✧✦Catherine✦✧

sorceress

#searchable

Public Key

npub1mh0qfzf7tna7wfk0mjrzqy542mxj5qzwtl6x0menkuhypld20n0savgga4

NIP-05 Address

whitequark@social-treehouse-systems.mostr.pub

Profile Code

nprofile1qqsdmhsy3yl9e7l8ym8aep3qz224dnf2qp89lar8auemwtjqlk48ehcpz4mhxue69uhhyetvv9ujuerfw36x7tnsw43qudz3l8

Publishing to

relay.ditto.pub

Author Public Key

npub1mh0qfzf7tna7wfk0mjrzqy542mxj5qzwtl6x0menkuhypld20n0savgga4

Show more details

Last Notes

2026-05-12 20:30:55 UTC

aw, recent pybind11 commits are like, nearly 100% llm stuff. i'm not exaggerating, i've never seen so many in a row before

2026-04-24 19:45:58 UTC

can someone explain to me why server motherboards are so fucking awful (it could also just be supermicro. i've had two server motherboards and both were from supermicro)
the cheapest, most cost-reduced, aliexpress tier desktop motherboard easily outperforms both of those supermicro boards, combined, when it comes to things like "fucking booting like a normal PC"

2026-04-18 11:07:42 UTC

my code has started (and ended) romantic relationships, changed how people view the world around them, and brought people bits of otherwise unachievable joy
it is embarrassing to tell on yourself that you can't do the same, but it is downright ignorant to claim it cannot or shouldn't be done at all
https://nondeterministic.computer/@mjg59/116424721686619312

2026-04-17 11:10:18 UTC

https://www.ietf.org/archive/id/draft-meow-mrrp-00.html

2026-04-13 10:08:56 UTC

hey. you know how every code forge in existence offers you two types of downloads: tar and zip?
i wonder if you can make that into one file that is both tar and zip.

2026-04-08 15:46:56 UTC

Where selected by your Test Sponsor, you agree that Pearson VUE will collect your palm vein pattern at the test centre on the day of your exam and retain that information, to the extent permitted by law.
what the hell

2026-04-05 17:17:05 UTC

- reply

@nprofile…r9vg right, yes, I got that. it's the "somehow" bit that makes me go "wtf"

2026-04-05 16:27:06 UTC

okay so what's going on with office software
the OnlyOffice mess is garden variety xenophobia that people who know better on here are amplifying, that's easy
what the fuck is going on with LibreOffice and Collabora

2026-04-04 23:01:12 UTC

IETF KITTEN WG?
https://cache.treehouse.systems/media_attachments/files/116/348/922/399/505/838/original/a5b4235dfe4017ea.png

2026-04-01 20:45:31 UTC

now, do you really want to remind soon to be citizens of this?
https://cache.treehouse.systems/media_attachments/files/116/331/402/290/203/354/original/87cafc58325a70cc.png

2026-04-01 20:28:06 UTC

- reply

what kind of test autofellatio is this
https://cache.treehouse.systems/media_attachments/files/116/331/333/619/502/603/original/77deae7489b9f1ff.png

2026-04-01 20:13:15 UTC

motherfucker why is this on the citizenship test
name one single reason
https://cache.treehouse.systems/media_attachments/files/116/331/275/492/679/991/original/db35a058f746f481.png

2026-04-01 19:48:36 UTC

- reply

@nprofile…yzy8 what

2026-04-01 16:44:15 UTC

Tom Scott is absolutely awestruck by the concept of "molten metal" and I love it. just minute after minute of exposition that boils down to "metal liquid? liquid metal hot??" https://www.youtube.com/watch?v=GpaNijzRaJI

2026-04-01 15:06:24 UTC

- reply

@nprofile…mv7s a DRAM controller?!

2026-04-01 13:48:35 UTC

as a toolmaker, there's an inherent tradeoff that I encountered years ago when I just started working at ChipFlow; what I was asked was essentially to develop Amaranth further as a way to de-skill the hardware design (RTL) field. I agreed because I don't really value the skill of knowing every one of the five hundred different ways in which SystemVerilog is out to fuck you over; I think we'd be better off with tooling that doesn't require you to spend years developing this skill, and that would be a lot more friendly to new RTL developers, and people for whom RTL isn't the primary area of work.
I also knew that ChipFlow was on the lookout for opportunities to shoehorn AI somewhere into the process. (at first this was limited to "test case generation"—frankly ill conceived idea but one I could hold my nose at and accept—nowadays they've laid off everyone and went all-Claude.) however, it was clear pretty early on that making hardware development more accessible to new people inherently means making it more accessible to new wielders of the wrong machine. benefiting everyone (who isn't a committed SystemVerilog developer) means benefitting everyone, right?
you can trace this trend in adjacent communities as well. Rust and TypeScript have rich type systems that generally help you write correct code—or bullshit your way towards something that looks more or less correct. I'm pretty sure it's a part of the reason Microsoft spent so much money on TypeScript.
so today I find myself between a rock and a hard place: every incremental improvement in tooling that I build that makes the field more accessible to new people also means there's less of a barrier to people who just want to extract value from it, squeezing it like Juicero (quite poorly but with an aggressively insulting amount of money behind it). so what do I do now?..

2026-03-31 13:00:54 UTC

happy trans day of vengeance 🏳️‍⚧️

2026-03-31 08:09:26 UTC

- reply

even if you make every machine single-tenant, most configurations of Forgejo Actions runners would enable malware to escape the build container, persist itself and infect all future releases

2026-03-31 07:52:23 UTC

the expectation of being able to run docker whenever in CI jobs is probably the single worst outcome of free GitHub Actions minutes because reproducing it in a bring-your-own-compute environment is borderline impossible unless you make every machine single-tenant

2026-03-28 19:35:25 UTC

- reply

@nprofile…yzy8 thats an analog PIR sensor, it's sort of like a whetstone bridge but made out of photodiodes (to crudely simplify)

2026-03-28 18:30:07 UTC

- reply

@nprofile…tpfq oh no I tried to give it 12V from a power supply for funsies. (it just sits there consuming no power)

2026-03-28 18:25:54 UTC

- reply

@nprofile…tpfq what does it need to start booting?.. does the damn thing has a power switch?

2026-03-28 18:00:49 UTC

so i took apart an NVIDIA BF-3 card and would anybody who understands it please enlighten me why it has on board:
Puya P24CM02 (256 KB)
Winbond W25Q256JVBM (32 MB)
Winbond W25Q512JVBQ (64 MB)
MXIC MX66L2G45G (256 MB)
a Kingston EMMC128-TY29 (128 GB)
a Microchip CEC1736 root-of-trust controller (which embeds 2 or 4 MB of internal SPI flash and is connected to the other three SPI flashes... I think)
and an ASPEED chip with a p/n I did not extract from under thermal compound looking suspiciously similar to those awful BMC devices, connected to 1 GB of Samsung DDR4
I wasn't prepared for a 1 Tbit eMMC connected directly to the main ASIC. what could it possibly be storing there? the field engineer's porn stash?

2026-03-28 08:20:53 UTC

- reply

@nprofile…v2v2 yeah you're on point

2026-03-27 23:02:01 UTC

people: ask their dependencies to follow semver, for fuck's sake already
also people: make a surprised pikachu face when the major version is incremented with every release
(i have been both, at times. this is about me. this is also about others who i've seen be a lot more militant about this issue)
the thing is, if you have a sufficiently complicated application it is not feasible to determine what is a "breaking change" or not. this complexity limit kicks in long before you get to a "browser" or a "JIT compiler" but it is definitely well applicable by that point
i think what people mean when they do both of those things are a mix of "please stop adding features entirely. only fix bugs" and "please only make changes i like, but not the changes i dislike" depending on maturity level. that's not really how open source software works though

2026-03-27 22:19:49 UTC

- reply

@nprofile…mv7s GIS?

2026-03-27 01:44:07 UTC

okay! check out this cutie. this is the new CI/build server that i've been assembling for the last month (hard drive for scale, there are no hard drives inside)
Chinese X99 motherboard (excellent)
20-core Xeon E5-2630
64 GB RAM + 32 GB swap on Optane
2.5 Gbit ethernet
3840 GB of NVMe cache
all in a tiny and more importantly quiet envelope
https://cache.treehouse.systems/media_attachments/files/116/298/601/011/917/303/original/22eb1b8dd8e664cc.jpg

2026-03-26 20:46:12 UTC

please enjoy this piece of modern art i've just constructed
https://cache.treehouse.systems/media_attachments/files/116/297/428/147/303/620/original/55894f8d42f592ef.jpg

2026-03-25 21:25:20 UTC

huh apparently there's a reimplementation of the discord server software https://docs.spacebar.chat/

2026-03-25 20:52:14 UTC

calling the language ortho-caml to piss off literally everyone except like One chemist using the language

2026-03-24 08:33:04 UTC

- reply

@nprofile…ds0l it is incredibly brazen and apparently the test used to be more in line with international practice but Davey Cameron (may he get fucked by the pig for the entire afterlife) considered it a big achievement to have really turned up the racism dial

2026-03-24 07:05:41 UTC

- reply

you can't be fucking serious
https://cache.treehouse.systems/media_attachments/files/116/282/880/197/744/339/original/3d484d5c80e5e204.png

2026-03-24 06:56:13 UTC

why do i need to know this for settlement
https://cache.treehouse.systems/media_attachments/files/116/282/842/769/013/964/original/208f54af85f32501.png

2026-03-24 06:45:22 UTC

- reply

@nprofile…mv7s yes this is how I've implemented this type of preference before

2026-03-24 06:25:30 UTC

- reply

political satire isn't supposed to be timeless!

2026-03-24 06:25:15 UTC

this scene was shot back when "West Berlin" was still an existing entity and basically nothing has changed in the 30 years since. normally i would praise timeless fiction but in this case it's just depressing https://www.youtube.com/watch?v=IX_d_vMKswE

2026-03-24 04:24:10 UTC

i just found out that semantically, javascript arrays are not special and so x[123] and x['123'] (as well as a syntactically invalid but theoretically sensible x.123) are the exact same thing
that's upsetting. i know engines don't actually convert indices to strings, but it is still upsetting

2026-03-24 02:59:20 UTC

honestly ircv3 might just make me change my mind on whether multi-client coordination for instant messengers can work well in practice (after xmpp and matrix being very visible examples of how it doesn't)

2026-03-24 02:30:12 UTC

irc has typing indicators now
https://cache.treehouse.systems/media_attachments/files/116/281/795/781/945/332/original/a38e56297a92388f.png

2026-03-23 23:08:56 UTC

- reply

this isn't satire, this is real research published by IEEE/ACM

2026-03-23 23:03:39 UTC

i'm at a loss of words after reading a paper about reformatting code using an ML model that has a measured statistical quantity A_c which says how often the reformatted code behaves the same as the original
the "ideal" (their choice of words) case is 64.2%

2026-03-23 21:03:05 UTC

- reply

@nprofile…yzy8 which p/n?

2026-03-23 19:55:03 UTC

intel, who hurt you to make this typography possible
https://cache.treehouse.systems/media_attachments/files/116/280/243/117/244/528/original/e2c680f282c2c6b4.png

2026-03-23 19:07:05 UTC

💭 the difference between "AI" and "ML" is whether it primarily benefits capital or labor

2026-03-22 21:28:42 UTC

- reply

@nprofile…h6ty @nprofile…0x09 principles don't exist as platonic ideals; they are solely the manifestation of the collective will of their practitioners
e.g. c is a bad language because the collective of its practitioners is unable to use it correctly. same shit. arguing about platonic principles outside of practice is a waste of text

2026-03-22 21:25:50 UTC

- reply

@nprofile…v2v2 i'm ambivalent about it. fossilized stuff like x11 has been subject to fascist takeover (enabled by the things you list). doesn't seem like a universally good outcome
(this is just the "how replaceable should labor be" discussion all over again, but less corporate now)

2026-03-22 21:13:20 UTC

- reply

@nprofile…h6ty @nprofile…0x09 if the most widely deployed unix doesn't follow the unix philosophy then it ain't very good is it mate

2026-03-19 23:50:02 UTC

the Big Short (a long and very boring film about PCB assembly rework due to an incorrectly connected power pin)

2026-03-19 19:45:50 UTC

anyway if you want a python dependency manager not owned by openai, try pdm

2026-03-09 23:48:32 UTC

- reply

@nprofile…8d0p are you talking with policymakers about this?

2026-03-09 06:08:08 UTC

- reply

@nprofile…8m52 @nprofile…9xu8 i don't think a generalized notion of intelligence is meaningful

2026-03-09 06:06:10 UTC

- reply

@nprofile…9xu8 in a sense, this is a superhuman intelligence that escaped its sandbox. it just didn't do this because it was smart. it did this because humans were greedy

2026-03-08 00:49:19 UTC

- reply

@nprofile…ffys @nprofile…mv7s this finally prompted me to buy cheap UV glasses for curing resin. this tiny LED can't have more than a few W of input power but still

2026-03-08 00:23:29 UTC

current status
https://cache.treehouse.systems/media_attachments/files/116/190/701/397/052/199/original/647f7b0f0bb3b3db.png

2026-03-07 22:22:18 UTC

i found a new semiconductor vendor unit that is straight out of @nprofile…kjet
"Standard logic longevity based on package pin count"
https://www.nexperia.com/support/customer-support/longevity
https://cache.treehouse.systems/media_attachments/files/116/190/224/689/505/065/original/c17fb9e1164a7724.png

2026-03-05 19:51:52 UTC

me: makes a harmless, pleasant pun
her: "Why are you like this?! What God-forsaken neuron miswiring are you suffering from to emit these words in this order?!!"

2026-03-05 00:09:44 UTC

i've been seeing "A COMPUTER CAN NEVER BE HELD ACCOUNTABLE / THEREFORE A COMPUTER MUST NEVER MAKE A MANAGEMENT DECISION" image a lot lately and i've come to queston it
sure, a computer cannot be held accountable. but a computer can have properties that are much more valuable than "can be held accountable" (which has happened so little with corporate management that i'm not sure it's even a relevant question anymore): it can be verifiably transparent, and it can be modified in a predictable manner.
try doing that with a CEO's brain! (i mean, i would try, no claims about success though)
now of course i say "can" because the most recent movement in computing is to completely negate those two important properties

2026-03-04 20:42:33 UTC

codeberg is down, post ponies

2026-03-04 18:40:54 UTC

what fresh hell is "PaymentDefermentRisk", DHL
https://cache.treehouse.systems/media_attachments/files/116/172/367/514/405/507/original/bed41f0820eeead3.png

2026-03-04 01:16:28 UTC

- reply

@nprofile…mv7s do you ship flatpaks or snaps?

2026-03-03 02:47:01 UTC

- reply

@nprofile…yzy8 seems kinda normal

2026-03-03 00:05:46 UTC

DUT (device under titties)

2026-03-02 15:23:58 UTC

pov: you have kvm at home
https://cache.treehouse.systems/media_attachments/files/116/160/266/668/335/009/original/86dc3550f40403ea.png

2026-03-02 02:41:51 UTC

hot singles in your area! (measuring over 100 uSv/hr)

2026-03-01 19:02:18 UTC

- reply

@nprofile…mv7s this is somewhat cursed but you can use nginx to do it. think of it as doing the inverse of the usual "reverse proxy" setup (so "reverse reverse proxy" I guess?): set up a server for "github.com" and location for "/user/repo" then proxy_pass it to actual github.com
obviously needs your cert to be trusted; and HPKP doesn't exist anymore so I don't think anything will complain that doesn't hardcode GitHub's TLS cert

2026-03-01 18:57:32 UTC

- reply

@nprofile…mv7s HTTP proxy?

2026-03-01 17:28:33 UTC

- reply

@nprofile…g0e5 me

2026-03-01 11:30:20 UTC

- reply

@nprofile…v2v2
https://cache.treehouse.systems/media_attachments/files/116/153/687/695/920/357/original/6b315c61a5f82576.jpg

2026-03-01 00:36:16 UTC

- reply

@nprofile…mv7s @nprofile…klv7 ohhhhhh.

2026-02-28 19:10:51 UTC

- reply

@nprofile…mv7s yep! Samsung got up to 400 layers last year

2026-02-28 18:21:00 UTC

hey do you wanna see a ludicrously high resolution image of a microSD card i'm deprocessing? enjoy
https://upload.whitequark.org/1772302728-20260228_175143.JPG

2026-02-28 17:14:12 UTC

- reply

if you think that the precariously balanced tower of flash above is an inaccurate representation of what's going on, i can reassure you that
(a) this is exactly how it looks like, and
(b) SK Hynix is up to 24 layers already (not sure in what application)
https://news.skhynix.com/semiconductor-back-end-process-episode-4-packages-part-2/
https://cache.treehouse.systems/media_attachments/files/116/149/376/322/617/263/original/5ea56180f88bc28e.png

2026-02-28 17:09:29 UTC

today I learned that we (as a civilization) now can put 16 silicon flash dice into a single 700µm microSD card on a commercial scale https://www.ose.com.tw/en/ic-services/memory-product-package/microsdsd
https://cache.treehouse.systems/media_attachments/files/116/149/358/453/020/449/original/a7d0782a7aeeafa4.png

2026-02-28 12:58:44 UTC

I would like to formally announce that the #GlasgowInterfaceExplorer revD development (the next iteration: with the same FPGA but 32 digital I/Os and 8 single-ended analog channels that can be configured as 4 differential ones) is well underway with a CrowdSupply campaign planned later this year; stay tuned!

2026-02-28 12:55:13 UTC

pov: you wake up to receive a package with a vacuum chamber and discover that the vendor is a crackhead and a menace to society
this is a vacuum chamber. the power supply that comes with it has a type-C connector and it puts 12V on Vbus unconditionally. the pump works off both 5V and 12V, so it can pull vacuum off a normal type-C power supply. but the solenoid release valve (that lets you open the chamber once you're done) requires their illegal type-C-shaped power supply.
to add insult to injury, it's using a NEMA 1-15P plug.
https://cache.treehouse.systems/media_attachments/files/116/148/352/557/585/322/original/b913a2942610ffb0.jpg

2026-02-28 01:08:48 UTC

the SI people are right and everybody doing it otherwise is wrong; you should never use the "K", "M", etc unit prefixes to mean factor of 1024. sure, with small enough sizes you can usually ignore it. but scale it up and you quickly run into trouble

2026-02-28 00:42:06 UTC

- reply

@nprofile…yzy8 give me a bit

2026-02-28 00:41:05 UTC

- reply

@nprofile…yzy8 hm. etching things into sapphire is probably as far as we can get towards "storage that isn't decaying"...

2026-02-27 15:04:52 UTC

heyyyyyy. check this out
i bought one of those chinese motherboards which get the UEFI package from American Megatrends and then enable options with the guiding principle of "YES."
check out how many juicy bits it has
you can turn the memory scrambler on and off! it even tells you the seed it has on, i think, this specific boot?
https://cache.treehouse.systems/media_attachments/files/116/143/199/339/070/253/original/a765655decfa93ba.jpg
https://cache.treehouse.systems/media_attachments/files/116/143/199/357/325/143/original/dd49047b8dda8875.jpg
https://cache.treehouse.systems/media_attachments/files/116/143/199/636/079/448/original/4269af315bb2b24e.jpg
https://cache.treehouse.systems/media_attachments/files/116/143/200/168/747/946/original/76dd2c6a26d5e00d.jpg

2026-02-26 12:28:34 UTC

i checked up on OsmAnd and apparently they're doing some pretty advanced stuff for offline route calculation now! beyond A* https://osmand.net/blog/fast-routing/

2026-02-26 09:48:56 UTC

firefox's on-device translation feature seems to work ok

2026-02-25 18:52:36 UTC

i've been doing a deep dive into SPI NOR flash command set and it is horrifying. i don't know how anything works. none of these devices agree with each other on anything, even devices from the same family of the same vendor with the "same" part number (but suffix J instead of suffix V or whatever)

2026-02-24 21:51:17 UTC

- reply

@nprofile…mv7s laughed out loud

2026-02-24 14:30:37 UTC

i wonder how much consumer demand is driven entirely by "I've gotta buy [thing] before AI outcompetes it out of existence via the supply chains"

2026-02-24 09:45:10 UTC

- reply

@nprofile…v2v2 I had very good experiences with the Zoom app (and it had been a critical piece of my life for like a year), mainly because nobody else has screen-sharing that works well enough for videogames

2026-02-24 04:59:13 UTC

8086 instruction set implemented in pure css, amazing https://lyra.horse/x86css/

2026-02-22 19:55:04 UTC

prompted by a discussion about naming things elsewhere
https://cache.treehouse.systems/media_attachments/files/116/116/030/171/601/426/original/8b5d599b76c4b88c.png

2026-02-22 17:22:28 UTC

- reply

@nprofile…v2v2 please be aware that Gandi got bought out by private equity which did the thing everybody expects it to do (jacked up prices by at least 2x compared to equivalent registrars, and made services worse)

2026-02-22 15:29:05 UTC

today's project for #GlasgowInterfaceExplorer: reworking the memory-25x applet
this is a very old applet, one of the first introduced. it works mostly fine, but it was designed and implemented before I really knew how to make good applet UIs, and long before I added QSPI support, so it's not using all of the available bandwidth & it requires specifying a bunch of geometry parameters by hand even when it is trivially extracted from the flash itself
I am planning to redesign the Flash command engine to support not just 1-1-1 (normal SPI) commands, but every mode including 1-1-2 and 1-1-4 (dual/quad data) and 1-2-2 and 1-4-4 (dual/quad address and data). to do this I will implement an abstraction for a (Q)SPI command set, which can then be extended and specialized by an individual flash
this will also handle devices like Winbond's and Micron's QSPI NAND devices (W25N series and MT29F series, respectively).

2026-02-22 11:07:26 UTC

- reply

i bought five of them to decap and analyze!

2026-02-22 10:06:49 UTC

- reply

@nprofile…v2v2 @nprofile…u57k I think (I am very much not an expert) that NAND flash suffers from, basically, rowhammer and everybody accepts it because that's just one of the things you use ECC for. I did find it that if I write a page a few times withour erasing in the middle it can kill the entire eraseblock, as in the LUN reports erase failures on it afterwards! I wonder what the physics is for that.
did you know that on leading-edge non-3D MLC NAND (a few years ago; people don't really make such flash anymore if they can help it), a single gate could hold as few as 80 individual electrons? this is bonkers to me

2026-02-22 09:52:40 UTC

- reply

@nprofile…v2v2 eraseblocks are generally isolated from each other; pages are generally not isolated. there is a million nuances I'm sure a NAND vendor could correct me on, but generally they all say or imply that (a) page writes can disturb adjacent pages and (b) if you find one bad page, condemn the entire eraseblock. subpages must always be written as 1-2-3-4 from what I've seen. (I wrote a simple FTL recently)
I did see some weird behavior on a QSPI NAND chip where every other page was failing with recoverable ECC errors on most blocks before I wrote and erased most of the chip a few times. no idea how that happened. something to do with the physical layout? (it's a single LUN... I think, did not decap it)

2026-02-22 09:42:36 UTC

- reply

@nprofile…v2v2 I'm very well aware yeah! I want to get a 5-pack of those and read out the ONFI NAND info from the testpoints

2026-02-22 09:23:28 UTC

apparently on aliexpress you can get specifically a small capacity microsd card. we talking 64MB to 512MB product lineup. for the same price as 1000x bigger ones lol
this is actually something i would want, one could hope they use SLC or pseudo SLC NAND there...

2026-02-21 17:17:35 UTC

- reply

it's 2 years newer than Xbox 360 and that's definitely retro, we all agree on that, right

2026-02-21 17:14:43 UTC

Cortex-M3 is a retro CPU

2026-02-21 13:02:53 UTC

3d printer havers: do you routinely correct for thermal expansion when manufacturing your designs? why?

2026-02-21 09:07:00 UTC

- reply

@nprofile…v2v2 that was slightly before when I had any capacity to understand computer architecture, but yes, I do remember the claim

2026-02-21 06:38:55 UTC

nitrogen blast from the past https://www.youtube.com/watch?v=z0jQZxH7NgM

2026-02-21 04:15:14 UTC

(archive.today) I think DDoS is essentially a reasonable response to someone doxing you masquerading as a "journalist", but modifying snapshots of webpages clearly crosses a line