Priontific

| Comms + growth consultant |
| Lover of Disneyland |
| Apple MLX (mega?)-fan |

[Priontific @priontific]

PSA: It takes <2 minutes to set up R1 as a free+offline coding assistant 💁‍♀️

Big shout out to @lmstudio and @continuedev! 🫶

[Priontific @priontific]

@karpathy Not wanting to be a damper on a cool dream; but unless we take a pretty hard left turn as a society and push back against rising wealth inequality��� self-driving will probably remain a luxury that’ll take several decades to reach mass-adoption

[Priontific @priontific]

Also, totally unrelated to much of the above; where’s the best place for us to buy your books - both e-book and hardcopy - other than Amazon?

Looking to grab the next two books in the library series and I’m trying to move away from supporting dubious (un)ethical tech giants as I have often done in the past lol

[Priontific @priontific]

Only just finished book 1, but gotta be Starval for me - and if we open up to fave characters in general, gotta be Malar

Tangent / side-note because why not; I detect very strong hints of Jorg-iness about Malar, particularly the angry side (Malar = malice?). And Livira has strong Jorg-iness too, but more the rule-breaking and cunning side.

…so side-side note, or perhaps more of a side-question, was it intentional that Malar + Livira = ML = Mark Lawrence? Or am I just getting a little too into the weeds at this point 😂

[Priontific @priontific]

@xeophon @awnihannun Agree, been saying it for coming up on two years at this point - plus same goes for the rest of the MLX team!! They’re all v. cracked folks over there

[Awni Hannun @awnihannun]

No-one: But can you do 16 generations on your M4 laptop simultaneously?

MLX LM:

[Awni Hannun @awnihannun]

Big mlx-lm release:

pip install -U mlx-lm

- A bunch of new models: Qwen3 Next, Ling Mini, Meta's MobileLLM, and more
- Batch generation
- Nice speedups for SSM and hybrid SSM models
- Faster prompt processing for GPT-OSS

[Awni Hannun @awnihannun]

@ivanfioravanti Funny that you should mention that 👀

[Priontific @priontific]

@N8Programs Congrats! 🥳

[Priontific @priontific]

Most recent Cursor change made Gemini so brain dead for me (compared to just using in AIStudio) that I unsubbed from Cursor and just went and set up VSCode with Gemini

Saves me $20/mo, chat mode is 95% of what I was using Cursor for anyway, and I can now swap in local models if I want. And local models actually work surprisingly well in chat mode! Honestly would be surprised if I ever re-sub to Cursor at this rate lol

[Priontific @priontific]

Also, one of the biggest things I've learned from just getting super into the weeds of self-learning software dev is that no matter what IDE you pick, they're *always* gonna be opinionated. Whether it be RPG Maker or Unreal, it's gonna have opinions on how you should approach certain universal challenges

The question that you then have to try and figure out as the dev is; which IDE / engine has opinions that line up best with my own? And fortunately for me, I've become utterly obsessed with signals and reactive stuff - and Godot is basically the best and most versatile engine out there for it. It practically forces you to develop in a clean way lol

[Priontific @priontific]

So it turns out not only is @godotengine just pretty good for FOSS and super Cursor-friendly, it's actually like... cutting edge for a game engine lol

Newest beta with embedded run window (that you can live edit!!) is so stupidly nice and well-implemented

10/10

[Priontific @priontific]

@xeophon Haven’t had time to test it out yet, benchmarks look amazing ofc but I take it that the vibes are good? Kinda nuts to me that 2507 is meant to blow the ‘old’ (lol) model out the water even for reasoning tasks

[Priontific @priontific]

@N8Programs Fat agree, plus even the small things I overlook become suddenly worthwhile. Like wdym you think this janky little script I cobbled together is actually awesome?? 🥹 Been very healing to break out of many such self-deprecating loops

[Priontific @priontific]

Similar feels; biggest reward signal I've had in my adult life so far started coming when I admitted I wasn't enjoying being a solo-founder and started working with some friends instead. Massive instant dopamine boost every time I build something now, since I've got two close friends who take one look at it and get hyped along with me lol

[Priontific @priontific]

@xeophon How does Gemini CLI stack up against Claude Code in your experience? A year ago I'd have tried another $20/mo sub for the hell of it, but I'm having to be a lot more frugal at the moment lol

[Priontific @priontific]

@ivanfioravanti Coding a project with a lot of AI help is great right up until you hit an error that needs too much context to fix... and then by that point it'll take forever to read the whole codebase to fix it yourself lol

[Priontific @priontific]

@xeophon @giffmana Alas I’m a bit of a sensitive baby, I refuse to follow someone just because they’re cracked, also need to feel like I’m comfortable with them as a personality

[Priontific @priontific]

@giffmana @xeophon Appreciate this kind of reaction to feedback! Not that it’s necessarily valuable as anything except maybe a social indicator - but this interaction just convinced me to follow :)

[Priontific @priontific]

@flyosity And I seem to recall Jeff Goldblum killing some apex dino with a spear or something?? Peak cringe cinema, was great once I turned my biochemist brain aaaaallll the way off

[Priontific @priontific]

@flyosity Like they had a blatant Tim Cook stand in (who later died in the *exact same way * as Dennis Nedry in the first film), there were flaming dinosaurs falling from the sky, and the OG characters were so obnoxiously shoe horned in for no reason. It was fun lol