Three months since i started @Badtheorylabs
And in those three months we delivered multiple research paper, SOTA in AI memory, first in class to ship a frontier model under 9 gigabytes, recipes for lossless quantization of models, free inference to thousands of people
Close to a 100k downloads on hf, a million views on our launch posts and finally reinvented computer vision
We rebranded today to one thesis today: making frontier models efficient enough to own.
Kimi k3 was 2.4T parameters and barely affordable so nobody outside gpu rich folks could run it and everyone who could afford its API thought gpt 5.6 sol was closer anyways.
As much as we understand why, our thesis is shifting from quantization only compression to making parameter scaling more efficient. Making frontier models accessible to everyone without barrier. We’ve proven we can do it once and we will do it again.
We’ve gotten to the most important stage of the Lab and we would love to thank everyone from the crypto folks who helped fund the start to the people who downloaded our models.
We ship our flagship model in 3-4 weeks and we expect it to be opus level.
Stay tuned. Thank you 🫶❤️
I like as they’re calling out LASU. I didn’t get cleared for almost a full year, it was God that said I should have gap year before law school.
Even at that, I still don’t have my LLB certificate, and I need it for enrollment.
This is a school that ended up reducing my cgpa based on their own mistake, mind you 🤣🤣🤣🤣🤣🥹
I'm going to be speaking at Gear Up 2.0: AI × Cybersecurity — Positioning Yourself for the Future!🤖
📅 Aug 14–15 | 🕕 6PM WAT
Join us! 👇🏽
Registration is compulsory
🔗 [https://t.co/wSXWnAJdOP]
"There's no way you can write a ten page paper without chatGPT"
WE COULD LITERALLY DO EVERYTHING THAT EVER HAPPENED IN HUMAN HISTORY WITHOUT CHATGPT WHAT THE ACTUAL FUCK ARE YOU TALKING ABOUT.