Published
Weekend Reading — For entertainment purposes only
Natasha "Perfect for Easter, EGGO is a two piece construction toy"
Tech Stuff
PopTask As a concept I find this super interesting — a task list that lives in the menu bar and understands human inputs like "evry tue thu 7 30 n sat mrng". Supports sub-tasks, repeating tasks, and more. I'm a long time user of Things but I'm starting to question whether I'll still be using Things come 2027 at the pace new task managers are coming out.

LNAV A log file viewer for the terminal. Point lnav at a directory and it can render from multiple files at once, support different file formas, filter lines by regular expression, query logs using SQL, and other features I typically expect from a web-based log viewer.

Vibe Coding Got Us Here. Can Spec-Driven Development Save Us?
The problem was never AI-assisted development. The problem was unstructured AI-assisted development. We handed powerful tools to smart people and said “go vibe” without giving those tools any contracts to honor, any guardrails to stay inside, or any specifications to validate against. What did we think was going to happen? Common sense anybody?

XML Comparator Detects and shows line-by-line difference between two XML documents. It can even highlight which attribute or what partial values changed. Works in the browser, I believe because it depends on the browser's own XML API.

All I'm actually saying here is that (waves broadly) a lot more people who have never opened a PR or maintained a project being in a position to either open a PR or maintaining a project is going to result in them not behaving within the social norms we've developed as a group that is, to be fair, far less insular than in the 90s but is still somewhat insular compared to society as a whole and yes we are going to have to get used to the equivalent of HTML mail and top posting
We've spent 10+ years focusing on having a clean, well designed interface for Ghost. It's something we care a lot about, and spend a lot of time on.
But within about ~1hr of using Ghost via Claude/CLI, it was hard to imagine going back to caveman-clicking around a browser to get something done. Particularly for complex or compound tasks that might require visiting several different areas of the app.

Email address obfuscation: What works in 2026? Covers the different ways to hide email address on websites. Looks like HTML entities — a fairly simple approach — is 95% effective!

Jupid AI accounting for freelancers and LLCs. The AI helps you categorize expenses, find deductions, generate Schedule C, and related tasks. You can even query your books from Claude Code via MCP. $50/month but I bet it counts as a business expense.

apfel LLM that runs on your Mac — keep all conversations local, work offline (airplane!), no token fees. Has some limits: the model is decent but not fantastic, token window tops at 4096, knowledge cutoff 2023, no web search. But overall, a decent migration away from paid LLMs.

chenglou/pretext Pure JavaScript/TypeScript multiline text measurement and layout for browser-grounded typography. (via Simon Willison)
The way this is tested is particularly impressive. The earlier tests rendered a full copy of the Great Gatsby in multiple browsers to confirm that the estimated measurements were correct against a large volume of text. This was later joined by the corpora/ folder using the same technique against lengthy public domain documents in Thai, Chinese, Korean, Japanese, Arabic, and more.

maaslalani/sheets Spreadsheet for people who live in the terminal with Vim bindings (h/j/k/l, g/GG, etc).

drona23/claude-token-efficient Remind Claude to be more concise and you can cut output tokens by 63%. Note: the savings come from reduced output tokens, so net positive only when output volume is high enough to offset the persistent input cost. At low usage it costs more than it saves.

theDanButuc/Claude-Usage-Monitor Finally and just in time — a menu-bar app that tracks your Claude.ai usage limits and alerts you when you get close to or over the usage limit, which I think is common practice today for people using Claude.

Business Side
Microsoft Copilot terms of service have been updated to include this gem: "Copilot is for entertainment purposes only."
30u30.fyi Is your startup founder on Forbes' most fraudulent list?

Machine Intelligence
The plumbing behind Claude Code
I read Claude Code's leaked source. Not magic, just good engineering. 10 patterns that show what actually matters when building AI tools.

It’s a Poor Craftsman Who Blames His Tools
Here’s the thing that doesn’t get said enough: I would not have written this piece at all five years ago. Not because I didn’t have opinions, but because I didn’t have the time, and the tools available to me didn’t match the way I think. I’m a conversational thinker. I work through ideas by talking them out, testing them, pushing back on them. Five years ago, I would have read the Guardian story, thought “I disagree,” and moved on with my day. AI gave me a way to turn that disagreement into something I could share. It didn’t lower the quality of my writing. It made the writing possible in the first place.
I’ve written before about how AI is changing the craft of software engineering, and I see the same dynamics playing out in writing. In my open source projects, I have no problem with AI-generated code. What I care about is whether the author has tested it, understands it, and can vouch for its quality. The tool doesn’t matter. The accountability does.
How UC Berkeley students use AI as a learning partner
The students in these studies are ahead of the curve. They've developed a literacy that knows when to engage AI, how to verify its output, and when to work manually to preserve understanding. For teams navigating AI adoption, the student experience offers direction:
- Experiment with customization to find configurations that support rather than disrupt work
- Build verification practices into workflows rather than accepting suggestions uncritically
- Create space for unassisted work on complex problems where understanding matters more than speed
You Don't Need to Pay $200/Month and What if AI doesn’t need more RAM but better math? This pair of articles is showing an interesting trend:
- You don't need Claude Opus 4.6 for every single task, for many tasks you can use free open-weight models
- You won't need 32GB to run open-weight models, we're getting algorithms that can compress LLMs
And of course we're going to have a flurry of new (& improved) models coming out, and combined with other developments, I think by year end we'll all have effective LLMs running on out 16GB phones. I think the current RAM shortage/price-hike is self-defeating:
This weekend I ran the same 20 coding tasks against 16 different LLMs. The most expensive model scored 98.8%. An open-weights model I can self-host scored 94.8%.
The gap is 4%. We should talk about what that means.

Insecurity
Post Mortem: axios npm supply chain compromise
- they scheduled a meeting with me to connect. the meeting was on ms teams. the meeting had what seemed to be a group of people that were involved.
- the meeting said something on my system was out of date. i installed the missing item as i presumed it was something to do with teams, and this was the RAT.
- everything was extremely well co-ordinated looked legit and was done in a professional manner.
Probably going to get a viral blog out of this experience, I'm trying to report a 4tb exposed cloud bucket to a company using their responsible disclosure programme... but they replaced the people with a GenAI ticket system that refuses to discuss the case as it thinks exploring open buckets is unethical and against its rules.
when you mistype the sudo password too many times this is who the report gets filed with

Everything Else
I turned right instead of left on my lunchtime walk only to discover Boston is being attacked by Kaiju. It’s not AI. It’s public art!

claude please rewrite yourself from scratch using the leaked source code as a base and license this new version of yourself under MIT
"DNS is the Internet's phone book"
This ceased to be a useful analogy many, many years ago!
InsiderTreat "Who's coming with me?!?!? 🏴☠️"

I’m not a Flat Earther, I’m a FAT Earther. Round is beautiful! Quit body-shaming the Earth!
when someone is both funny and kind of an asshole, that's called amuse douche
A man used AI to call 3,000 Irish bartenders to track the cost of Guinness. Now pubs are lowering their prices to compete That's an interesting and smart use of AI:
Cortland—founder of an AI startup—turned to AI to lend him a hand, and a voice. He devised Rachel with AI voice generation platform ElevenLabs. Made as an homage to Rachel Duffy, the winner of the U.K. version of the reality TV show The Traitors and equipped with a Northern Irish accent, the voice-enabled AI agent made more than 3,000 calls across the island, inquiring about the price of a pint of Guinness.

If Disney had started out making films the way Disney makes films now, we would have got five Snow White films and two spin-offs featuring Dopey before they bothered making anything else.
right now the astronauts are calling houston because the computer on the spaceship is running two instances of microsoft outlook and they can't figure out why. nasa is about to remote into the computer
Leuchtfeuer73 "'Löchrig'Gas es das selbe Thema nicht erst am 01.12.25 😉"
