All posts

Deep dives and essays, newest first. Pick a series tag to narrow the list.

  1. agy-staff Bench v1: Evaluating Agent Orchestration and Harnesses

    Zen of Harness22 min
  2. Blogging, Aesthetics, and Taste

    Scaling Thoughts11 min
  3. Harness Optimization in Practice: Agent Tool Optimization | Notes on Iterating agy-staff

    Zen of Harness8 min
  4. agy-staff: Making Gemini a Staff Member for Codex / Claude Code

    Zen of Harness6 min
  5. How to Think About the grill-me Skill?

    Zen of Harness9 min
  6. Why Is Codex Always Auto-Compacting?

    Zen of Harness18 min
  7. One Poem Suffices: Agent Skills

    One Poem Suffices50 min
  8. When AI Becomes a Way of Understanding the World: Notes on a Demis Hassabis Interview

    Scaling Thoughts9 min
  9. Distilling Claude Code Source into an Agent Skill — A Harness Engineering Practice

    Zen of Harness25 min
  10. When Do You Need Multi-Agent Systems?

    Thinking in Context20 min
  11. The Nuclear Proliferation of Seedance 2.0

    Scaling Thoughts4 min
  12. Context Engineering in Codex

    Thinking in Context18 min
  13. 2025 in Review: Wanting to Do Everything, Going Deep on Nothing

    Scaling Thoughts9 min
  14. One Poem Suffices: Just-in-Time Context

    One Poem Suffices35 min
  15. When Tokens Become a New Means of Production

    Scaling Thoughts11 min
  16. At $125 a Month, Can Productivity Scale with Token Usage?

    Scaling Thoughts2 min
  17. Best Practices I

    Zen of Harness10 min
  18. One Poem Suffices: Context Engineering

    One Poem Suffices22 min
  19. One Poem Suffices: Multi-Agent Systems

    One Poem Suffices45 min
  20. Treating an LLM as Your Personal Secretary

    Scaling Thoughts3 min
  21. Introduction and Problem Breakdown

    Zen of Harness3 min
  22. One Poem Suffices: MCP (Model Context Protocol)

    One Poem Suffices50 min
  23. Full-Stack Development in the GenAI Era

    Scaling Thoughts2 min
  24. One Poem Suffices: Prompt Caching

    One Poem Suffices55 min
  25. When Coding Becomes a Basic Skill

    Scaling Thoughts3 min
  26. Renting Out Your Time and Working for Yourself

    Scaling Thoughts10 min
  27. One Poem Suffices: From Workflows to Agents — Reading Anthropic’s Best Practices

    One Poem Suffices40 min
  28. One Poem Suffices: Mixture-of-Depths and Dynamic Compute Allocation in Transformers

    One Poem Suffices38 min
  29. One Poem Suffices: Basic LLM Inference and Generation

    One Poem Suffices30 min
  30. One Poem Suffices: ZeRO (Zero Redundancy Optimizer)

    One Poem Suffices34 min