Weeknotes 263
12th July, 2026
“Persistent disks”
-
Once your identity is invested in a tool, admitting its flaws starts to feel like admitting something about yourself. So people don’t just tolerate the flaws—they defend them, and eventually flaunt them. You cannot have an honest conversation about a tool with someone who’s decided the tool is part of their personality.
This is why you need to be careful criticising CSS.
To be clear, I’m not saying text editors don’t matter to your workflow. I’m questioning the near-religious devotion people have to a tool because it gives them a “hacker vibe” – which is basically the whole appeal for newcomers to vim or emacs.
I definitely feel this way about Neovim. When I first wanted to learn Vim the primary reason was because it just seemed so cool and hacker-like. It also seemed near impossible. Everything about how I manipulated text on a computer was different. Somehow I muddled through, and got to a point where I use it every day. Now I’ve sunk so much time and effort into it that the fact I can never remember how best to search and replace after 10+ years of use doesn’t seem to be the issue that maybe it is.
I do keep meaning to try Zed seriously. It looks cool.
-
The 10x AI Delusion & The Real Bottlenecks
The moment humans stop reading the code, the repository becomes a black box. You inherit a system nobody understands, maintained by agents nobody is auditing, generating dependencies nobody can explain. For a regional bank or a healthcare provider, this is not a bottleneck; it is a regulatory event.
-
Worktrunk is a CLI for Git worktree management, designed for parallel AI agent workflows
OK.
But the git worktree UX is clunky.
They’re not wrong about that.
I’m not sure I need this, yet.
-
Germany’s richest man takes on Big Tech
Dieter Schwarz made his fortune from supermarkets, becoming Germany’s wealthiest person. Now he wants to take on Google, Microsoft and Amazon and an entire region stands to benefit.
-
The latest ChatGPT Mac app:
Shouldn’t LLM use encourage the development of native apps? It should be so much easier to support multiple native versions of an app now, no?
Surely Electron was about utilising many of the same skills people making your web app already had to also build a cross-platform Desktop app? Small teams would use toolkits like Electron so that you could have an app at all.
-
Do you really need separate systems for caching, queues, search, documents, and vector embeddings when you already have Postgres?
Postgres indeed has many things.
-
I was talking to someone (I forget who, sorry!) about how when I started out programming the first language I used was PHP paired with MySQL. It was drummed into me via books of the time that performing any database query was terrible thing to be long considered. All JOINS were bad. A subquery, heresy!
But both hardware and software are both a completely different beast now – it’s been a long time, especially in computer years. But those lessons learnt early on can stick in your mind. I’m not saying we don’t need worry about performing needless queries, but I definitely need to worry less and take note of the numbers, and if required, measure.
-
My Thoughts on the Bun Rust Rewrite
I don’t know an awful lot about much of this but this does seem like exactly the sort of JavaScript runtime I would not use and also exactly what everyone feared might happen with AI slop.
From the Reddit thread where I found this:
This whole repo is insane, an AI agent picks up the issue, makes changes and creates a PR, then other AI agents review and merge it.
And it’s true, it’s quite something. Bun are now owned by Anthropic so is this just a big PR stunt or something anyone would actually rely on?
-
Bunny DNS is now free (although they have a $1/month minimum spend – to stop fraudsters I assume?). I wonder how this compares to DNSimple.
-
I’ve noticed various options springing up that are marketed as ways of running LLMs in the “cloud”. Interestingly these are all “stateful”. They have persistent disks and checkpoint features as a way of saving and restoring the machine to certain points.
-
opencomputer 🤷♀️
-
exe.dev which I’ve mentioned before.
-
sprites.dev from the folks at Fly.io. This is the only option I’ve tried. Getting up and to a running machine was very quick and easy but I haven’t gotten much further than that yet.
I watch with interest.
-
-
Break large changes into small, reviewable pull requests that build on each other — with native GitHub support and the gh stack CLI.
Nice to see native support for this coming. I’ve done this manually many times and it can get very tricky to manage.
-
Umami — another option for privacy focused web analytics
-
Using AI to write better code more slowly
I find this style of coding to be a more super-powered version of the kind of programming I was already trying to do before LLMs: careful, methodical, quality-obsessed, focused on making things better for the next coder.
Cowboy coders have always been writing terrible software, and they’ll continue to write terrible software but faster via “vibe coding”.
-
The Minimum Viable Unit of Saleable Software
Don’t get me wrong, I hate Jira just as much as anyone who’s ever used it and have a nearly uncontrollable urge to want to rebuild it too, but the math here doesn’t pencil out
-
The Beginning of Programming as We’ll Know It
The time may come, perhaps even soon, when AI takes over programming completely. But in the mean time, a programmer who embraces AI, yet is skeptical about everything it creates, is better-equipped than any comparably-skilled human in programming history.
-
Redis Cluster Support, Without Writing Code – This is a very interesting post about adding new big features to an Open Source project using LLMs.
My biggest takeaway is that engineering is still engineering, whoever does it. Directing agents to the right architecture, pushing them to write more tests, steering them to handle edge and failure cases—these are all things that I had to do.
Yes, ignore the people “one-shotting”.
The most stressful part of all this is that I’m now responsible for code that I did not author. That’s common in the workplace, but not so much in OSS and especially in my personal experience, as I maintain quite the big bunch of libraries where I’m the sole author. I can read the code again and again, but there must be some different process happening in the brain compared to reading code that you yourself have written. Only time will tell how things will go in this regard.
Indeed. Whilst I am “writing code” with the LLM now I am doing my very best to read the code and understand what it is doing. I actually feel more invested in some aspects of the code than I did previously too. But it’s different.
-
I finally finished watching The Strain. It’s quite good, stupid, fun.