This is my blog. Sometimes in these posts I’ll talk about research or art projects I’m involved with, sometimes I’ll just ramble about other stuff which is keeping me up at night. If you’re interested in a certain topic, click on a tag to see just the posts with that tag:
aiatprotoccccomp4020cyberneticsdevextemporegigslenslivecodingllmmetapolicyreimagineresearchteachingweb
If anything here sparks your interest (or your ire!) then get in touch via email or discuss on HN.
- Extempore is alive on aarch64 (part deux)26 Feb 26
A quick update on Extempore on aarch64:...
- The road to COMP4020: climbing the "good" gradient24 Feb 26
- The road to COMP4020: the core mechanic20 Feb 26
- AT-URIs as persistent identifiers for scholarly blogging19 Feb 26
Every post on this blog now has a persistent identifier on the AT Protocol. If you scroll to the bottom of any post, you'll find a "Cite this post" section with a BibTeX entry that includes an AT-URI alongside the regular URL. I did this because I wanted a citation identifier for my blog posts that's more durable than a bare URL but doesn't require the institutional overhead of a DOI---and because the standard.site spec gave me a clean way to do ...
- Ben's dev setup 2026 edition18 Feb 26
New year, new dev setup. This year the theme is:...
- The road to COMP4020: providing sharp tools17 Feb 26
- LLMs Unplugged at Brimbank Tech School9 Feb 26
Sign up for the workshop...
- AI Agents on (in?) the Atmosphere6 Feb 26
I've been writing about agentic AI for a while now---from LLMs gaining the ability to act on the world, to what coding agents mean for how we think about code, to the power of new interfaces for agentic systems. But over the last few months something different has been brewing. People are building stateful agents---systems that don't just respond to prompts but persist, remember, reflect, and (at least in the eyes of their creators) grow over tim...
- Out of the browser: on the power of interfaces for agentic AI5 Feb 26
When OpenAI launched ChatGPT on November 30, 2022, it was a text box in a browser. That conversational call-and-response interface was a big part of what made it a hit---the fastest-growing technology product in history, reaching 100 million users in two months. But that's not what this post is about....
- Coding agents and the plasticity of code23 Dec 25
Code has always felt rigid to me. Brittle, even. Most of the ways you poke, prod, or change it will break it. And even when you know what you're doing and when you're making a change that's unambiguously a good one you often need to break it in one place and then manually fix it up in a dozen others to compensate....
- COMP4020/8020: Rapid Prototyping for the Web19 Dec 25
These are draft learning outcomes---subject to change as the course develops....
- Extempore's aliiiiive on aarch6417 Dec 25
If you're on the extempore mailing list you'll have seen the chatter over the last year or so about how development has (sortof) stalled due to the "core team" (@digego and @me, but calling it a "core team" makes it sound more grown-up) having other committments these days and not so much time to hack on it....
- Fixing slow Zed agent startup times17 Dec 25
I've been using Claude Code via ACP (Agent Control Protocol) in Zed, and while the integration is brilliant, starting a new Claude session took noticeably longer in Zed than running claude directly in the terminal; ~30s vs about 5s. And it was particularly frustrating because there's no /clear slash command via ACP---the only way to clear the context, which I d all the time is to re-start the agent, which (because of the delay) breaks my flow....
- Interactive neon perceptron visualisation11 Dec 25
I'm about to design and build a modern interpretation of Rosenblatt's Perceptron. It's called the Neon Perceptron becausever wire will be a flexible LED that'll light up with it's "activation". I'm doing the software, my colleague Brendan Traw is designing some custom PCBs and other hardware, and we're working on the overall design together....
- LLMs Unplugged: teaching language models with pen, paper, and dice10 Dec 25
The LLMs Unplugged site is at llmsunplugged.anu.edu.au. All materials are CC BY-NC-SA licensed for educational use. The code's on GitHub if you want to dig into the implementation details or contribute....
- Blog archaeology3 Dec 25
After the recent switch to VitePress, I became curious about how many of the previous iterations I still had access to. I've been an obsessive git packrat for years, and I figured that there'd be some secrets in the git repo. So I went spelunking in the history....
- Switching from Jekyll to VitePress2 Dec 25
This site has run on Jekyll since 2014 (or even before, I think; but some secrets are lost to time). Ten years is a good run for any technology choice, but I've finally made the switch to VitePress. The source is open, so you can check it out for yourself....
- 11ty and Vite for modern static websites24 Nov 25
I've been using 11ty for static site generation for a while now (not this site---that's Jekyll---but for other stuff) and it's great. But for a recent project I needed proper asset bundling, hot module replacement (HMR), and access to the npm ecosystem without building my own pipeline. Enter Vite....
- Seeing AI tasks through a TAM lens12 Nov 25
When it comes to AI adoption research, we keep asking "do you use ChatGPT?" when we should be adding "for which specific tasks?"...
- Tracking APS AI Transparency Statements11 Nov 25
New Cybernetic Studio project announcement: the APS AI Transparency Tracker (GitHub link)....