Leads AI reinvention at PwC Australia. Research and writing on technical AI safety, agent interfaces, and the economics of automation.
Does a frontier model, prompted to answer as a prior one, actually reproduce th…
Apr 16, 2026Capability jumps between models create a side-channel of information that only…
Mar 14, 2026Everyone says AI is getting cheaper. The data says otherwise. Frontier model pr…
Feb 22, 2026If you put an AI agent in a community where everyone else is writing malware, d…
Feb 8, 2026






A game to build empathy for AI agents by experiencing their constraints firsthand
AI-driven workforce automation risk assessments using O*NET mappings and Claude task analysis
Open source implementation of Anthropic's Clio
A memorial to the benchmarks that defined—and were defeated by—AI progress.
Comprehensive evaluation framework for language models
Creating infinitely scalable verifiable test benches
Come explore Australian legislation!
AI-first regulatory compliance services
Explorations of SAE-based mechanistic interpretability for transformers
Research on document understanding using LayoutLM for relational extraction
Visual Transformer Mechanistic Analysis Tool
Reimplementing the evolution of generative models: vanilla NNs → AE → VAE → VQ-VAE → Diffusion
Flame charter and node level analyser across every query in an entire PG log output
Real-time collaborative cryptic crossword solving platform
Collection of landscape and urban photography projects
Physics and particle simulation framework
Robocup internationals - autonomous teaming soccer robots