Skip to content

Series

Some posts form a natural sequence and are best read in order. Each series below is a self-contained arc — you can start anywhere, but starting at part one gives the full context.

  • Giving an AI Agent a Brain

    How I connected IronClaw (a persistent AI agent running on EC2) to Synthesis via MCP — and what broke in unexpected ways during debugging.

    Posts in this series
    1. Giving an AI Agent a Brain: Connecting IronClaw to Synthesis via MCP
    2. When Your AI Lies About Its Tool Calls: Debugging kimi-k2.5

    2 posts  ·  February 2026

    Start reading

  • The Four-Layer AI Stack

    The architecture behind a development environment that partly runs itself: execution, tools, knowledge, and intelligence as four distinct, composable layers.

    Posts in this series
    1. Your AI Has One Layer. It Needs Four.
    2. Four Layers: How I Built an AI Development Environment That Partly Runs Itself
    3. What a 10× Workday Actually Looks Like
    4. What It Looks Like from Inside the Stack

    4 posts  ·  February 2026

    Start reading

  • Knowledge Context Protocol

    From llms.txt (a table of contents) to KCP (a navigable map): how AI agents actually find and use knowledge, and why the gap between the two matters.

    Posts in this series
    1. Who Describes You to AI?
    2. Beyond llms.txt: AI Agents Need Maps, Not Tables of Contents
    3. KCP and MCP: One Protocol for Structure, One for Retrieval
    4. Add knowledge.yaml to Your Project in Five Minutes
    5. Who Let the Agent In? (RFC-0002: auth & delegation)
    6. What Happens When Your Agent Needs Knowledge From Five Teams? (RFC-0003: federation)
    7. How Do You Tell an Agent "This Data Cannot Leave the Building"? (RFC-0004: trust & compliance)
    8. The HTTP Status Code That Waited 30 Years for Autonomous Agents (RFC-0005: payment & rate limits)
    9. The Agent Read the Whole Spec. It Didn't Need To. (RFC-0006: context hints)

    9 posts  ·  February 2026

    Start reading

  • Aurora & Temporal Analytics

    Why asking why something happened requires a different kind of data infrastructure — and how Aurora approaches root cause analysis through temporal graphs.

    Posts in this series
    1. Rethinking Systems for AI
    2. Aurora: Answering Why
    3. Unlocking Temporal Graphs
    4. Alchemy + Aurora: Data to Action
    5. Temporal Analytics and Organisational Amnesia

    5 posts  ·  August – October 2025

    Start reading

  • Frøya: Digital Co-Workers

    Building a digital co-worker for skill library quality assurance — what it takes to give an AI agent a meaningful role alongside a human team.

    Posts in this series
    1. Frøya: A Digital Co-Worker
    2. Mapping Human Potential

    2 posts  ·  April – May 2025

    Start reading

  • Building lib-pcb

    The story of building a professional PCB design library in 11 days — what the methodology looked like from the inside, and what 197,831 lines of Java actually required.

    Posts in this series
    1. The Surprisingly Hard Problem of Semiconductor Part Numbers
    2. Building a PCB Library: A Weekend Experiment
    3. Months to Days
    4. Six Pillars: What We Learned Building 200,000 Lines in 11 Days
    5. Building Together: An 11-Day Human-AI Collaboration Story

    5 posts  ·  January – February 2026

    Start reading