#asahi · 2026-03-15 04:52
chadmednicolas17: yeah i mean, the hassle of repairing these things is not worth it when theres 1000-2000 of the things in the fleet and they cost basically nothing #asahi · 2026-03-15 04:27
nicolas17(yet it's better than other Macs where you can't replace the keyboard at all and need a whole new top case)#asahi · 2026-03-15 04:26
nicolas17the keyboard is apparently the biggest pain in the ass to replace on the Neo#asahi · 2026-03-15 04:26
nicolas17<JamesCalligeros[m]> Im betting schools and such will just write them off if they get busted keyboards and other things like that given how cheap they are#asahi · 2026-03-15 00:38
yvainhi all#asahi · 2026-03-14 23:26
fl0_idbut agree it acribes too much intent#asahi · 2026-03-14 23:26
fl0_id(like more like actually also reacting properly to a no)#asahi · 2026-03-14 23:26
fl0_idJamesCalligeros[m] i kinda keep up to date on it for work, for relevance to IP, data protection and licensing. some have gotten a lot better with it though I think#asahi · 2026-03-14 23:16
JamesCalligeros[m]Yeah thats what i meant. The maintainers of these systems just do not see this as an issue to be fixed#asahi · 2026-03-14 23:15
chaos_princessthat is ascribing too much intent to the llm. it is not intentionally violating consent, being evil, or whatever, it just is. and ofc the weights nudges the response towards continuing to output, that's just what the training data looks like#asahi · 2026-03-14 23:04
chadmedyou never get a "no" option with big tech, only "maybe later" #asahi · 2026-03-14 23:04
chadmedbig tech has never, ever respected consent in anything theyve built#asahi · 2026-03-14 22:40
chaos_princesstrains stochastic parrot to always keep yapping and being pretend-helpful. tells no to said parrot. "no" fails to override the parrot's training data, and the parrot keep yapping. surprised pikachu#asahi · 2026-03-14 22:38
dystivthis "no" that the ai doesnt accept/understand is so hilarious#asahi · 2026-03-14 19:05
flightlessowlokay much appreciated#asahi · 2026-03-14 19:05
chaos_princesskb backlight is about as simple as it gets#asahi · 2026-03-14 19:04
flightlessowlany recommendations on a starting point? I don't know how to tell the difference between a simple and complex driver#asahi · 2026-03-14 18:57
flightlessowlokay#asahi · 2026-03-14 18:55
chaos_princessnot useful now, and won't be for a while. M4+ need a certain prerequisite (SPTM) that will take a lot of work#asahi · 2026-03-14 18:54
flightlessowlUnderstood. When considering simple drivers to reimplement, do mean ones found inside of the forked linux repo? Or are you recommending a simple driver not necissarily related to asahi? I have an M5 mac and I haven't seen that in the documentation so maybe not a useful testing device at this point?#asahi · 2026-03-14 18:46
nicolas17yeah the easy tasks have already been taken#asahi · 2026-03-14 18:44
chaos_princesswe don't really have any low hanging fruit anymore really, so, maybe consider re-creating a simple existing driver without looking at it's code. just to learn how it all works#asahi · 2026-03-14 18:41
flightlessowllearning/contribution?#asahi · 2026-03-14 18:41
flightlessowllol thats pretty good. Just completely disregarded it. I can learn without them. I've been finding it helpful reviewing docs in school but do not use it to write code there. But these days very much encouraged at my job. But point taken no with the link to the dictionary definition. I am completely new to systems level programming. Where would be the best place for me to start in terms of #asahi · 2026-03-14 18:35
nicolas17chaos_princess: LLMs don't understand that word https://tl.nulldata.foo/uploads/61e105bfd3234b4d/yes.jpeg#asahi · 2026-03-14 18:31
chaos_princesshttps://www.merriam-webster.com/dictionary/no#asahi · 2026-03-14 18:30
nicolas17is it *so* hard to just not?#asahi · 2026-03-14 18:29
flightlessowlcurious how far this goes#asahi · 2026-03-14 18:29
flightlessowlHi everyone. I am new to this IRC channel. Also, I am a newer developer who is working in AI but have been following this project for a while. I know there is no LLM generated code that can be contributed to the project. But I want to learn and contribute to this project because I love the work y'all have done. Is it okay to use LLMs as part of my learning process but not code generation? I am #asahi · 2026-03-14 08:44
JamesCalligeros[m]Perfect for scratching up traces/other destructive RE#asahi · 2026-03-14 08:43
JamesCalligeros[m]Institutions will treat them as consumables and therell be heaps of still working examples with minor damage only for bargain bin prices#asahi · 2026-03-14 08:42
JamesCalligeros[m]Im betting schools and such will just write them off if they get busted keyboards and other things like that given how cheap they are#asahi · 2026-03-14 08:41
chaos_princess"broken" ?#asahi · 2026-03-14 08:40
JamesCalligeros[m]Given the target market im sure there will be heaps of “broken” machines on the secondary market before too long#asahi · 2026-03-14 08:39
svengonna be much slower than usb2 but should work #asahi · 2026-03-14 08:38
svenyeah, was just gonna suggest to solder to the hw UART testpoints instead #asahi · 2026-03-14 08:38
chaos_princessgood thing the mbn is only $500, easier to justify soldering to test points :P#asahi · 2026-03-14 08:37
jannaua considerable part of the hypervisor is implemented in python to be run on the host. you would need to port that to run on the target device#asahi · 2026-03-14 08:33
jmax[m]jannau: Can m1n1 be updated to trace to DRAM for the period of the experiment, then the dump can be 'played back' into the m1n1 python host parser after the experiment is concluded? I guess this wasn't needed in the past but could be useful for tracing things like USB resets or entering/exiting super low power states.#asahi · 2026-03-14 08:25
svenI’m really not looking forward to that mess :/#asahi · 2026-03-14 06:21
jannauI guess at least the usb2 part won't need resets as the hub is always connected (at least as long as one port is used)#asahi · 2026-03-14 06:06
LarstiQyeah exactly#asahi · 2026-03-14 06:05
nicolas17._.#asahi · 2026-03-14 06:05
LarstiQwhat are the odds of that?#asahi · 2026-03-14 06:03
nicolas17maybe the USB controller is not fucked up aynmore and you don't need to reset? ._.#asahi · 2026-03-14 06:02
jannauWe can't keep the same way as on m1/m2 currently because the resets would break the other port on connect. we can't trace because that consumes the usb port. best case I can think of is if debugusb works on the usb2 port without interfering with the other port#asahi · 2026-03-14 05:57
jannaunicolas17: there is a single USB controller but USB2 and USB3 are always separate. See "Linux Foundation 2.0 root hub" and "Linux Foundation 3.0 root hub" you get for each USB3 controller. getting usb to work relieably on the neo will be annoying#asahi · 2026-03-14 05:17
TramtristI have an refurb japanese M4 pro with pro chip comin tomorrow. Will bide my time#asahi · 2026-03-14 04:41
Ardu0usYeah and the amount of stuff that just worked right out of the box on my M1 Max was super impressive.