Parslow.net

Notes, articles, papers and teaching resources on artificial intelligence, education and software engineering — all topics of my own choosing, most of them explored in collaboration with AI.

One strand of the site is PatLang — a programming language I designed, but which was coded, tested, and largely documented entirely by AI, with none of the implementation hand-written. I strongly suspect this leaves real security gaps, not least the ability to run arbitrary code generated at runtime.

The rest of the site — the articles, teaching material, and topic pages that make up most of it — comes from a different process, and a genuinely collaborative one. The ideas are mine: which questions are worth asking, which connections between fields are worth chasing, and which direction an argument should take when that isn't obvious yet. The AI's role is closer to a research assistant's than an author's: finding the literature behind a claim, checking that a cited paper actually says what it's being used to support, and pushing back with evidence when an idea doesn't hold up against what's been published.

Working this way has changed what I know, not just what I produce. I regularly end up reading research well outside my own specialism, following a citation trail I wouldn't otherwise have had the time or background to pursue, and coming back with a link between two fields that neither literature obviously anticipated on its own. Cross-disciplinary reading that used to be slow and tiring to assemble by hand now happens fast enough that following a hunch across fields is something I actually do, rather than something I mean to get around to.