Orch Lite Rebuild

Complete Research Library — April 2026

6
Documents
6
Research Pillars
148
Sources Cited
Research Pillars
RESEARCH
Heavy User Configs & Real-World Patterns
Dotfiles & settings.json from named 2026 power users — what they actually run.
  • Steve Yegge / Sourcegraph config
  • Geoffrey Huntley dotfiles
  • Harper Reed + Simon Willison setups
  • Mitchell Hashimoto workflow
  • Hooks actually in settings.json
  • Slash commands that earn rent
  • Worktree + parallel session patterns
  • Claude Code Discord power configs
  • 32 sources cited
 RESEARCH
Autonomous Build-Test-Ship Loop
CC automation primitives + real autonomous loop implementations in 2026.
  • PostToolUse hook auto-test pattern
  • CronCreate scheduled agents
  • /loop self-pacing primitive
  • --print headless mode usage
  • Aider vs Cursor vs Devin loops
  • Mutation + property-based testing
  • Security scanning in agent loops
  • Real public loop configurations
  • 29 sources cited
 RESEARCH
UI Iteration & Visual Feedback Loop
Agentation + Playwright MCP + Chrome DevTools: sub-30s UI fix cycle.
  • Agentation MCP tools inventory
  • Agentation install + wiring
  • Playwright MCP fix-verify loop
  • Chrome DevTools MCP vs Playwright
  • Visual diff in agent loops
  • End-to-end UI bug pipeline
  • Real configs from practitioners
  • Sub-30s feedback loop patterns
  • 38 sources cited
 RESEARCH
Context Continuity & Instant Resume
Sub-3s full-context boot: CLAUDE.md, skills, batons, memory MCPs compared.
  • CLAUDE.md size vs. boot cost
  • Skills: rent vs. dead weight
  • Baton / handoff protocol patterns
  • MEMORY.md index approach
  • --resume vs /clear tradeoffs
  • Vector-store memory MCPs
  • Sub-3s startup benchmarks
  • Embedding retrieval vs. flat files
  • 16 sources cited
 RESEARCH
Multi-CLI Parallelism & Conflict Prevention
4+ parallel CLIs: worktrees, conflict detection, merge gates that work.
  • Worktree-per-CLI vs branch strategy
  • Concurrent edit detection
  • SQLite vs GitHub Issues tracker
  • Lock-free coordination patterns
  • Pre-push hook configurations
  • Merge-gate in production
  • 4+ parallel CLI real setups
  • File-level conflict prevention
  • 15 sources cited
 RESEARCH
Claude Max Token & Cost Optimization
Claude Max caching, model selection, manifest hygiene: real benchmarks.
  • Prompt caching hit rate benchmarks
  • Opus vs Sonnet vs Haiku ROI
  • MCP manifest size vs cache cost
  • Always-on vs on-demand MCPs
  • Subagent dispatch ROI rules
  • Claude Max rate limit patterns
  • Fast mode quality tradeoffs
  • Weekly cap burst behavior
  • 18 sources cited