@flowblok@tech.lgbtSite Reliability Engineer 
@flowblokSome statistics from my #AdventOfCode this year (for full gory details, see this thread: https://tech.lgbt/@flowblok/109438322711775692).
Of the 25 languages I used, I had used 13 of them for last year’s AoC.
Of the 12 fresh faces, I had previous experience in 3 of them: I learnt 9 new languages.
And finally, here’s my personal stats, annotated with each day’s language:
My god https://github.com/ArcaneNibble/i-cant-believe-its-not-webusb is an incredible hack but also I'm now wondering whether there ought to be a spec for devices to explicitly opt into WebUSB
Whilst I'm talking #copyright today, it seems to me that the Trump administration has repudiated the AUSFTA by unilaterally imposing tariffs on Australian goods.
Which means we're now free to reverse the damaging 20-year extention to copyright terms the AUSFTA imposed:
@chockenberry @maccatalan kind of amazing (or unsettling) that you could paste this paragraph into a Microsoft press release about Copilot AI *today* and it wouldn’t seem out of place.
@flowblokSpent this evening generalising the implementation and making the URLs shareable, have a play: https://static.flowblok.id.au/pub/ringsteady-art/#spec=1-16,64,16-1&scale
(apologies that it's not mobile-friendly, though it should still work)
@flowblokCoded up some art this evening.
I'd originally intended to do some maths so I could do a different kind of art, but decided this was easier and more fun.
Then you have little problems like _this_:
https://landley.net/history/mirror/institutional_memory.html
3:O)>
I have been asked strange questions about old software, including "Do you have any contact details for $Person?" by someone wanting to know about the Y2K fix some 20 years on... (Because they will need to repeat the exercise in 2027 at the latest, but they didn't say that.) I have a broad outline but no specifics, and gave them the LinkedIn profile details - that they had and were not working. Oops!
on my blog!
https://dotat.at/@/2025-02-13-pcg32-simd.html
the cunning insight that makes it possible to get eight successive results from one instance of pcg32 in little more than a nanosecond
TIL big specialized forums have started backdating millions of LLM-generated posts. Now you cannot be sure a reply from 2009 on some forum for physics or maps or flower or drill enthusiasts haven't been machine-generated and totally wrong.
@flowblokI’ve been writing this up as a blog post, except I got to multiplication, realised there’s a faster way to do one of the steps, and now Part 1 of Day 7 runs almost 5x faster. Oops.
@flowblokOH: I cleaned the kitchen because GitHub have an outage.
Regarding authority:
Colocation datacentres are shared telecommunications facilities — Lots of ISPs, phone companies and large corporates all renting space in the same place, sometimes the same room, so they can connect to each other.
There’s an obvious need for strict norms of conduct: Don’t cause outages, don’t touch anyone else’s stuff, don’t move cables unless you know what they’re connected to, and so on. Most facilities have a site induction manual which goes into some detail about the rules of engagement, and a small team of irritated-looking 24x7 security guards to make sure standards are maintained.
One of the rules at one of the facilities in Sydney is that you’re not supposed to have cardboard anywhere in the technical spaces. If you take delivery of equipment, you’re supposed to unpack it outside in the corridor and carry it into the colo space without any packaging.
The ostensible reason is that cardboard is not only a risk in the event of a fire, but it also exudes dust, and these spaces are supposed to be very clean and well filtered. Shared space, your dust might affect others.
Backing-up the cardboard rule is a price-listed entry on the facility’s service schedule for rubbish removal. If you leave cardboard in the space, they’ll send someone to dispose of it and charge you some three-figure sum of dollars as a form of liquidated damages. /1
@flowblokAlmost there, just have Part 2 of Day 24 left. I have two ideas for how to go about it, but my brain is done for today, so I'm going to add some cosmetic touches and then go play games.
@flowblokPart 2 of Day 21 is also out, but then I went to release it and discovered it was too big for the JS interpreter to handle. I'd also noticed it had gotten rather slow to compile…
I’ve now reduced this down to a minimal test case, and I (proudly? shamefully?) present https://inform7.atlassian.net/browse/I7-2584.
@flowblokThe performance for Part 2 of Day 23 is terrible[1], but I have gotten the correct answer.
[1] I have discovered that not only does my list type lack amortized constant time append, it's also built on top of a linked list, though in a way that I don't yet have a good mental model for.
@flowblokPublished another update, with everything except Part 2 of Days 17, 21, 23 and 24.
