re: chatgpt is bad at software
@noracodes a fun question is whether it’s entirely hallucinating (coming up with vague but convincing-sounding bullshit using the words it is given), or it has enough capacity in its parameter space and enough examples in the training set to encode the connection between the two fields (affine types for borrow checking vs linear types to encode unitary operators in functional quantum programming languages) to some degree and thus is more likely to come up with this specific sort of bullshit
nevertheless, it’s pretty useless without any form of logical grounding, to say the least
re: space, the final frontier
Q: When is it okay for a program to reach out to the internet?
A: In response to an explicit user request for it to do so.
Q: How much data should be included in network requests?
A: The minimum necessary.
hot take, telemetry
this makes sending telemetry impractical
that's the point
I had a systemd timer set up to take a #btrfs snapshot every hour and delete the snapshot taken 25 hours ago (so there are 24 snapshots at all times)
since the last time I upgraded (to #linux 6.1.9), removing the old snapshot pins a CPU thread to 100% with
btrfs-transaction for several minutes, making IO latency (including network) on the system generally terrible
time to figure out what has happened. there isn’t any substantial btrfs change in 6.1.9 compared to 6.1.8, so I’m unsure whether it’s a regression, or just a coincidence and I broke something else
@halcy i tend to conceptualize it like “it outputs what a random webpage on the topic would say”, and that seems to elicit the right level of source criticism
re: how to clean your mouse (much less interesting than training your dragon)
another cleaning tip: use a thin piece of plastic, about 3 mm wide (a cut it out from a transparent plastic sheet meant as a cover for documents) to knock out stubborn pieces of lint from under the wheel in its casing once the mouse is disassembled
had to get rid of my matrix homeserver, because it took up way too much resources compared to how little (facebook and irc bridge, essentially) i was using it
now i’m back at the stage where i can postgres
VACUUM FULL; without buying more disk space for my vps
today is probably the first time tex told me
Something's wrong--perhaps a missing \item. and indeed a
\item was missing
how to clean your mouse (much less interesting than training your dragon)
the iFixit guide is essentially correct, but here are some other notes if the scroll wheel starts feeling gummy. for me, gently pulling the wheel up got rid of the gummyness until i pressed the wheel, but turns out this has nothing to do with the click spring below the wheel. i simply had to clean the lint and dust from the inside of the wheel:
- turn off the mouse
- peel off the 3 feet of the mouse (the front one can stay) and remove the screws underneath with a 00 philips bit. the ifixit guide say the feet aren’t reusable unless heated beforehand, but if you dig deep enough (i used the guitar pick-like ifixit opener thing) you can pull off all the adhesive together with the feet in one piece, which can be just stuck back later. also remove the exposed screw on the front with a T5 torx bit
- pull off the shell from the mouse. i used the guitar pick-like thing liberally. there are two challenges: there is a very stiff clip below the right button, so the base has to be slid forward to disengage it, but the scroll wheel and the lip of the base next to the horizontal scroll wheel will prevent the base from sliding forward. my tactic was to pull the base downwards while pushing it as forward as possible, and poking with the guitar pick-thing below the right button
- disconnect the ribbon cable from the base. make sure not to break both pegs of the clip holding the cable. the shell is free now
- clean around the wheel and especially the teeth of the internal gear on the while. i wrapped a microfibre cloth around the (closed) jaws of plastic tweezers to get between the teeth properly. compressed air would also help here, i believe, and maybe even some silicone oil for the axle
- reconnect the ribbon cable. this may require some maneuvering, because the shell obstructs the space for the fingers
- turn the mouse on and test whether it still works (it’s a pain to have to disassemble right after you have put it back together). make sure not to touch any PCB, then turn the mouse off again
- snap the shell back to the base. the challenges are the same as for the disassembly. i could get it to snap back fully together except for the stiff clip below the right button. a combination of pushing the clip itself back (using a piece of plastic) from the gap left at the front and pushing from the top helped
- put back the screws and stick the feet back
i also had some luck previously with rejuvenating the rubber coating eaten by skin oils with some isopropyl alcohol and cotton balls. i hope to use this mouse for a long time, despite it accumulating wear and tear quite readily
the mx master 3 is such an annoying piece of gear to reassemble
why are there a myriad clips around the base that have to be pressed together at exactly the same to hold the shell? weren’t the screws on the back enough already?
it’s technically well repairable, one you figure out the puzzle that is its geomery
> absolutely cake on layers and layers of workarounds
> it's working now
> move the workarounds one by one to see which one did it
> i removed all the workarounds
> it's still working
what is this sorcery?
sealing ceremony, but instead of doing mormon stuff they crimp tungsten wire into a glass stem via an intermediate layer of tungsten(VI) oxide
been experimenting with using #vite #vitejs
import.meta.glob as a headless CMS
it feels quite magical, but, in contrast with #nextjs, i think i understand what it’s doing