Stephen Downes

This article doesn't actually offer 10 simple rules - that's a different document - but it does lay out four themes based on the ten rules (and some personal experience). The themes are reasonable - you still need to know what you are doing, you need to know how to work with coding assistants for things like context management, you need to work within a testing framework, and you need to ensure the code is valid - that it actually does what the AI says it does. The article made me wonder: how would the same rules apply for using AI to write articles (assuming we need more articles in the world, a proposition I am beginning to doubt)? 

A couple of weeks ago Emil Stenström wrote How I wrote JustHTML using coding agents. I read it at the time - I thought maybe I had posted it here, but I guess I hadn't (it's one of those niche posts that really interesting to me but maybe less interesting to the broader e-learning readership). It describes using AI to write a fully compliant HTML5 parser in Python (not trivial, because there are so many ways to write HTML incorrectly, and a parser can't choke on them). It was significant to me because it suggests that testing, rather than reading lines of code, will be how we validate software in the future. Anyhow, in this article Simon Willison describes porting the software from Python to Javascript - "It took two initial prompts and a few tiny follow-ups... Time elapsed from project idea to finished library: about 4 hours, during which I also bought and decorated a Christmas tree with family and watched the latest Knives Out movie." So is that what software development is now? "Is it responsible and appropriate to churn out a direct port of a library like this in a few hours while watching a movie? What would it take for code built like this to be trusted in production?" Here's the playground for the new software. Works perfectly.

I was reviewing something I had written the other day when I realized I had used the word 'delve'. What did this mean? Had I succumbed to parroting the AI style? Wasn't I afraid of being labeled an AI user? Neither: it was just the word that seemed to fit. I left it in. So anyhow, this article wrestles with a similar dilemma for academic writers. What happens to creative thought when AI is everywhere? "Users become both architects and artefacts." I had a few responses to the paper: one, a visceral dislike for the writing style; another, wherein I commented on LinkedIn, "Makes you pine for the days when only the rich could afford intellectual dishonesty and unearned advantage." Another, where I considered the possibility that there was a genuine issue being raised that addressed both human and AI content equally. In the end, all I was left with was 'meh'.

The most recent bit of AI jargon is something that might take you back to the days of the Amazon Echo and Alexa: skills. Today's version, however, comes from Anthropic and is being (quietly) adopted by OpenAI. "A skill is just a folder with a Markdown file and some optional extra resources and scripts, so any LLM tool with the ability to navigate and read from a filesystem should be capable of using them." Here's where they show up in OpenAI. "Skills are a keeper," says Simon Willison.  

I didn't have much money when I was young so for me reading meant buying used mass market paperbacks. These were at best temporary books, and would often fall apart. But I wasn't building a library to decorate a house! Anyhow, with the arrival of the internet I was able to read as much as I wanted essentially for free, and I stopped buying them (I remember the point in time exactly - I had read the first two Game of Thrones books and was waiting for the third to come out in paperback... and it never did (maybe it did eventually, but I waited years). No way I was going to pay for a hardcover, so that was the end for me.) Anyhow, Ben Werdmuller writes, "ReaderLink, the largest book distributor in America, is going to stop carrying them - and we'll notice the effect immediately."

Alan Levine makes a point of offering a Hat Tip for the OER Foundation. He writes, "The courses shared through OERu are built on the web first approaches of collaborative authoring via a wiki and a dynamic distribution system based on open protocols. This is a credit to the vision and commitment of Wayne Mackintosh over 16 years. And so much was built and share in not only the tools, but the methods, but technologist Dave Lane, who always shared his expertise with me for years and years." I do hope someone in the world can step forward to host the content and services.