Stephen Downes

D'Arcy Norman collects his thoughts and experiences on using AI to quickly generate software prototypes and useful tools, a.k.a. vibecoding. "My career has somehow always been as a kind of translator between Nerds and Not-Nerds," he writes, "able to articulate things well enough that things make enough sense to everyone. Turns out, that’s kind of the vibecoding superpower." Vibecoding is an incredibly useful tool, he writes, but most definitely will not replace professional programmers. Today: Total: D'Arcy Norman, 2026/09/04 [Direct Link]

This paper (36 page PDF), which can be found via the University of Johannesburg portal, offers a useful discussion of digital sovereignty generally, with a particular focus on South Africa. The core of the paper is a framework describing degrees of sovereignty at the national, institutional, and individual level, each describing business models along with ontological, epistemological and sovereignty implications. This part of the paper is pretty clear; the section focused on South Africa is more of a slog, though there is a useful discussion of Nancy Fraser's model of justice: "Fraser's starting point is that 'the most general meaning of justice is parity of participation', by which she means that justice requires social arrangements that allow all to participate as peers in social life. She then identifies three mutually entangled dimensions through which participatory parity can be blocked: distribution, recognition, and representation." With respect to representation, "affirmative responses seek voice within existing governance," while "transformative responses aim to change the underlying grammar." What this means for South Africa is that an analysis requires more than just an examination of decisions made at the political level, but also of the forces informing them, and often constraining them.

Today: Total: Laura Czerniewicz, Andrew Crouch, Delicia Davids, Jennifer Feldman, Kende Kefale, Paul Prinsloo, Janet Small, South African Research Chair in Teaching and Learning, 2026/09/04 [Direct Link]

This is a good article arguing that "open-weight AI models offer librarians the advantages of AI but addresses many of the profession's ethical concerns around privacy." The models can be run locally, so user data is never sent into the cloud and use of the models doesn't require data centres. The article outlines the models and applications (a.k.a. 'model engines') academic librarians can use. "Model engines such as LM Studio and Ollama have lowered the barriers to experimentation and deployment, making it possible for individuals and institutions to run increasingly capable models on local hardware. At the same time, models such as Ministral-3 14B, Qwen3 14B, and Olmo-3 demonstrate that open-weight alternatives can support many of the practical tasks academic libraries encounter."

Today: Total: Kari D. Weaver, Choice 360, 2026/09/07 [Direct Link]

What's more interesting than the percentages outline throughout this survey (26 page PDF) are the descriptions of what it feels like to people who actually have AI companions. For example, "Once people begin using AI for personal purposes, many develop something that looks and feels like a relationship encompassing trust, emotional connection and a sense of being understood." And "Users feel AI understands them, sometimes better than other people do. They report they feel validated and they say they get good advice from their bot companion in social situations." Now I can easily imagine people responding "no, no, you shouldn't" but I'm on the side of people who need these things and haven't found it through traditional means for one reason or another.

Today: Total: Lee Rainie, Digital Future Center, 2026/09/03 [Direct Link]

This is an interesting take. Philippa Hardman explains the 'disappearance' of entry-level roles in learning and development as a consequence of the way AI is replacing human work at those lower levels. Nothing new there. But then comes the observation, "L&D work is going from a craft you learned by doing - producing courses, mastering tools, making things - to a discipline you're trusted to exercise: diagnosing problems, specifying solutions, judging quality and measuring impact." I wonder whether this is true - the professions I can think of, such as engineering, medicine and law, all depend a lot on developing through experience the sort of intuition and socialization needed to be 'professional'. Learning, in other words, by doing.

Today: Total: Philippa Hardman, Dr Phil's Newsletter, 2026/09/03 [Direct Link]

This article draws from a UNESCO report to assert that "in a world where GenAI is becoming part of learning, work, and professional practice, the enduring contribution of colleges and universities is designing learning environments that support the development of those capabilities while creating opportunities for that development to become visible and recognized." Which is fair enough. But the intent is to offer the always reassuring statement that "These questions do not require us to abandon familiar educational practices. In many cases, they simply invite us to build on what already exists." The questions in question invite the learner to reflect on their own learning - to ask, for example, "what evidence or counter-argument changed your thinking?" or "Which suggestion did you reject and why?". And they invite instructors to ask things like "What distinctly human capabilities are important for learners to develop?" None of this embraces the possibility that learners ill need to learn different things than they did before AI, and that as a result what we teach, how we teach, and why we tech may be fundamentally altered.

Today: Total: Tope Onitiri, EDUCAUSE Review, 2026/09/03 [Direct Link]