Blog
Original writing on the tech that's actually worth your attention: using AI well (especially for code), running models on your own hardware, weekend projects you can follow along, and the practices that hold up. Every piece is researched and sourced — no hot takes without receipts.
Follow along: subscribe with 📡 RSS and new posts land in your reader automatically — no email, no sign-up.
Personal Agent Protocol: what Sierra and Meta are proposing
Sierra and Meta announced an OAuth-based open standard for how AI agents deal with businesses. What it says it does, how it compares with MCP and Visa's protocol, and what is still missing, including a published spec.
Home Assistant starter guide: automate three things this weekend
Get Home Assistant running on a Green or a Raspberry Pi in an afternoon, then build three automations worth having: lights before dusk, a door-left-open alert and an empty-house sweep, plus when not to bother.
Password managers vs passkeys: where we actually are
Passkeys fix phishing and are now the default on the big platforms, but nearly every account still has a password behind it. Where adoption, recovery and portability really stand in October 2026, and the setup to have this year.
Choosing a GPU for local AI: what the numbers actually mean
For running language models at home, memory capacity decides what fits and memory bandwidth decides how fast it talks. How to read a spec sheet, size a model against a card, and when not to buy one.
How image generation models work: diffusion, explained gently
AI image generators start from pure static and remove noise step by step. How that works, how your prompt steers it, what steps, guidance and seeds actually control, and where it still falls short.
E-ink dashboards: build a low-power family display with a Raspberry Pi
Pair a Raspberry Pi Zero 2 W with a colour e-ink panel for a wall dashboard that looks like paper: the parts list, the header gotcha, the software, and why "low power" only describes the glass.
Backing up your digital life properly: the 3-2-1 rule in practice
Three copies, two kinds of media, one off-site. What each number is actually protecting you from, why cloud sync doesn't count as a second copy, and a setup you can build in an afternoon.
What AI agents actually are (and what they aren't)
An agent is a language model running in a loop with tools. What that buys you, what the measured evidence says they can do as of mid-2026, and the failure modes demos never show.
Weekend build: an ESP32 temperature and humidity monitor with a web dashboard
A follow-along build: wire a sensor to a $10 ESP32, flash ESPHome, and get a live web dashboard on your network — no soldering, no C code, plus the honest limits of a cheap sensor.
Self-hosting 101: what's worth running on your own hardware in 2026
The hardware you actually need, why Docker makes it manageable, and the handful of services — Immich, Jellyfin, Home Assistant, Vaultwarden — genuinely worth running yourself.
Prompt patterns that actually work for coding agents
Six prompting patterns that reliably improve coding-agent output — specificity, plan-first, verification and the three agent reminders — each with a concrete before-and-after.
Using AI to write code: what actually works
AI assistants make good developers faster and careless ones slower. The workflows, review habits and guardrails that separate the two.
Running AI on your own hardware: local LLMs explained
What it takes to run a capable model on your own machine — the hardware that matters, the tools to start with, and what to realistically expect.
Weekend project: block ads on your whole network with a Raspberry Pi
A follow-along Pi-hole build: parts, setup, and what changes for every device in the house — no soldering, one evening.
How large language models actually work
Tokens, attention, training and why models hallucinate — the mechanics behind the chatbot, explained without the math degree.
Which AI model should you actually use? A practical 2026 field guide
The major AI models as of mid-2026, what each is genuinely good at, and a decision guide by task and budget.
Looking for evergreen how-tos tied to the tools? Those live in Guides.