Portfolio

Tools, toys, and experiments. Some you can play with; some you can pull apart.

Rama & Agent-o-rama

Exploring stateful agents and research tools with Rama and Agent-o-rama. My LoF Research Lab repository brings together paper-ingestion code and notes on running the stack.

Red Planet Labs' Agent-o-rama provides agent graphs, tracing, evaluation, and durable storage in Java or Clojure.

DistinctionBench

A Laws of Form toolkit for evaluating LLM reasoning. The expressions are tiny: adjacent marks ()() become (), while a nested pair (()) cancels to nothing. Can a model keep track when the brackets get noisy?

Explore the benchmark and examples.

House of Bao

A playable puzzle prototype where numbers are containers you can move and transform. The idea: make arithmetic feel like something you can see and touch, rather than a pile of symbols you have to take on faith.

Play House of Bao: a grid of visible quantities and containers

Play House of Bao · Code · Why iconic arithmetic?

Printing accessible playing cards

I wanted to play Netrunner with my legally blind girlfriend. I tested digital settings and oversized prints, ending up with cards around 89 × 123 mm: readable, but still shuffleable. The write-up includes settings, print sizes, and what worked for us.

Read the making-of and printing notes.

More experiments

  • Rama backend harness — my fork of Red Planet Labs' harness for testing LLM-built backends against implementation challenges, including correctness and fault-tolerance checks.
  • Math prototype — an earlier exploration of visible arithmetic. Rough edges included.

More writing, or email me to talk about a project.