Profile avatar
alanbuxton.bsky.social
Serial CTO etc in tech startups. Based in London. Also sometime classical guitar player. Working on an open(ish) data side project https://syracuse.1145.am
44 posts 481 followers 1,988 following
Regular Contributor
Active Commenter

More isn't always better, part 9 billion

Cribbed from LinkedIn.

I wrote a long and slightly rambling post about the hot new way to code that all the cool kids are doing. 2point0.ai/posts/vibe-c... #ai #coding #vibe-coding

Did my first successful headstand today. Still got some work to do before I can catch up with my 6 year old daughter. But we all need to start somewhere.

Fascinating Hacker News comment from Tom Gally, a professional translator (Japanese to English) who uses LLMs as part of his workflow, which he describes in detail: news.ycombinator.com/item?id=4289... Wrote a bit more about this on my blog here: simonwillison.net/2025/Feb/2/w...

THE COMING AI CONTRADICTION In today's Odd Lots newsletter, I wrote about the gulf between how traders are thinking about AI vs. how Silicon Valley is. At some point, this is going to have to break

I build software products. I spend a lot of time in hotels. Here are some learnings about UX that I've taken from hotel elevators. They aren't about software UIs which makes them super helpful analogies when talking about general user experience.

That's one way to game your NPS

I wrote, a bit tongue-in-cheek (but not much) that the UK's tax regime is encouraging second-time tech entrepreneurs to move to Dubai. I picked Dubai more or less at random based on 2 people I'm aware of who moved there.

I fixed an "off by one" error in one of my codebases. It had been in there for best part of 3 years and been executed a few million times without it manifesting. Just shows how long some bugs can sit around below the surface.

Lesson in importance of testing your backups. I have two similar databases, both being backed up the same way. Or so I thought. I'd tested one a lot and it worked fine, so I assumed the other would work too. It didn't.

The trees behind me trying their best to protect the frost from the oncoming sun.

got into a huge fight with the wife but she apologized by buying me a cybertruck and a brand new life insurance policy

A year ago I blogged about the AI slop increase in bug and security reporting: https://daniel.haxx.se/blog/2024/01/02/the-i-in-llm-stands-for-intelligence/ It has only gotten worse since then, but I wouldn't call it a serious problem just yet. The trajectory does not look good though.

My open-ish data project is mainly used by bots and scrapers. I want to keep things as open as poss so here are my latest commits to put pages behind a login but to make the login shared. github.com/alanbuxton/s... And then github.com/alanbuxton/s...

So... after clicking around I found my big refactor broke some pages. Those pages weren't being tested. Good opportunity to write some tests for them :) github.com/alanbuxton/s...

So today I committed a big refactoring in my side project. It highlighted 3 big things: 1. Something that seems like a relatively small refactoring can turn out enormous 2. Thank goodness for decent test coverage (see 1) 3. I've now got ChatGPT writing useful code. github.com/alanbuxton/s...

variable names *are* code comments

Definitely.

It's that time of year where I post this one

To everyone out there who needs help figuring out their 2025 objectives I give you a tried a tested process: Ancient Persian Decision-Making. According to Herodotus (writing around 425 BC):

I had my first success with ChatGPT writing a good piece of code for me. In the past I've given it tricky problems which it struggled with, or relatively simple things but in a language I'm not so up to date on (e.g. Bash scripts). But today it did a neat python function for me. Very impressed.

Pournelle's iron law of bureaucracy, in academia In any bureaucracy, the people devoted to the benefit of the bureaucracy itself always get in control and those dedicated to the goals that the bureaucracy is supposed to accomplish have less and less influence, and sometimes are eliminated entirely.