I like 🖥 computers, and I like 📝 writing about them! I also have 🧠 an Obsidian vault where I dump most of my notes. I make 🎵 music as slono. I like 🎨 drawing and 🖊 handwriting too!
Yesterday the LLM sliced through a problem like butter. Today it can't follow a three-line instruction without going sideways. You press regenerate, rephrase the prompt, add qualifications — "no, I meant this, not that" — the context window slowly fills with noise, further degrading the model's output. A roulette wheel of code: "maybe the machine will do what I ask for, maybe it won't, but I have…
The avoidable process error was mine: I briefly launched two flash/probe jobs in parallel against the same /dev/ttyACM0 device. That is invalid for this setup and could have polluted the evidence, so I killed both jobs, wrote down the mistake here, and added a standing repo instruction to AGENTS.md so future work treats each serial device as single-owner during flash/probe work. This is what an AI…
Everyone is trying to go faster, running more agents to run more agents, to open more pull requests to orchestrator agents to then get reviewed by another fleet of agents. The arrival of coding agents has turned software development into something that looks like a factory (or casino) floor — and the temptation is to optimize for throughput, to feel the dopamine hit of watching five Codex tasks…
(Side note, you can also find this article on substack: like and subscribe if you want to help me spread the good word ) A key attribute of good software has always been simplicity; the ability to say: this is the problem, and this is the smallest thing that solves it. The layers of technology we've added since the early days of computing of are mostly overhead we've added for ourselves, not for…
Using notation, tools or code don’t make the model smarter. They make the task simpler. That’s the quiet trick behind “tool-using agents,” DSL prompting, and “just write code.” These prompts rearrange the work the model has to do so that the parts requiring real generalization get pushed into deterministic machinery—databases, runtimes, calculators, typecheckers—and the model is left with what it…
There's a myth floating around that large language models (LLMs) like sonnet or gpt are mainly good for greenfield, one-off projects. It's true they're fantastic for brainstorming new ideas, prototyping quickly, or scaffolding out fresh codebases. But what's surprising—and actually much more exciting—is how effectively you can deploy them against legacy code. I've spent a fair bit of my life deep…
You've probably heard about vibe coding—using large language models (LLMs) like ChatGPT or Copilot to program via natural language. There’s been a fair amount of hype, skepticism, and frankly, confusion around whether these models can really handle the complexity of big codebases. Let's cut through the noise: vibe coding on large-scale software is not only feasible—it's downright practical, as…
(This is part of a longer series of short essays I'm writing as part of a marketing writing course. This is the first article I think is meaty enough to be published here. I will rework most of my material once the course is over.) Unit tests enable bold refactors and protect against subtle errors introduced by Large Language Models. However, while unit testing strengthens codebases for the long…
I am notoriously distracted and both my long and short term memories are less reliable than a hobbyist's kubernetes cluster. After 20 years of python, I still can't remember what the arguments to split are. Thankfully, these days are now over, as I can ask a just as clueless LLM to answer the myriad of questions I would otherwise have to google everyday (please don't use google, use kagi , you'll…
I have been playing with Tarot a little bit over the last year, inspired by hanging out with mysticism connaisseur Bob Doto and Xe Iaso's Tarot for Hackers . I like divergent creative methods and my brain seems to gel with that kind of thinking anyway. I also consider large language models to be akin interpreting tea leaves, except that instead of tea leaves, I have semantically reach language…