me

I'm an engineer-researcher at Anthropic, where I've moved between projects as demand dictates. So far that's meant building much of our early RL infrastructure, our sandboxing infrastructure, leading the multimodal effort and working on big-model training.

Outside of my day-to-day, I'm particularly interested in test-time compute, science-of-AI, interpretability, and the social impacts of AI - especially with regards to future-of-work, concentration-of-power, and the dynamics that might emerge when many people start deferring their decisions to their AI assistants.

I'm based in DC but regularly travel to SF and (as a Brit) semi-regularly to London.

The best way to get in touch with me is via email!

writing

anthropic

• Red-teaming Language Models to Reduce Harms: Some of the earliest results on automated red-teaming.
• Predictability and Surprise in Large Generative Models: Connects scaling laws (and lack thereof) with near-term impacts.
• Constitutional AI: Harmlessness from AI Feedback: Replacing the 'human' in 'human-feedback'.
• Training a Helpful and Harmless Assistant: The third worst way to align an AI, RL on human feedback.
• A Language Assistant as a Laboratory for Alignment: The two worst ways - prompting and fine-tuning - to align an AI.

indie

• Scaling Scaling Laws with Board Games: Studying big AIs on big problems using small AIs on small problems.
• Debugging Reinforcement Learning Systems: Debugging reinforcement learning implementations, without the agonizing pain.
• A Clearer Proof of the Policy Gradient Theorem: A policy-gradient proof with less algebraic manipulation.
• The Recursive Approximation Algorithm, Animated: How n-body problems are solved in linear time, without any maths.
• Post-Mortem Python Plotting: Fixing numerical bugs after they've caused a crash.
• AI Safety Needs Great Engineers: AI safety labs are bottlenecked on great engineering.
• Are we in an AI overhang?: AIs might get much larger very quickly.

projects

major

• megastep: A 1m FPS RL environment.
• pybbfmm: A GPU-accelerated black-box fast multipole method.

minor

• fayin: An attempt at using WebSpeech to help learn Mandarin pronunciation.
• coolgpus: Custom GPU fan control for headless Linux.
• kvbtests: Heteroskedasticity-and-autocorrelation-robust statistical tests.
• modulepickle: Makes distributed dev feel local by auto-shipping your working dir to workers.
• zonotable: Nails Zotero's paper ingestion onto the side of Notable.
• tablatex: Adds LaTeX symbol autocomplete to vscode.
• nosearch: Adds reverse search to Jupyter notebooks.
• noterminal: Adds disposable notebooks to Jupyter.
• stripcommon: Strips the common whitespace when copying into Jupyter.
• RL discord: One reinforcement learning Discord to rule them all.

old

These are from my first few years out of college.

• Commutes & Rent: An interactive visualization of commutes and rent in London.
• A Global Shallow Fluid Model: An interactive model of the atmosphere on a rotating planet.
• Segmenting Microarrays with Deep Neural Networks: c. 2014, automation of a labour intensive part of DNA microarrays.

website

This website is rendered from Markdown into a minimal template by a short Python script.