Writing

Essays, research notes, and stories — including the Inkhaven 30-day run.

Linkpost: Sanity-Checking 'Incompressible Knowledge Probes'
Q&A
Attachment Theory Is Extremely Cool and Useful
From Output to Connection
Why Are Anglophone Countries Unhappier Than Their European Counterparts?
The Great Smoothing Out
Do Emergently Misaligned Models Believe What They Say?
Whatever I Was Doing in 2020, I Should Do Again (Continued)
Whatever I Was Doing in 2020, I Should Do Again
Giving AI a Childhood
Contributing to Technical Research in the AI Safety End Game
People That Inspire Me
Which Information Sources I Listen to the Most
Do Models Know They're Lying When Claiming Fake Identities?
You Can Just Build Things with Claude Code
A Case for Persona Robustness as a Research Area
Characterising the Views on Safety from Frontier AI Labs: OpenAI
Characterising the Views on Safety from Frontier AI Labs: Anthropic and DeepMind
Is the Adult in the Room with Us Right Now?
How to Make a Problem Disappear
How South Africa's Electricity Catastrophe Was (Mostly) Fixed
This Is Not Financial Advice
Revisiting GSM-Symbolic: Do 2026 Frontier Models Still Fail at Confounded Grade School Math?
Spend Money to Buy Opportunity
How Does an Agent with Multiple Goals Choose a Target?

How a maze-solving RL agent internally represents and switches between multiple sequential goals, discovering spatial gating through negative activations.

Spectator Sport

Watching a cybersecurity competition where the most promising minds have become spectators to their own tools.

The Garden (Complete)

Two angels debate whether paradise requires suffering. One builds elegant systems; the other asks why they can't do better.

The Garden (Part 1)

Two angels tasked with building paradise discover that creation demands uncomfortable trade-offs.

The Physics of Great Storytelling

How the best fiction earns its weight by letting the world do the work — using Hunter X Hunter and A Wise Man's Fear as examples.

How did the United States hand over its future to corporations?

Tracing how a single memo by a corporate lawyer led to the systematic dismantling of the policies that built America's middle class.

Whole Brain Emulation as an Anchor for AI Welfare
Why pursue conceptions of agency for AI safety
Vipassana Meditation and Active Inference: A Framework for Understanding Suffering and its Cessation
Lessons from my first 10 day Vipassana