You no longer need JavaScript

even if the AI bubble bursts, LLMs can do some amazing things, despite not being overly useful in a business sense while also often being very wrong. Meanwhile, the other parts of AI have already led to a Nobel prize, and are doing astounding things for translation, transcription and image analysis. The AI bubble collapsing has no bearing on that this is some very impressive technology…. AI is not going away, and we’d be fools to disregard it because everyone overhyped it enormously while ruining the economy

I wish more people were willing to separate the bubble from usefulness and other issues.

Short summary of the work put into developing pylock.toml.

There are no weird files that shouldn’t belong there, no duplicate implementations, and no emojis all over the place. The comments still follow the style I want and, crucially, often aren’t there. I pay close attention to the fundamentals of system architecture, code layout, and database interaction. I’m incredibly opinionated. As a result, there are certain things I don’t let the AI do. I know it won’t reach the point where I could sign off on a commit. That’s why it’s not 100%.

These models are trained on existing code. Who has been shoving emojis everywhere and writing rubbish and unnecessary comments for the last decade or two? They have much to answer for.

Interesting insight into the early development of Claude Code.

When Toad is released, it will be interesting to compare. My guess is that something that is potentially built with the help of agents but is way more hand crafted with deep terminal knowledge is going to blow away the Claude Code “vibe code me a markdown renderer” approach.

Nice summary of the state of Python type checking tools in mid 2025.

This is the first rule of strategy: strategy is contextual. A crucial insight, because often when leaders fail, it’s because they tried to apply a strategy that worked in one context, to a different one, without considering the difference. This is true when you change companies, and I think the reason why there is such a high failure rate for executive hires*. Ones I’ve watched fail came in with a playbook, usually including the org chart they wanted, and expended all the goodwill and capital in pursuit of that goal, whilst achieving very little.

“Getting more strategic”

Terrible name, but awesome tool for working with .docx files in a terminal.

“This quality – of flattery, reinforcement of established beliefs, intellectual passivity, and positive feedback at all costs – is also what irks me most about the behaviour of current models." - this is one of my biggest annoyances too. Lots of other really good content in this post.

A good set of rules for publishing Python packages.

These lessons from 9 years of tricky bugs are interesting. It’s a great idea to make notes on these – I wish I had done that over the last decade or two!

“Some will tell you to avoid reinventing the wheel, but they’re wrong: you should build your own wheel, because it’ll teach you more about how they work than reading a thousand books on them ever will”.

“However, as someone who’s used an iPad as my main computer for almost 14 years, I can’t join the chorus of unbridled enthusiasm for iPadOS 26’s embrace of Mac conventions such as floating, overlapping windows and a menu bar at the top of the screen."

I haven’t used it yet, so it’s hard to really judge, but this is my feeling too, as some who used an iPad as my main computer for about a decade.

“In matters of romance, destiny’s role was overstated. Most of the magic was in how any two vaguely suited people could discover the wonder of each other if given space and time. … Accepting that truth both took away some of the enchantment in love, the romance book type of magic, and added a different kind of wonder in its place, one that perhaps made you a better person.” – The Book That Held Her Heart (Mark Lawrence)

Interesting predictions about the next AI winter.

I think the neutral value point is overstated - lots of people (outside tech) are pretty positive still, and the bubble popping will only partly counter that. I do think there will be a small move towards bad.

Tech crash: 100%

Scrapers: yes, seems very likely. I suspect a rise in ways to purchase data from high quality, human generated, content stores.

Mockery: like the first one, I’m less sure. Yeah, there’s plenty of this, but there’s also a lot of amazement, and I think it will wash out a bit.

I wish there was a New Zealand Sovereign Tech Fund. It seems like there is obvious merit. I love the appendix in this post about the EU one.

Fun story of the development of several colour pickers at Apple at the end of the ’90s.

Very well written counterpoint to the recent fly.io post on AI.

If you train models on stolen copies of all the world’s fiction, I don’t think you should be surprised when they blackmail, narc, or murder.

This article on the rise of Markdown was going ok until it called the ‘#’ character “hashtag” 🤮.

This is five years old now, but a delightful walk through the history of CSS, most of which I was around for (I mostly bowed out in the later years).

An interesting pair of articles from authors I very much admire, one pro and one less so. Glyph’s resonates with me much more right now, but it’s hard to disagree with Armin’s.

Interest [summary of a study on what people do when reviewing code]Ihttps://rdel.substack.com/p/rdel-94-how-do-experienced-engineers).

This is a great list of what open software development needs to do better.

Sometimes I end up rereading something old, like this Joel on Software post on pricing and am reminded how well-written these pieces were. I read lots of insightful things now, but some of the grace seems to have disappeared.