#claude
6 posts tagged with this
-
Opus Plans, Sonnet Builds, Haiku Documents. The Hard Part Is the Handoff.
Picking a model per sub-agent takes one line of config. Making the work survive the trip between them takes a handoff contract — here's how to design one.
Read more -
Claude Has Three Different Memories. Here's What Each One Actually Does.
claude.ai, the Claude API, and Claude Code each ship something called "memory" — and they're three unrelated systems. Here's what each one actually stores, and which one you're using right now.
Read more -
Caching While Coding with Claude Code — What's Actually Happening Under the Hood
I tracked 65 days, 1,142 sessions, and 1.83 billion cache read tokens. Here's what's actually happening every time you run Claude Code — and why you're probably already benefiting from it without knowing.
Read more -
How Claude Code Image Paste Actually Works
You see the word "image" appear in the terminal. Here's what's really happening underneath — from keystroke to API call.
Read more -
Claude Code Hooks: Auto-Label Every Terminal Session
Replace Claude Code's random session hashes with meaningful labels like S1 and S2, so notifications tell you exactly which terminal needs your attention.
Read more -
Claude Code Hooks: Get Notified When It Needs You
Stop babysitting the terminal. Claude Code hooks let you run scripts on task completion, permission prompts, and idle states — so you get notified the moment Claude needs you, no matter what you're doing.
Read more