I’m a Bayesian statistician who works out whether an AI system can be trusted: how to evaluate an agent that no human is watching, how to score every suggestion a model makes, and how to govern a system that is allowed to rewrite itself. That work is jack, a court for a self-modifying agent, and the book about it, A Court for Agents, in which every number can be recomputed from the court’s public record. My PhD on Chain Event Graphs (Warwick, 2010, supervised by Jim Smith) established methods for Bayesian structure learning that have been cited over 1,750 times.
In industry I design the evaluation that decides whether an LLM system ships. At Booking.com the measurement redesigned the system: task completion up 2.7×, latency down an order of magnitude. At HSBC I design and measure LLM workflows for vendor compliance and fraud. When a one-line ablation beat my own agent, I said so.
Oxford mathematics. Warwick PhD. Hong Kong, by way of Tel Aviv; a UK and Israeli citizen.
Open to research and senior scientist roles in AI evaluation, oversight and assurance, and to contract and consulting work. Hong Kong, Singapore, London, Tel Aviv or remote.
If you need to know whether an AI system works, and to be able to show it — let's talk. A two-page CV is at /cv/.
What I Do
Evaluating AI systems
Evaluation that is fixed before the run and can tell a real regression from an artefact: LLM-as-judge with its variance decomposed, ablations, power analysis, proper scoring rules, and criteria signed before the code exists. The method is in Signed Before the Code Existed; the ship decision it produced at Booking.com is on the CV.
Governing self-modifying agents
A fixed core the agent cannot reach, a free edge it may rewrite, and a score that is rigid everywhere. jack is the court that runs it; A Court for Agents is the book that derives it from results that were mostly settled decades ago. The founding argument is The Science Was Done. The Engineering Wasn't.
Bayesian decision theory
Agents whose beliefs are probabilities and whose choices maximise expected utility, with no other machinery. proplang makes the architecture unsayable to break; wald is the exact decision kernel that audits production choices.
Hong Kong data
Renavon collects and serves the territory's public data and preserves Webb-site's 35-year financial database. doorplate.dev resolves any Hong Kong address, in English or Chinese, to the government gazetteer with a Bayesian posterior and an honest refusal.
The book
A Court for Agents Law, scoring and the self-modifying machine
The monograph of jack, a court for a self-modifying agent, written as a construction: each chapter builds the smallest thing that answers what the last one left open, and every number carries the entry of the record it came from. Written with two Claude sessions under the court's own rules, every figure checked by script against jack's public record. A first draft, open to read while the layers cross.
First draft, November 2026. Next: chapters 6, 7 and 13 are completed from the C2a and C2b window exits; C2a ends on 15 November at the latest. The first edition is cut at C3's exit, when the agent arrives; it scores chapter 15's bets. Send a correction · Follow revisions
Read the draft →Start here
Five essays that carry the argument. Not the newest — the ones worth reading first.
Why We Stopped Using the Mathematics That Works
Start here literally. The mathematics for making decisions under uncertainty has existed since the 1960s and the field walked away from it — not after weighing it against something better, but because ImageNet made a different bet pay off and nobody looked back.
The Science Was Done. The Engineering Wasn't.
The argument. An agent allowed to change itself needs a core it cannot reach, an edge it may rewrite, and a score that is rigid everywhere — and nearly every piece of that was proved decades ago. The founding note for jack and the book.
Signed Before the Code Existed
The method. The acceptance tests were written, hashed and signed before a line of the implementation existed, then handed to a coding agent that could not change them. Pre-registration, made binding.
The Ablation That Beat the Agent
The integrity. A one-line ablation beat the agent I had announced a paper about, so the paper never went out and the record of why is here instead. What survived is narrower and better than the claim.
The Court That Tried Itself
The cost. The predecessor court reached its seal and a measured verdict, and it also cost 121 of my repairs in four days. The book begins where this essay ends.
Series
- A Court for Agents 1 post
- Disciplining the Coding Agent 6 posts
- Credence 7 posts
- proplang 5 posts
- The Tool-Call Governor 4 posts
- Agents from First Principles 5 posts
Recent Posts
Projects
jack
renavondata/jack
A court for an agent that is allowed to change itself. At its centre is an append-only, hash-chained, signed record; a 450-line checker that accepts or refuses each entry; and a judge that runs the checker over every pull request and refuses what breaks the law. The parties are one person and several language-model sessions, and every suggestion any of them makes is a forecast scored by log-loss against a base rate, …
Kana Derivation Chart
An interactive map of how Japan's two phonetic alphabets grew out of kanji. All 46 basic syllables sit in a grid by consonant and vowel, colour-coded by script; click any cell to trace both descents …
proplang
A small language built on one claim: an adaptive Bayesian decision-theoretic agent needs no adaptation machinery. If hypotheses are programs and the prior is 2^-|program| under a grammar, then …
life-agent
The composition root for a personal life-management agent — an assistant that remembers your whole digital life, reasons under uncertainty, and acts across your tools. It reinvents nothing: every …
credence-pi
A Bayesian governance layer that decides which of a coding agent's tool calls to allow, block, or pause for human approval — and learns where the line sits from the agent's own behaviour. The design …
pkm
A content-addressed extraction cache and catalogue for personal knowledge management. It ingests documents, routes each through a format-appropriate extractor — pandoc, docling, unstructured — and …
docaviv+
A faster, phone-friendly way to browse the Tel Aviv International Documentary Film Festival than its official site. Bilingual Hebrew/English with full RTL, fuzzy search, a day-by-day schedule grid, …
accessinfo.hk
Hong Kong's Freedom of Information platform: citizens file requests to government bodies and track the replies in the open. Every exchange stays on the public record, accreting a searchable archive of …
Webb Data
An open archive of 35 years of Hong Kong corporate data — directors, board memberships, CCASS shareholdings, and company-registry filings — cross-linked so you can follow who sits on which boards and …
doorplate.dev
Resolve a Hong Kong address written any way a person might write it — English, Traditional or Simplified Chinese, mixed, abbreviated, misspelt, in any order — to the premises it names in the …