Blogs
Opinion pieces, notes and long-form arguments.
-
From Stage Gates to Belief Flow: Continuous Decisions and Dynamic Pricing for R&D in the Age of AI
Biopharma R&D has long made its decisions at a handful of stage gates. This piece proposes belief-flow pricing: describing a project's state by its posterior beliefs as well as its stage, so that v...
-
After Seven Hundred Manuscripts: A Meta-Analysis of OpenAI's Mathematical Results, and Seventy Years of Machine Proof
On 6 October 2026, OpenAI released in one go 722 mathematical manuscripts in 372 result families, generated by an internal model. This piece sets out what the release actually contains, works throu...
-
Is Consciousness a Product of Reality, or Its Ground? From Laszlo's Akashic Field to Glattfelder's Information Ontology
The systems philosopher Ervin Laszlo and the complex-systems scholar James Glattfelder both reject physicalism and place consciousness at a more fundamental level of reality. This piece traces thei...
-
Does Biology Have Its Own Conservation Laws? A Test from the Central Dogma to Metabolic Scaling
The central dogma describes the mechanism of information flow; it doesn't draw the boundary of the states life can reach. Biology still lacks universal constraints of the conservation-law kind, met...
-
Of Every $100 of AI Capex, Whose Profit Does It Become?
An estimate chart cited by a16z Growth splits $100 of AI capex across the layers of the supply chain. Checked against public filings, the spending shares are broadly credible, but profit is highly ...
-
Once Lab Automation Turns Craft into Data, Who Is Still Watching for Anomalies?
Automation encodes the operating knowledge that can be written into a protocol, but struggles to encode the judgment that "this time is different". As reproducibility improves, the chances to disco...
-
Data Is Not Oil (a four-part series)
For nearly twenty years, "data is the new oil" has led people to think the data business is about extraction and sale. The oil industry took half a century to learn that a resource owner's way forw...
-
From Calling Tools to Operating Instruments: A Review of Anthropic's Model Hardware Standard (MHS)
The Model Hardware Standard (MHS), released by Anthropic as a research preview on 27 August 2026, is a shared specification that lets AI agents discover, understand and safely operate laboratory an...
-
The Line-Shaft Effect in the Age of AI (a four-part series)
The electric motor replaced the steam engine, but what changed the factory was that power no longer had to be distributed from one centre. AI is replacing a great deal of cognitive labour; its bigg...
-
Will AI Repeat Electricity's Forty Years? What Three Industrial Revolutions Tell Us About Diffusion
Steam waited sixty years, the electric motor forty, the computer twenty. Going back over the accounts of three industrial revolutions: where the lag comes from, what determines it, and where AI sit...
-
Causal-Gene Certainty, Not Statistical Significance: Decomposing the Value of Genetic Evidence
Splitting the composite claim that genetic evidence doubles drug success into two stages — getting started and getting through. The effect sits almost entirely in the first; what does the work is c...
-
Age at Onset and Time to Progression in the Leading Causes of Death: A Region-Stratified Synthesis
Age at onset and disease-progression timelines for the leading causes of death, stratified by WHO region and income group — and how much of the regional difference is real biology versus a differen...
-
A New Start in 2025
A poem for the year ahead.
-
Reading: The Power of Trees
Notes from a small book about trees, and about Pando — arguably the largest living thing on Earth.
-
A Few Flags for 2023
Six challenging targets for the year, and the paper of the day.
-
A Short Review of 2022
Work, organisation, hobbies: what a year of staying in one city looked like.
-
Revive!
Try to make use of github to host again the personal website. It looks like Jekyll is by default supported.
-
What AI Taught Me About Management
Objective, model, search strategy — the three things you fix before a machine-learning project, applied to work and to life.
-
Made By It, Undone By It?
Written the day after the gene-edited babies were announced: on democratised technology, outliers, and where the line between good and harm actually sits.
-
On Evolution, and the Evolution of Intelligence
From the physics of the origin of biological function to a set of open questions about simulating the evolution of intelligence itself.
-
Winning, Losing, and the Original Mind
Written while watching a game of Go between a human and a machine, on what it means that the board has not changed at all.
-
Where It Gets Hardest
The hardest part of building a company, for me, is restraining my own ideas: choosing a direction, and then building the team around it.
-
If You Came to Interview With Us
Notes from the hiring side of the table: what makes a bad impression, and what to prepare instead.
-
Character Before Craft
Two things I keep coming back to while building a company: hold your possessiveness loosely, and know where your line is.
-
On Rational Deliberation
An opening note on why I would rather we argue from logic than from authority.