Benchmarking Self-Organized AI Swarms in Steering Complex Systems
MAM-Bench tests whether decentralized groups of coordinating AI agents can intentionally steer an emergent system toward a specified outcome they only partially control.
5/10/15/20-Minute Walksheds for Chicago 'L' Stations
An interactive map built from 580 inbound Google walking isochrones, showing which parts of Chicago can reach any 'L' station within 5, 10, 15, or 20 minutes.
2024 U.S. Voters by Birth Year and Estimated Childhood Blood-Lead Exposure
A cohort-by-cohort view of reported 2024 voters, partitioned by estimated early-life blood-lead exposure.
/skill:interactive-understanding
A local Docling-based skill for giving terminal agents enough PDF context to read academic papers--text and page images.
Pi VS Code Context Bridge
I wanted terminal agents for the agent loop and VS Code for reading code, so I reverse engineered Claude Code's IDE bridge and pulled the useful bit into Pi.
Training a Local GLaDOS TTS Voice
I pulled GLaDOS voice lines from the Portal game files, reconciled transcripts against the Portal Wiki, hand-reviewed the messy cases, and trained a local TTS model.
STT Benchmark: Cohere Transcribe vs Gemma 4
Gemma 4 accepts raw audio as multimodal input now, so can it match a dedicated ASR model for transcription? Ran 50 LibriSpeech samples on an RTX 3090. Not even close.
Toucan Corne Keymap Reminder
Working through a keymap to a new Corne layout keyboard. First time using ZMK and doing layouts via firmware files in C
PhD Advisor Analyzer Agent Experiment
An experimental targeted deep research tool I built to help in my own PhD application process. I've rebuilt it to be more robust and open sourced it to share it with the community. The former hosted demo is now archived.
Resistance Cascade: An Agent-Based Model of Civil Resistance Dynamics
This article looks at the empirical data on protests in authoritarian countries as a function of regime type and information control, then constructs an agent based model to examine how the effect of uncertainty can help explain the differences in protest and resistance movements in these differing regime types.
Using AI
A quick jot of notes on how I use AI tools
ML Prediction of Agent-Based Model Outcomes: Schelling's Segregation Model with 3D Manifold Predictions
Thomas Schelling's segregation model demonstrates how individual preferences can lead to complex social patterns. Here I visualize the results of a grid search across two parameters as a 3D manifold with ML outcome predictions.
