@ferrari955@trashh_dev I've bought two high-end System76 laptops 10 years apart and regretted it each time. Support is generally slow and expensive and the longevity/build quality just isn't there.
@rfleury 100%. My kids instinctively hate anything to do with AI because of this inherent cheapening. It's interesting to watch how visceral their rejection is.
@BenjaminDEKR The age of typing code character by character is over. We now generate it in chunks. Not sure this implies that one doesn't "write code anymore".
Pretty neat little visualizer created by Gemini showing the effect of the `shift` parameter in rectified flow models in ComfyUI: https://t.co/UAHf2w8Y7y . I'm reasonably sure it's not upside down.
https://t.co/cTf47C4P8c gives a 406 Not Acceptable to Linux-based User-Agents. Paging someone who may care at any of @UtahGov@KSLcom@UtahGOP@UtahDemocrats. This is a potentially serious access issue for people who need court info/access.
@GameChangerzUT Hey man, my son has a 3DS with a broken speaker and really wants that repaired. Is that something you can handle? We may swing in some day this week. Love what you're doing, figured Twitter would be faster than sending an email.
@MaximusMoroni@Ch_JesusChrist How many times did Joseph allow the Saints to be run out without "standing up"? What happened after Sidney Rigdon suggested fighting back in his infamous July 4 address? Like it or not, this is what God expects of his people.
Stay tuned for GC in a couple weeks.
@hadilq@tsoding It is fully decentralized -- the whole repo's contents sync to all clients. It's not mutually compatible with git, on purpose; it made choices to be different from git intentionally. There's `jj` if you're looking for something git-compatible, but it doesn't have wikis, etc.
@dhh Every time a new person mails the list I have to screen their message in. There's no concept of lists or list-threading, which makes following the list hard to do. I messaged support about this early on and was told it just wasn't supported yet. It's now been 3+ yrs.
@latkins So I pulled this, and have tried several times, but I get a 500 error on `chat/completions` no matter what I try. I've tried extracting the tool_use template independently, and still get a 500. The default template seems to work, but has no tool tokens. I'm on vLLM main.
Also used grok3 to bolt an OpenAI-compliant API to a model that isn't supported by any of the major inferencing platforms yet (Ovis2). I should share the code it wrote, but it feels weird making a PR with mostly AI-written code...
Using grok3 to troubleshoot/debug and fake-bisect an issue in the `v4l2loopback` driver's git branch:
https://t.co/isRtGsFArI
Needed some hand-holding and made some errors, but overall I'm impressed!