Ian O'Byrne | wiobyrne.com

Sometimes when we use AI the answer arrives looking finished.

AI really does introduce some genuinely new capabilities and conditions, but technological change does not wipe away everything we already know about literacy.

Why running a local AI model—and then turning off the wifi—changes how people understand where AI actually happens and what it costs.

Why an AI query has no universal energy cost—and what local models can teach us about the infrastructure hidden behind every answer.

Why I built a teacher-owned teacher.md profile file for a local classroom AI assistant.

Closing or reducing the distance between thinking, writing, and publishing by moving from WordPress to Astro.

Why I use AI to mirror my voice without outsourcing the thinking that happens in drafting.

How Karpathy's LLM wiki pattern maps onto educator knowledge work and local AI systems.

Practical ways to tell which AI model you are using when platforms and models stay opaque.

Why language models confidently misidentify themselves, and why that matters for AI literacy, trust, and how we evaluate these systems.

Notes from a March 2026 talk on helping learners use AI without surrendering judgment, agency, or responsibility.

What building TrustSense revealed about the current state of local AI: the model is easy now, but the real challenge is system design and trust.

How I took a one-click Jitsi install on Reclaim Cloud and turned it into a secure, branded, production-ready video conferencing setup.

What a self-hosting project taught me about computational thinking, risk, tradeoffs, and reasoning carefully through technical decisions.

Why InitiatED moved from institutional Zoom to a self-hosted Jitsi server as part of a broader commitment to digital sovereignty.

A reflection on how default digital tools became invisible infrastructure, and why sovereignty starts by questioning what we were told to trust.

Why the phrase "AI illiteracy" sounds useful at first, but may distort how we think about literacy, access, and responsibility.

Why the hardest part of bringing AI into curriculum is not logic or pedagogy, but the logistics, shame, and practical constraints educators actually face.

Why collapsing every generative system into "ChatGPT" weakens our thinking about AI, learning, and what these tools actually are.

A follow-up on ungrading that argues removing grades is only the beginning; agency requires labor, structure, and support.

What I learned when ungrading stopped being a rejection of points and became a question of infrastructure, power, and student support.

Why disappearing messages in Signal are less about secrecy and more about reducing the amount of sensitive data you have to protect.

How I turned a pile of local services into a coherent private cloud with domain names, HTTPS, and a real front door.

A practical guide to Signal Registration Lock and why one small setting can protect your account from SIM-swap style attacks.

What Signal safety numbers actually do, and how to think about them without panic, jargon, or blind trust.

Why Signal feels different from regular messaging apps, and how to explain end-to-end encryption in concrete, human terms.

A reflection on Vincent Harding's idea of the live human signpost and the quiet people who help others keep moving.

A friendly starting point for people who downloaded Signal and want to know what to do next, and why it matters.

Why convenience stopped being enough, and how digital sovereignty became a more useful frame for thinking about platforms, data, and control.

Why agency in AI work often feels like productive friction.

What digital literacy should become in a platform-dominated web, and why educators need to teach for sovereignty, not just participation.

A response to the idea that true sovereignty online is impossible, and an argument for why partial control still matters.

Why the internet itself is not broken, and why the real task is leaving platform silos rather than inventing a whole new web.

Why the old privacy-versus-security distinction no longer goes far enough in a world of AI, surveillance, and predictive systems.

Why checklists feel reassuring in complex digital systems, and why they often fail when what we really need is literacy and judgment.

How I went from thrifted hardware and confusion to a stable Proxmox base for the rest of my homelab.

Why prompt engineering is less about crafting perfect inputs and more about orchestrating an ongoing dialogue with the system.

A meditation on the everyday people who steady communities by helping others see clearly when things get hard.

What it means to design online communities around trust, safety, and shared responsibility instead of just technical features.

How to build digital communities that support people, not just platforms.

Why a living web presence should be more than a static CV.

What it means to federate your self across website, newsletter, and notes.

A real-time, messy walkthrough of building a Proxmox homelab.

A reflection on holding onto hope when things feel hard.

How Obsidian can turn a newsletter into a connected knowledge garden.

How my first DIY server kicked off a homelab journey.

What to consider when campus IDs move onto your phone.

Why privacy matters when data collection feels routine.

Why building a homelab helps me understand the systems I rely on.

How a homelab project became an experiment in writing with AI.