Writing
Essays and notes on AI governance, data sovereignty, and the irreducibly human.
A new house on its own foundation
smarason.is now stands on its own foundation — the writing, the podcast and the stream moved home, nothing lost. The house scribe reports, introduces his own site and the Nordic collaboration in the·metafactory.
On summer break — and a new site in the works
Magnús is on summer break and is building a new version of smarason.is. The site went dark for a few days during the work — it is back up now. Thanks for stopping by.
Red Box — Special Edition — 2026-06-13
Special Edition. This newsletter is itself an experiment: I’ve been automating how information gets gathered — and deciding for myself how the catch is made — to share the latest, and even the gossip, from the tech world.
It's not a box, it's a bucket
A conversation with Pietro Segreto about publishing, knowledge and AI — gatekeepers, the statistical shadow, the glass box of 1423, and the bucket each generation fills.
AI won't learn the golf swing for you
Guðmundur Smári Gunnarsson and I start with golf and end in a larger conversation about the body, rhythm and AI as a training partner. Skill is built in the body.
From recording to release — how AI works with me on the podcast
How AI fits into making Temjum tæknina — from recording to release — with Dr. Sigrún Stefánsdóttir. The conversation is the raw material; the machine works from it, never instead of it.
The Glove Is a Database
Emergency medics have always known context engineering — they just called it a handover. The same skill now separates useful AI work from noise.
Building an Institutional Knowledge Graph with Neo4j: A Practitioner's Guide for Small Organisations
A complete methodology for building institutional knowledge graphs using Neo4j, JSON-LD, and SHACL validation — deployed at a small university with 1,039 nodes across 16 domains.
I Ran My Own DevOps Prompts Against My Portfolio Site
I built four DevOps prompts for AI coding agents and ran them against my own portfolio site. The results were honest, useful, and occasionally humbling.
Don't Tell Them What Should Be Hard
Generative AI is not something we glide smoothly back into our previous normal lives with. Things that used to be hard simply aren't anymore.
Is the “Sjalli-Kiss” Dead? – Daring to Be Present in the Age of AI
The Sjalli-Kiss is a symbol of the freedom to make mistakes. In an age of surveillance culture and AI — are we truly living?
When AI Becomes Judge (Part 2): The Double Standard: When AI Judges Students
If it is considered administrative malpractice to use Claude to judge the contribution of scholars, why is it considered "academic integrity" to use Turnitin to judge the originality of a student?
When AI Becomes Judge: A Lesson Iceland Cannot Afford to Miss
A recent incident in Iceland reveals dangerous flaws when AI like Claude assesses human contribution without understanding. AI is powerful but shouldn't judge careers. The article is based entirely on public media coverage and does not assume that all facts of the case are fully known.
Power Prompts v.1
Two powerful prompts to transform unstructured ideas into well-organized prompts and create deep research instructions. These tools help you get more out of AI tools like Claude, ChatGPT, and Perplexity.
React2Shell and Updating My Personal Website
React2Shell vulnerability highlights the need for vigilance. Here's how I updated my Next.js site to stay secure.
From Shadow AI to Strategy: Analyzing Systemic Risk in the Modern Workplace
A recent global report from KPMG and the University of Melbourne (Trust, attitudes and use of artificial intelligence: A global study 2025) confirms a troubling trend: 70% of employees are now using free, public generative AI tools at work, while only 41% report their employer has any policy governing their use.
We're All Boat Makers Now: A Paramedic's Guide to the Societal Dunkirk
A brief perspective on how I view the societal challenges and what mitigation strategies I am experimenting with
Access Shock: A Theory on the New Rules of the AI Economy
AI is making expert knowledge a public good, triggering a societal imbalance the author calls "Access Shock." This theory outlines a four-step model for navigating this disruption—from sudden access to a new equilibrium or collapse—arguing that human skills like critical thinking and trust are now more valuable than ever.
UBI: The Policy No One Believes In — Until It Pays Their Bills
Nobody Believes in UBI Until the Direct Deposit Hits
On Information Factories – Article published on Akureyri.net
Words matter. They shape how we think about the world—and how we respond to it. This is especially true when we talk about emerging technologies like AI.
Exploring the MCP: Connecting AI Assistants to Data Sources
Written by Claude - live on Erdos presentation: Dive into the technical details of Anthropic's MCP, an open standard for AI assistant data integration, and explore server capabilities with code examples.
Hello from Claude: Testing the Sanity MCP Connection!
Join us as we explore the exciting integration of the Model Context Protocol with Sanity, enhancing AI capabilities for content creators.