Weeknotes 320

Wassup.

Three gyms. Chicken bones on the gym stairs. It’s actually quite helpful to have bullied myself into this Monday/Wednesday/Friday pattern because it’s eliminated any need to make a decision when I wake up groggy at 5:30am.

Shroud is intriguing so far, although as always I’m having to push through the discomfort of memorising yet another list of made-up names for the crew of a spaceship. It was slow to start but things are happening now.

Irrationally I’m not particularly enjoying Sophie Aldred as a narrator, mainly because she sounds a bit too much like Harriet from The Traitors.

I cycled to the Barbican Waitrose to buy more of their own-brand bran sticks. They’d completely sold out so I had to buy the vile Kellogg’s ones instead.

Look: if you’ve been using my weeknotes to triangulate the hottest sources of alt-bran in London and then clean them out, please stop it because I can’t survive much longer on this absolute horror of a remastered breakfast cereal. Also, rude.

FRY-UP FOLLOW-UP: the mystery of what’s going on with falsehood-themed vegan comestible Just Egg has been somewhat resolved by this article which says it’s in limbo because of regulatory delays preventing the sale of mung bean protein and — probably more importantly? — everyone at its UK distributor being made redundant. So I won’t hold my breath.

TEDIOUS GIT CORNER: Whenever I want to execute some command (e.g. run the tests) on every commit in a branch, which I probably do tens of times per day, I have an ancient habit of using GIT_SEQUENCE_EDITOR=​true git rebase -⁠-⁠interactive -⁠-⁠exec <cmd>. It’s a bit contrived because I don’t actually want to do an interactive rebase, but “[-⁠-⁠exec] can only be used with the -⁠-⁠interactive option”. By setting GIT_SEQUENCE_EDITOR I’m telling Git to use true to “edit” the todo list instead of Vim or whatever, i.e. not bother opening an editor at all, to allow the rebase to proceed immediately as if it were non-interactive.

However, this week I discovered that a decade ago Git decoupled -⁠-⁠exec from -⁠-⁠interactive, and now “[-⁠-⁠exec] uses the -⁠-⁠interactive machinery internally, but it can be run without an explicit -⁠-⁠interactive”. So for ages it’s been possible to use the much simpler git rebase -⁠-⁠exec <cmd> to achieve the same thing and I’ve been wasting my life. What a world!

I happened upon my RubyEvents profile and realised it needed some cleaning up because my name isn’t unique.

I also noticed that several of my talks aren’t listed there, mostly LRUG ones. This turns out to be because Murray’s valiant import of LRUG meetings intentionally only goes back to 2020 since all earlier videos have disappeared down the Skills Matter memory hole.

Which made me consider — ignoring RubyEvents for now — whether I should do something about all the broken video links on my talks page. Is it worth recovering the videos and sticking them on YouTube, like I did for The DHH Problem? They have little intrinsic value so I don’t really care about them being unavailable today, but maybe in future I’d regret letting them bitrot into oblivion? Also it just kind of looks bad to have dud links on my website.

I haven’t decided what to do yet, but obviously the first step in doing anything would be to get the videos.

Years ago, in anticipation of this sort of thing, I downloaded videos of all my talks, stuck them in an Amazon Glacier vault and forgot about it. Then in October last year AWS emailed to say they were deprecating the standalone Glacier product and not accepting any new customers because it’s effectively been replaced by the Glacier Instant Retrieval, Flexible Retrieval and Deep Archive S3 storage classes. I ignored that message because it didn’t seem like something I’d ever need to deal with.

But now, I suppose, I might want to download those videos, so it might, I suppose, be worth moving them to S3 so they’re not trapped in a moribund service that nobody uses any more, like Heroku.

I tried to migrate them with the official process which involves launching a CloudFormation stack to run an eyewateringly complex set of services to copy files from one place to another, but that failed because my AWS account is somehow limited to a low amount of memory for AWS Lambda functions, and the CloudFormation stack’s Lambda functions somehow need to exceed it.

I opened a ticket with AWS support and asked them to increase my memory quota, and they replied and said I can do it myself, which I can’t because that quota is special and doesn’t show up on the self-service quota dashboard.

At this point the whole enterprise is starting to feel a bit Gyfordian so I should probably cut my losses, but, as is fitting, the degree of pointless inconvenience has galvanised my determination to complete the task regardless of whether it’s necessary or even desirable. One day someone’s going to watch my shit 2009 talk about vector spaces whether they like it or not.

I really don’t want to give these people the oxygen of attention, but I must briefly note that Sam Altman justified the energy consumption of ChatGPT by saying “it also takes a lot of energy to train a human” and “it takes like 20 years of life and all of the food you eat during that time before you get smart”, revealing that the entire purpose of human experience is to produce economic output by answering inane questions.

If you’re the kind of chuckleheaded ghoul who claps in glee at this absolute dogshit, I hope you’re being honest with yourself about the future you’re implicitly promoting: a ruined planet owned and operated entirely by people who think like this guy. Don’t forget to tell your kids what you did.

While riding my bike along Whiston Road I saw crocuses peeking out in Haggerston Park and a weight was lifted.

Also, I baked some bread.