look, okay, these things happen sometimes

Kali's Tumblr | They/Them ’nouns Roughly ¼ original content by weight Extended Post Writer, Engineer, Terrible Computer Nerd, Infrequent Embroiderer. Kalium on Slashnet IRC, kilovoltamp on hachyderm, kva on cohost. Ask box usually open.

[andmaybegayer]

Penrose quilt is officially finished! 2022-07-30 to 2023-04-02, all hand sewn by me and my mother.

It would figure that the Einstein would be found like a month before I finished this! Honestly probably for the best, the Einstein has too many concave angles.

We'll embroider patches for it soon but for now we're calling it, it still needs its first wash and I'll probably have to patch up some quilting after that.

[theoriginalsupermodels]

Balenciaga - Spring 2002 RTW

[memories-of-ancients]

Anthropeum

A daily game that challenges our understanding of human cultures. Ten objects. 5,000 years of human history. Guess where and when each artif

An interesting game where you are presented with 10 artifacts from the MET. You have to place where the artifact is from and what time period it is from. Each artifact scores up to 10,000 points, and you lose points the further away your guess is and how far off in time you are. You can only play once a day. Thanks to @baebeylik for showing this to me.

Today I scored really well. Yesterday ... not so much.

Anthropeum.com · Jun 8 2026 🟩🟦🟦🟩🟩🟩🟥🟦🟦🟩 79,001 · top 3% of players today!

[motherfuckingnazgul]

oh this is extremely fun. i did NOT do all that well but i can see myself getting good. i will be doing this regularly.

Anthropeum.com · Jun 8 2026 🟩🟦🟦🟨🟨🟦🟥🟩🟩🟦 68,088 · top 12% of players today!

[andmaybegayer]

One City

London, United Kingdom

[memories-of-ancients]

Anthropeum

A daily game that challenges our understanding of human cultures. Ten objects. 5,000 years of human history. Guess where and when each artif

An interesting game where you are presented with 10 artifacts from the MET. You have to place where the artifact is from and what time period it is from. Each artifact scores up to 10,000 points, and you lose points the further away your guess is and how far off in time you are. You can only play once a day. Thanks to @baebeylik for showing this to me.

Today I scored really well. Yesterday ... not so much.

Anthropeum.com · Jun 8 2026 🟩🟦🟦🟩🟩🟩🟥🟦🟦🟩 79,001 · top 3% of players today!

[motherfuckingnazgul]

oh this is extremely fun. i did NOT do all that well but i can see myself getting good. i will be doing this regularly.

Anthropeum.com · Jun 8 2026 🟩🟦🟦🟨🟨🟦🟥🟩🟩🟦 68,088 · top 12% of players today!

[andmaybegayer]

gaggle of french sales staff visiting the office bewildered by the concept of the "unisex toilet"???

[defendglobe]

project hail mary but grace's translation program uses that awful fucking default tiktok TTS voice for rocky instead.

[andmaybegayer]

watching four people who all haven't played or even looked closely at Disco Elysium talk about why Disco Elysium is bad

(better yet: one person who has played Disco Elysium telling them it's good, and them going, but I heard it's bad.)

[andmaybegayer]

One City

London, United Kingdom

[andmaybegayer]

Last Monday of the Week 2026-06-08

it'll happen any day now

Listening: New Modest Mouse just dropped:

Modest Mouse changes a lot over the years but it somehow maintains a kind of vital essence. He's a very wet type of musician.

Also because I bought this because I was like hang on this has several of my favourite songs: Aims by Vienna Teng, which has the GOAT Landsailor

I love the feeling of like. A down home city song? Everyone wants to be from the countryside or from the small town but this is like no! I'm from the cities! I am of an Industrial Society! I would have died several times over without the wonders of the modern world.

Watching: Fatal Deviation at Bad Movie Night, a 1998 martial arts movie out of Ireland made for zero dollars by some Van Damme fanboys

This movie is dogshit in an endearing way. Incredible small town Ireland vibes. The back half has some "stunts" that feel like they just got a nonexistent wrestling promo together.

Also getting my partner to watch Arcane with me, just cleared e3, which. They're like "It's incredible! I'm so sad!" which is my favourite part of Arcane.

Reading: Ancillary Justice by Ann Leckie, which I am like halfway through. I am having a great time with this. I love having most of my characters be objectively terrible people. War Criminal central. The deeply integrated class structure of Radchaai society is so juicy.

It has a gentle hand on the worldbuilding which I appreciate and I hope it keeps it up through the other books. There are "bov" but there are also just tamarind.

Making: Not much, tinkering with some software projects. Ordered the adapter I need to talk to my servomotor so hopefully project Sewing Machine will be back soon.

Playing: Doing the Forza Horzion 6 weeklies. I think they have definitely amped up the AI difficulty in this one? Highly Skilled feels much more like a decent challenge, I can beat it almost all the time but it takes some pretty precise driving and if you fall behind you often just can't make it up if you aren't driving dirty.

Tools and Equipment: I was having this discussion with a coworker and I will now convey it to you: have you heard the good news? Frozen food is good now!

Frozen pretty much everything is incredibly high quality if the cold chain is maintained and it is a great way to get things that are both very high quality and with good shelf life. I basically only buy peas, spinach and corn frozen unless I really need the fresh stuff for something. If you eat meat, finding a good source of frozen meat is a good idea, I tracked down a local supplier recently that I want to try.

If you are turned off frozen food because you had bad experiences with it 20-30 years ago, we've gotten much better at it! For one thing, most of the meat and fish you see in a grocery store will have been frozen anyway, shortly after packing, and it's been defrosted for you. You may as well cut out the middleman and just get it frozen.

[andmaybegayer]

Radchaai has a base-20 counting system, right?

[andmaybegayer]

@memoryisaneventhorizon said:

because a decade is made of of 20 soldiers? that’s super interesting, i never picked up on that!

There's a lot of twenties in this book, although some of that is just that a major event happened twenty years prior. People say "almost twenty" a lot. Decades have 20 lieutenants, and One Esk consists of 20 ancillaries, not sure about the other ancillary units, they aren't mentioned as much. People use "twenty" as a round number, I think you see more "tens" when someone is speaking a non-Radchaai language. There was a particularly weird one that caught my eye here:

For the next twenty to forty seconds we had nothing to do but wait, and fall.

twenty to forty seconds? that's not very naturalistic, unless you're in vigesimal!

[buttercupfestival]

Plish plash: Buttercup Festival 3-384

www.buttercupfestival.com

[regexkind]

Postmortem: algol-0001

Hypothesized Incident Onset: 2026-06-08 03:28AM PST

Oncall Ownership Onset: 2026-06-08 05:45AM PST

Amelioration Onset: 2026-06-08 05:50AM PST

Narrative: at 2026-06-08 3:28AM PST, the tumblr-rotator daemon running on minicomputer-2 attempted to POST the 208th avatar of the day to the Tumblr avatar upload endpoint.

During nominal functioning of the daemon, after every upload attempt, we fetch (without intermediary sleep) the avatar from Tumblr's servers. In order to verify the fidelity of Tumblr's cache invalidation (and therefore their ability to faithfully broadcast avatars sent to them by the tumblr-rotator daemon), we use a perceptual hashing algorithm (ImageHash) in order to check that they are not serving a stale or otherwise incorrect image. This threshold is tuned by-hand to accept uploads where the pHash is less than 25 according to the snippet "abs(imagehash.phash(src, hash_size=16) - imagehash.phash(srv, hash_size=16))".

The perceptual hash algorithm has previously ensured that we re-attempt uploads as-needed; Tumblr has been known to among other things reset the regexkind avatar to a days-old version of the regexkind avatar. When a mismatch is detected, an upload is reattempted for a maximum of 5 times (including the initial attempt), with a fixed sleep of 5 seconds between each upload¹.

After 5 attempts elapse, the method run_avatar_stage surrenders control to the scheduler, which under nominal conditions will sleep to allow 60 seconds to pass between each successful avatar push, giving viewers adequate time with each avatar. After that delay passes, the last unpushed image is scheduled for upload. This means that the retry-logic in run_avatar_stage is somewhat misleading; in actuality, an image will be attempted indefinitely until one of two things happens:

A successful upload occurs,

We forfeit² the day

The exact sequence of images for 2026-06-08 had been pushed on previous days, but not since the update to this new scheduler algorithm; previous versions of the scheduler were more willing to allow "transmission gaps" where an image never made it to the viewer, which would require them to interpolate said image. It is assumed that on these previous days, the scheduler also saw a pHash distance of 26 (as it did today during the upload of avatar #208), attempted 5 times, and then gave up transmitting; due to a gap in our operating procedure, such events would not have been adequately investigated.

At 2026-06-08 05:45AM PST, the oncaller for the service saw the (silent³) notification on her phone. Noticing that the lag metric had reached 8260s, she immediately investigated the logs, which revealed that attempts had been steadily made to upload every 30s for the past few hours, but had "failed" due to the pHash value being 26 > 25. She observed visually that the current image at the avatar GET endpoint was the 208th image, implying that this hairsplit was a false positive for an upload mismatch. She tuned the threshold by-hand to 27, reasoning that the precise degree of acceptable pHash distance could be computed at a later time, so long as she unblocked the rotator.

She then ran "systemctl restart tumblr-rotator.service" as root to reload the rotator scripts, and the upload process immediately resumed and completed a successful reupload of the 208th avatar at 2026-06-08 05:50:58 PST, followed by a successful first upload of the 209th avatar at 2026-06-08 05:51:27 PST.

The following graphs illustrate the characteristics of the incident:

We anticipate the lag dropping to sub-minute intervals within two hours of this report, after which we will consider incident algol-0001 to be mitigated if not resolved.

Proposed fix(es):

The pHash algorithm is probably appropriate for detecting the difference between glyphs but is inappropriate for detecting other important features of the avatar, including the RGB value and [REDACTED]. Since RGB value and [REDACTED] can be more faithfully measured by other methods, we should ensure that the pHash algorithm is only used to measure the difference between glyphs, leading to a composite diff-measuring algorithm that is simultaneously more accurate at detecting differences between served and uploaded avatars, and less prone to false positives.

As mentioned in this report, this issue likely has occurred before and not been surfaced due to a presumption of fault on the side of Tumblr API and CDN. Since the number of faulty uploads is normally less than a dozen per day, and we expect that the oncaller will be able to review data from the daemon at least once every few days, it seems reasonable to snapshot each offending diff and compile them into a running "report" for visual inspection.

The first of these fixes is considered accepted, pending implementation and the second of these fixes is rejected due to a strong organizational value on never imposing toil except where absolutely required. Forcing our oncaller to view a report full of false positives every day does not align with our values.

¹Incidentally, this is not considered to be sound engineering practice--exponential backoff with jitter being the gold standard--but the author concurs with the writer of the script that this is "good enough" in a circumstance where we are likely the only client with the exact interaction characteristics of tumblr-rotator.

²The forfeit conditions are somewhat complex, but involve the scheduler computing that even if we uploaded an image at the maximum rate (30s between uploads) from now until midnight, we could not push every image in the schedule.

³All notifications for the tumblr-rotator service are silent; we cannot afford to pay our oncaller for the level of service implied by a paging alert system that can wake her up.

[regexkind]

The problem with that is that it's named after the demon star Algol and not the programming language ALGOL

[phaeton-flier]

(lens warping effect) in an efficient market, arbitrage opportunities like me would not exist. but this is not an efficient market

[andmaybegayer]

computers give me a kind of unreasonable expectation of how much you should be able to get done with a certain amount of power. The fan I use to circulate air around my living room pulls like 35W! That's like a whole laptop going full tilt!

[nyantara]

was looking up how long people in academia have hated pdfs and was genuinely surprised by the strident tone of this one paper

Get rid of PDF as the publication format. This is likely the only long-term solution. PDF is an objectively terrible format, designed for printing and not for reading, and most of the problems we have with making papers accessible are either due directly to it being a terrible format or are made worse because of it.

It treats the document markup and text as second-class citizens compared to visual layout, which implies that it will never be a great format for people with visual impairments or anyone who needs custom styling. Switching to another document format that preserves text order and allows the user to easily change font size and type would make this problem simple.

For historic reasons this is difficult, as our community is very familiar with distributing PDFs. Researchers have already spent decades making other document publishing formats accessible by default, most notably websites. Using HTML and CSS to format academic publications is a great way to solve the problem. PubCSS could be used to render the document in a way similar to the current proceedings template, if we want articles to still look like the printed page.

Another approach would be to use a document converter like Pandoc to produce a traditional PDF from a better markup language, e.g., using HTML and CSS or plain text. Using a document format that screen readers can access implies that it is easily machine-readable, which also benefits archival and analysis of our field. This is something the community seems to be interested in, as CHI recently changed the reference format partially to make it easier for machines to recognize papers and authors. [...]

It is sometimes suggested that an accessible version of the paper be supplied along with the PDF. In practice, this does not work because it is likely that the PDF will be shared, but not the supplementary information.

An Uninteresting Tour Through Why Our Research Papers Aren't Accessible

Funnily enough the only source is a PDF

[shocked--kwarrtz]

It's funny how the transformer architecture was originally invented for translation, but auto-regressive models have so thoroughly overtaken that as an application that even if your goal is to translate something the best approach is often to just ask a chatbot to do it.

I mean it makes a certain amount of sense. General-purpose LLMs have much more complete world models than a translation-specific model ever could, and that's often necessary to understand the full context of a translation request. I remember pre-2017, a talking point I heard a few times was that "any truly effective MTL system is necessarily AGI because understanding language well enough to translate it accurately also implies understanding all the things language can describe." It turns out they were right, in a funny sort of way; what they failed to expect is that we wanted good MTL so badly that it became the route through which AGI was accidentally invented!