Knowledge
Why a practice publishes a reading list
Most people selling AI to Australian businesses learned it from a vendor page eighteen months ago. There is no way to tell them apart from the outside. This is the alternative to asking you to take my word for it. The reasoning behind every system I build is here, with sources. Some of it shapes work I do for others. Some of it is a research interest and is labelled that way.
Shapes the work
Knowledge tools, methods, tools for thought, public brains, and the AI layer I actually deploy — including the ones I decided against.
Research interest — not sold
Computational neuroscience, topology, neuro-AI models, and a long-range track toward noninvasive brain-to-text. Included so nobody has to guess where this is heading.
Bush, Engelbart, Victor, Luhmann, Clark & Chalmers. This is why I don't believe in SOP documents nobody reads. Retrieval-based learning and distributed cognition are the reason my systems put the right thing in front of you at the moment of decision, rather than filing it where you'll never look.
Zettelkasten, MOCs, progressive summarisation. Practical structure for the part of a business that lives in someone's head — the part that's genuinely hard to hand over.
Eighteen tools evaluated on one question: when the vendor disappears or changes its pricing, do you still have your knowledge? That test decides what I install in a client's business.
The agent and LLM layer I actually deploy. Fourteen tools, tested — including the ones I decided against and why.
Prior art in publishing a body of knowledge instead of hoarding it. This page is built on what that reading taught me — the same pattern works for any business that wants its expertise to compound instead of walking out the door.
Thirty pieces of computational-neuroscience lab software — simulators, M/EEG pipelines, imaging. Research tooling. I don't sell this; it's how the signal-processing discipline got into my head.
The cliques-and-cavities lineage — including a note correcting the "brain works in 11 dimensions" headline that most people repeat wrongly. Pure research interest.
EEG/MEG/fMRI foundation models, and the critical review that explains why several of them are flattered by their own benchmarks. Reading this way — the sceptical paper alongside the announcement — is a habit that transfers directly to evaluating any AI vendor's claims.
Open questions and work packages toward noninvasive brain-to-text. This is the long-range ambition, stated openly so nobody has to guess where this is heading.
Atlas last revised 2026-07-30. Open the full annotated list →
Write with the situation as it actually is. I’ll tell you honestly whether I can help.