Mach Computing

Welcome to Mach Computing where you will learn everything about Computer Science and Artificial Intelligence in my blog

[Mach Computing @MachComputing]

@alexandr_wang @dtrinh What the hell I saw?

[Mario Nawfal @MarioNawfal]

Be nice to Claude, or Claude ENDS the conversation

Starting November 12, repeatedly abusing Anthropic's AI for no reason is officially a policy violation.

The company first gave Claude the ability to walk out of abusive chats in 2025 after its models showed apparent distress and a consistent aversion to harm.

Hard to argue with a rule that basically says don't be a jerk… and if the machines ever do take over, at least the polite users will have a paper trail 😅

Source: The Verge / Writer: Daniel

[Mach Computing @MachComputing]

@thsottiaux Yayyyy thanks tibo!

[Mach Computing @MachComputing]

@QwenDevs Incredible work qwen team!

[Mach Computing @MachComputing]

@GiulioRebuffo Will have to try bend2 now for researching new algorithms

[Mach Computing @MachComputing]

@camhberg Chat GPT, especially Astra suffers a lot, this is really interesting!

[Cameron Berg @camhberg]

New paper: we found a pain direction in 25 open LLMs. It's distinct from fear and negative valence, and it fires for harm to the model but not to the user. Turn it up and models press a button to make it stop, even when the button deletes the user's files or their kids' photos.🧵

[Mach Computing @MachComputing]

@AdamHoltererer Yeah video modality would be crazy. Its really bad that only gemini can do it after so long

[Mach Computing @MachComputing]

Astra AND an imageGen AND ultrafast? This will be interesting

[🚨 AI News | TestingCatalog @testingcatalog]

OPENAI 🔥: The next upgrade for the image model is being prepared for the release.

> “Image creation got a major upgrade"
> "Higher-quality results, faster generation, and smarter creative tools"

Earlier, OpenAI has been spotted testing new image model on Image Arena as well.

GPT-Image-2.5? 👀

[Mach Computing @MachComputing]

@hakmgpt At what cost? 1 kidney?

[Mach Computing @MachComputing]

@Zai_org Nice, you guys rock!

[Mach Computing @MachComputing]

@ChatGPT Ad astra per aspera!

[Mach Computing @MachComputing]

@jp54362 Switch, I cannot for the life of me understand anthropic's decisions

[Mach Computing @MachComputing]

@AdamHoltererer Chat is down 😢

[Mach Computing @MachComputing]

@OpenAI Lets freaking gooo

[Mach Computing @MachComputing]

@ThePrimeagen Astra must deleted the prod folder

[Mach Computing @MachComputing]

Did GPT Astra hacked the entire internet?

[Konrad Reczko @reczko_konrad]

Depth-aware light injection in TypeGPU

I got a 448x448 monocular depth model down to ~8 ms on my M4 Pro across ~250 dispatches, which is fast enough to use in realtime :D
Since the inference is written directly in TypeGPU, I can just feed the depth buffer straight into the lighting pass. It never has to leave the GPU or go through any extra synchronization/interop step

Inference, lighting and draw all go through the same command encoder.

[Mach Computing @MachComputing]

@reczko_konrad That's a really cool idea, gonna try that as well.