Stephen Downes

Here's the argument in a nutshell: "The pattern is clear: AI is collapsing specialist work, expanding what one person can do, and shifting the premium from niche expertise to end-to-end capability.The same pattern hasn't yet fully emerged in learning design - but there are signals to suggest that it is on its way." The instructional design function in particular seems headed for automation, but also a lot of the content-generation function, including in particular on-demand content generation, is headed that way. In my opinion, anyway.

OK, I haven't tried this yet, but it's in my list of things to do on my 'Stephen Follows Instructions' video series. "In this article, you will learn what Model Context Protocol (MCP) is and how to build a simple, practical task-tracker MCP server in Python using FastMCP." I can think of all kinds of things I want to do with an MCP server (and I'd like to do at least one of them before MCP servers are superseded by the next new technology).

Sure, you could use Microsoft Teams, if you enjoy complexity and pain. Or you could use Zoom, which works well but, as Ian O'Byrne says, is "a rented space where governance is centralized, and participants have no meaningful choice." Or you could use Jitsi. "Jitsi Meet is an open-source video conferencing platform. Think of it as the community-owned cousin of Zoom or Google Meet. It runs directly in the browser, requires no participant accounts, and supports encrypted video conferencing out of the box." Yes, you're hosting it yourself, but a service like "Reclaim Cloud offers a one-click Jitsi installer that handles most of the complexity while still allowing full control under the hood."

This is yet another RSS reader (YARR) that I'm mentioning here for a few reasons. One is the beautiful design, which shows what people can do these days. Another is the way feed data is stored locally (with an option to sync) which is how I built my own (comparatively very ugly reader). And third was the use of Turbopack, "an incremental bundler optimized for JavaScript and TypeScript, written in Rust, and built into Next.js", which should serve as a reminder of how much web infrastructure is available to people now that they can use AI without spending years (literally!) learning how to make it work.

I use a few bookmarklets I have written myself as part of my newsletter workflow. One sends items to Pinboard to read later. Another (which I am using as I type this) provides a direct way for me to write these posts right from the article I'm reading in my browser. This article is, as the title suggests, a guide to bookmarklets. This article discusses bookmarklets that change the style of the web page by altering the cascading style sheet (CSS). There are also some links to cool bookmarklets, like this set of web development bookmarklets.

I've mentioned this before but as things get increasingly fluid using 'students' as your survey population may become increasingly misleading. I think that's partially the case here. In addition to the generally positive remarks about AI (95% are using AI tools) we read, "students report that AI is enhancing their learning - rather than replacing traditional study methods." That's exactly what you would expect to hear from students, Properly So-Called. They are uniquely those people who have been able to access and succeed in higher education institutions. But what about the rest of the world? The report (30 page PDF) is behind a spamwall but you can access it without filling out the form here.