Matthew Garrett

16.1K Followers
168 Following
6.6K Posts
Former biologist. Actual PhD in genetics. Security at Nvidia, OS security teaching at https://www.ischool.berkeley.edu. Blog: https://codon.org.uk/~mjg59/blog . He/him.
Bloghttps://codon.org.uk/~mjg59/blog
Signal@mjg.59
I have finally actually got around to booking flights, and still looking forward to not having to participate in the "Can Matthew enter the UK without being arrested" challenge hurrah hurrah hurrah
For Reasons that are different to previous Reasons, I have an excess train ticket from Paddington to Ledbury and back, leaving Thursday and returning Monday, that I would happily hand off to someone making the trip to #emfcamp (I will be there, I am just bad at buying the correct number of train tickets)
(And, re: my previous comments about LLMs being an effective way to have non-coders make local changes they want anyway - if you don't care about upstreaming your code, that seems entirely fine, there's no need for it to be maintainable or extensible)
Ah right it's because it's written like it's being written to ensure something works rather than written with any understanding of the underlying design and that's how proprietary software can end up working in a sufficiently siloed environment and also what happens when something that doesn't understand software at all is asked to write more
Open source is extremely imperfect but from a practical perspective it does tend to end up with most new code going through some sort of review process by someone with a holistic understanding of the codebase as a whole
The code I'm looking at works just fine! But we end up going through certain code paths twice for no reason whatsoever, and in one of them command line options aren't applied, because those are parsed twice in two different places
And, to be fair, that's also the kind of thing I'd expect from a lot of proprietary software
LLM-written code is, unsurprisingly, in the absence of human review frustratingly annoying to modify by hand
(Which leaves me confused why they have so much difficulty following the first order)
Going through paperwork and realising that Tuesday will be 6 months since https://codon.org.uk/~mjg59/case/order2.pdf went into force, which means the case summary can be removed. The hearing concerning whether the Schestowitzes are in contempt for failing to follow the previous order (https://codon.org.uk/~mjg59/case/order.pdf) is scheduled for the 8th and 9th of July, about three and a half weeks from now, but the second order was either correctly followed or rendered moot by their bankruptcy.