Stephen Downes

I just want to say, I absolutely love the drawings in Jon Dron's presentation, Entangled Cognitions and AI: a Complexivist Perspective on AI and Learning, "an attempt to situate generative AI in the context of complexivist approaches to learning such as Connectivism, Rhizomatic Learning, Landscapes of Practice, Entanglement, Enactivism, and many other theoretical models that treat learning as a complex phenomenon." I'm not sure I buy into the new terminology of 'complexitivism', which to me looks an awful lot like connectivism, but that doesn't impact the value of this presentation, which I heartily recommend. And in fairness, I may be missing some nuance, since I don't see a recording or transcript (why people don't record their presentations remains a mystery to me)..

It's worth knowing that (a) some of the most important applications that hold the internet together are run by a small number of people doing it mostly in their spare time, and (b) government and industry have not found it worth their time or effort to provide any support at all to most of these people. Via D'Arcy Norman.

I think of these as design problems more than UX issues. Still, they're real and plague the fediverse even now. The list includes: instance selection issues, confusing timelines, difficulty following people, direct messaging issues, search, content discovery, and user discovery (the last three are variations on a single theme). Most of the solutions described in the post come in the form of third party apps: for example, "Services like tags.pub, relay.fedi.buzz, andtagpush.app already provide followable ActivityPub actors for each hashtag." Also, "Activity Intents just landed. Fediscovery is actually available to test." Will these solve the usability problems? I'm inclined to say no. This post is a follow-up to "Tim's Part 1 and Part 2, and Sean's solutions" and you may want to read them first, though it's not necessary.

This is a complex presentation of how "Léo Laporte runs AI at home and in the cloud: the machines in the attic, the models in the cloud, the agents, the voices, the memory, and what a year of it taught him. October 2026 edition" (Laporte is a tech journalist and host of TWIT). To figure out what's going on, your best bet is to scroll down to "Follow a thought through the house" and pick a scenario (such as "What's on my calendar?") The different stages will be listed on the right; click on them and watch as the components interact. If you want this at home, prepare to spend some money; the two DGX Sparks come in at roughly $10K per. 

In many parts of the world universities are mostly autonomous, even though they are largely funded by government. It can feel like a black box: put money in, and get what you hope are educated people and research output out the other end. This article looks at the concept of university autonomy in the UK, from its origins in universities as religious institutions, to its present-day role as a bargaining position and advocacy of the idea that the best people to determine how to teach and research are teachers and researchers. There's a point to that; my dental insurance thinks it knows more about dentistry than my dentist, which is how I got an $8 payment this week. The money quote is near the end, where author David Kernohan suggests that freedom of speech legislation "has, arguably, removed the ability of the academy to discard old, discredited, and harmful ideas with impunity."

Instructional continuity is "an institution's ability and commitment to sustain high-quality teaching and learning during and after a disruption." It's needed because "cyberattacks, technology outages, severe weather, public-health emergencies, political unrest, and infrastructure failures can cause disruptions." These are often hard to predict, and the biggest challenge is maintaining institutional commitment in the period between disruptions. Personally, I would have emphasized 'learning continuity', because that's what's important, but that's not how these institutions roll.