Series

Most of what I write here belongs to a longer argument. These are the sequences — each one read best in order, though every part stands on its own.

Disciplining the Coding Agent

4 posts

When a model writes the code you cannot supervise the lines, so you constrain the environment instead: a spec as contract, an adoption gate with the power to refuse, acceptance tests frozen and signed before any implementation existed, and finally an agent paid to break all of it.

  1. Keeping the Coding Agent on the Straight and Narrow
  2. The Gate That Said No
  3. Signed Before the Code Existed
  4. Fifteen Ways Past My Own Gate

Credence

7 posts

A Bayesian decision-theoretic framework for LLM agent tool selection: what it claimed, how it was measured, what shipped, and the one-line ablation that eventually beat it. Seven posts, ending in the retrospective that retracts the headline.

  1. Agentic AI Is Neither Intelligent Nor an Agent
  2. How Decision Theory Cuts Your AI Agent's API Bill in Half
  3. Sixty-Two Percent Correct and Winning by a Hundred and Twenty Points
  4. The Prompting Gradient
  5. The Agent That Invents Its Own Rules
  6. Ninety-Six Percent Cheaper and Slightly Better
  7. The Ablation That Beat the Agent

proplang

5 posts

A small language in which a Bayesian decision-theoretic agent is a program, cornered so tightly that the architecture is enforced by the grammar rather than by good intentions. Five posts on what it refuses to let you say, and why each refusal is load-bearing.

  1. What You Cannot Say, You Cannot Get Wrong
  2. The Alphabet Is the Prior
  3. There Is No Forgetting
  4. Think More, or Act Now
  5. The Agent That Prefers to Be Wrong

The Tool-Call Governor

4 posts

credence-pi: a Bayesian governor that sits between a coding agent and its tools, blocks the calls it wastes, asks before an injected action, and routes each turn to the cheapest model that can handle it. Four posts, from architecture to shipped plugin.

  1. The Brain is Opaque to the Body
  2. What a Regex Can't Do
  3. Make Your OpenClaw Cheaper and Harder to Fool
  4. Make Your OpenClaw Agent Cheaper, and Measure It Yourself

The Velotix Files

5 posts

A five-part investigation into Velotix Ltd, an Israeli cybersecurity startup that marketed angel investments through an ISA-licensed crowdfunding platform while under government enforcement for not paying its employees — and what happened after Part 1 was published.

  1. Part 1: The Investigation
  2. Part 2: The Serial Entrepreneur
  3. Part 3: The Marketing Machine
  4. Part 4: The Takedown
  5. Part 5: The Comeback

Agents from First Principles

5 posts

Five posts building a Bayesian learning agent from nothing: a creature that learns what to eat, an evolutionary sequel, a text adventure played through a value-of-information gate, the state-representation fix behind it, and the three types the whole thing reduces to.

  1. Building a Bayesian Learning Agent That Teaches Itself to Eat
  2. Evolution Discovers How to Think: A Philosophical Journey in Code
  3. Teaching Zork to a Bayesian
  4. The Loop Problem
  5. Three Types and a Funeral for Your Inference Library