I trained a sparse autoencoder on CLIP and found over 50,000 highly interpretable and steerable features. — June 2024
Research post on the effect of error correction in a maze solving agent — April 2024
Training a model to front run bad actors in crypto for fun and profit — January 2024
Guide to modelling algorithm performance on GPUs — December 2023
An on-the-go technique for writing notes using your voice, instead of on paper or using a device. — October 2023
I studied category theory for a week, and made some sketches of how learning assistants can be improved — October 2023
Experiments in mechanically finding activation vectors to influence how GPT2 responds — June 2023
Experimental user interface for AI feedback on system design — February 2023
Linen was an AI-powered journal for self-growth. — December 2022
Integral allowed users to ask questions about their network by synthesizing information across their email, Slack, and contacts management software. — August 2022
Interstitial logging app for tracking context switching — October 2019
A transpiled educational programming language with English-like syntax — January 2016