Shows how Claude Code stays aligned with an Android monorepo by using repo-level skills and rules and a company plugin marketplace to enforce standards across dozens of repos, demonstrates KV cache reuse and a 1M token context window (4.8 opus) to avoid repeated work, explains spawn vs fork sub-agent strategies, and proves MCPs plus parallel sub-agents can run real CI tasks and code-review pipelines end-to-end with measurable speed and cost benefits.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 2:11Claude workflows and prompts
- 7:29Sub agents for parallel tasks
- 11:03Agent loop under the hood
- 14:55Tokenization basics
- 17:48Transformer circuits explained
- 21:46Context window explained
- 24:38Multi-agent strategies
- 28:14KV caching process
- 31:45Prompt caching boundaries
- 33:15Reasoning budgets
- 37:26MCP versus skill trade-offs
- 39:35Q&A
related talks
38mOne Android Dev, Ten AI Teammates: Scaling Delivery With Claude Code
44mClaude Code for Spring Developers
01h 06mHow Claude Code Works
01h 06mHow I Tamed Claude
01h 00mHow I Tamed Claude
18mClaude Code & the evolution of agentic coding
27mAgentic Coding with Discipline & Skill: Practical Techniques and Tips
37mScaling App development at Zalando