News for hackers. Stuff that matters.

ixaxaar

I'm running it using a 4090 on using llama.cpp with Q5_K_S and its running at ~33 t/s
Re:The 'Father of the Internet' is finally retiring on Sunday July 12, 2026 @12:41PM (#48882453)
For a moment there I read "TC" and was like "the candidate"?!?!
Re:The 'Father of the Internet' is finally retiring on Sunday July 12, 2026 @07:32AM (#48880287)
Apparently there's "more to come" and its not a complete retirement.
Re:Midjourney Medical on Thursday June 18, 2026 @05:34AM (#48582976)
Well healthcare needs its own DLSS 5!
Okay so regardless of the model, platforms provide attestation from end to end that nothing is being logged from either input or output, including at the firmware and OS level to the extent that the customers have proof of the data never being saved. AFAIK, both GPUs and TPUs support this.
Re:Git 3.0 will use main as the default branch on Monday November 24, 2025 @02:30AM (#46031306)
From a third-world perspective, it feels like American politics being injected into the developer domain because of some previous biases that Americans had. Which is sad for a community that claims to be global.
Re:Cormac McCarthy's personal library on Wednesday October 01, 2025 @11:51PM (#45446174)
Or maybe he was searching for "structure", the way Grothendieck was - in pure math, in systems biology, in complex networks (e.g. civilizations, culture, occult) and in individual behaviors
I wonder how many of these hype-chasing AI founders are going to take the path of self-enrichment while capitulating their startups when they get exhausted of searching for a moat.
Re:Claude Code Best Practices on Sunday April 20, 2025 @08:48AM (#43743439)
> These are effectively identical to outsourcing your product to someone like Infosys.

But in my experience, the user has to be better than an Infosys employee to know how to convey the task to the LLM and then verify iteratively.

So more like an experienced engg outsourcing work to a service company engg.

Re:Nvidia-Ingest: Multi-modal data extraction on Friday January 10, 2025 @07:03AM (#42654912)
Ah so like NIM is a set of microservices on top of various models, and this is another set of microservices using NIM microservices to do large scale OCR?

and that too integrated with prometheus, 160GB VRAM requirement and so on?

Looks like this is targeted for enterprises or maybe governments etc trying to digitalize at scale.

Garbage in garbage out.
Re:Fly.io outage (restoration in progress) on Monday November 25, 2024 @11:16PM (#42242549)
Can you shed some more light on this "browning out" phenomenon?
What a sad fucking world. I like what China does in the regard to patents. That is exactly what patents deserve.
Re:Micron launches 60TB PCIe gen5 SSD with 12GB/s read speeds on Tuesday November 12, 2024 @10:48PM (#42122723)
Woohoo exactly!
Cool, rise of concert tourism? At least better than medical tourism ig.
Re:Healthcare Datasets for AI/ML on Thursday May 30, 2024 @05:30AM (#40521726)
Hey guys, I built this large list of healthcare datasets from last year when I spent some time in researching and downloading them to try out various things (AI).

Hope this is useful to someone!

Extreme pinky swear.
So a kind of lack of empathy? Do these guys have this image of "autists" and are basically filtering for them, cause this criteria seems to be favoring oppositional defiance disorder.

I mention this specifically because I remember mark andreseen comment something similar in lex fridman's podcast, something along the lines of getting "those creative people" together to build on ai.

Re:No "Zero-Shot" Without Exponential Data on Thursday May 09, 2024 @03:23PM (#40311996)
The physical world has limits, that's why sigmoids everywhere.
My AI chat thing: https://imgur.com/a/cxR8KpM (WIP). Connects to openai transcription, completion and tts APIs. Refactoring to use assistants, to use it to feed it my fridge's manual and have it think it's my fridge.
I wrote a part of the kernel (uitron) for Toshiba SSDs that went into macbooks around 2013 on kate. It was my first and favorite editor :D
Re:Meta MTIA v2 on Wednesday April 10, 2024 @01:39PM (#39993531)
If all this turns out to be useless, burning their cash for nothing seems like a great way to accelerate tech while going down. I guess that would actually be a positive thing.
Re:Ask HN: Who is hiring? (April 2024) on Monday April 01, 2024 @04:48PM (#39899123)
Hey love what Kagi is doing!

I'm based out of India, can I apply?

Re:Ask HN: Who wants to be hired? (April 2024) on Monday April 01, 2024 @04:41PM (#39899032)

  Location: India
  Remote: Yes
  Willing to relocate: Yes
  Technologies: Python, Scala, terraform, k8s, scikit-learn, pytorch, transformers
  Résumé/CV: https://github.com/ixaxaar
  Email: ixaxaar at geniusrise dot . ai
I've built early stage startup teams, starting at 4-5, and taking them to 40-50. I can handle pretty much all forms of tech work - management, backend, frontend, devops, data engg, data sci, AI/ML

Some interesting links: https://github.com/geniusrise - containerized inference servers for LLMs etc (like NVIDIA NIM) https://github.com/ixaxaar/pytorch-dnc - reverse-engineered implementation of deepmind's DNC, SAM & SDNC http://monoid.space - pure math and theorem prover tutorial

Any org? Would, for example, openai be included in your definition of "any org"?

Look, "in principle" stuff is not how the real world works. AFAIK, hacks happen mostly because of carelessness. No one cares because no one cares if they care (and the compensations etc reflect that). I know enough such cases in fintech (forget about other verticals), which are mostly stupid like wrong RBAC, open firewall, AWS keys taken by roommate etc and not public of course.

Good. Also, don't hire good devs, only hire the cheap ones in India.

Money saved can be paid to these providers. That way, the money stays in the us. A 10k IQ move that no one will understand.

/s ov course

There are also open source equivalents btw - https://github.com/geniusrise
Re:Show HN: Flash Attention in ~100 lines of CUDA on Saturday March 16, 2024 @01:10PM (#39727626)
You mean triton the inference server or triton the DSL for cuda?

Older »