Harrison Hines

6 months ago

💿 Open Source Release 💿 mdspan-cute: a zero-overhead bridge between C++23 std::mdspan and CUTLASS cute layouts. One header. Swizzled memory. No bank conflicts. Read the blog and check out the repo (links in reply)

1

11

1

0

2K

6 months ago

@KaroLyric 🤙

0

23

6 months ago

another cool open source release for Fleek. and several more in the pipeline

6 months ago

💿 Open Source Release 💿 nix2gpu: reproducible GPU containers that run anywhere. No Docker. No vendor lock-in. One command to deploy Same nix flake = same binary = same results. Read the blog and check out the repo (links in reply)

2

23

5

1

2K

3

9

1

0

467

harrisonhines retweeted

6 months ago

1/Yesterday we announced nix2gpu - a NixOS package for portable GPU containers. Portable containers prevent vendor inference lock-in. Here's why it's a big deal. #Nix #AIInfra

1

13

2

0

2K

6 months ago

@ceeychuks 🙏

0

1

0

25

6 months ago

if anyone can help me out with an arXiv cs.LG endorsement I would very much appreciate it. happy to send the paper we're planning to publish prior to endorsement if that is helpful. referrals or tips also welcome

3

10

0

3K

6 months ago

@ceeychuks thank you

1

0

112

6 months ago

@shreyanshsjains @KandaBhaji_x hey would you be able to do me a quick favor and endorse me for https://t.co/9JGbt69obI. i would really appreciate it: https://t.co/1nBPz8Wfws

1

0

73

6 months ago

this was fun. look forward to people getting to hear more about what we're cookin

6 months ago

Just wrapped our first pod recording 🎙️ Talked all things Fleek and efficient inference ⚡ We will drop it next week 🤙

5

20

4

0

3K

3

9

2

0

887

6 months ago

Another huge efficient inference deal/partnership. Probably something.

Yuchen Jin

@Yuchenj_UW

6 months ago

OpenAI’s multibillion-dollar partnership with Cerebras makes total sense. Sam made the right call. Cerebras chips are insanely fast at inference, sometimes 20x Nvidia GPUs, similar to Groq. My biggest issue with ChatGPT and GPT-5.2 Thinking/Pro is latency. Cerebras software stack is nowhere near CUDA, but for accelerating a small set of GPT models, it’s absolutely worth it, especially to win more users against Google Gemini. Honestly, I wouldn’t be surprised if Sam paid $20B to acquire them.

Yuchenj_UW's tweet photo. OpenAI’s multibillion-dollar partnership with Cerebras makes total sense. Sam made the right call.

Cerebras chips are insanely fast at inference, sometimes 20x Nvidia GPUs, similar to Groq.

My biggest issue with ChatGPT and GPT-5.2 Thinking/Pro is latency. Cerebras software stack is nowhere near CUDA, but for accelerating a small set of GPT models, it’s absolutely worth it, especially to win more users against Google Gemini.

Honestly, I wouldn’t be surprised if Sam paid $20B to acquire them.

41

399

26

82

60K

0

5

0

308

harrisonhines retweeted

6 months ago

* Mega AI labs subsidize it * AI startups bleed margins to it * Indie hackers & vibe coders can't afford it ❗ Inference costs are a problem❗ But don't worry. We're making inference so cheap that everyone can ship sustainable, world-class AI products. 🚀⚡

6

25

8

1

2K

6 months ago

@PartnershipDude 🥳

0

1

0

27

6 months ago

Excited to start open sourcing some of the cool tools and infra we are using in our AI stack. Starting with nimi. For any nix fans this one is certainly worth checking out.

6 months ago

💿 Open Source Release 💿 nimi: a tiny process manager that takes NixOS 25.11's modular services spec and runs it anywhere you need. One Service Definition to Rule Them All. Read the blog and check out the repo (links in reply)

4

22

4

1

4K

4

14

4

0

2K

harrisonhines retweeted

6 months ago

1/ Yesterday we released Nimi: a Rust process manager that runs the same service definition everywhere. NixOS, home-manager, containers, dev shells. One config. Zero rewrites. Today let’s dive deeper into how it works 🧵

1

20

7

1

2K