正在加载视频...

视频加载失败

Fable 5.2 wrote the transformer. Opus 5.5 generated the training data. The result: a 367K parameter language model that fits on a floppy disk and runs on a Pentium. AI is now building AI for hardware from 1995.

289,229 次观看 • 21 小时前 •via X (Twitter)

36 条评论

Hardik Gohil 的头像
Hardik Gohil18 小时前

AI really said we don't need more powerful hardware

Umair Shaikh 的头像
Umair Shaikh18 小时前

And intresting part is there's a lot to come for ai in future

Nell AI Labs 的头像
Nell AI Labs21 小时前

The Pentium is cool, but the real story is that one AI designed the brain and another wrote its textbook. That’s where things get weird.

Umair Shaikh 的头像
Umair Shaikh21 小时前

Yeah, the recursion is the part that actually feels wild

Impuls 的头像
Impuls16 小时前

If you'd shown that in 1995, they would have burned you at the stake

••• 的头像
•••19 小时前

This is absolutely insane! I was surprised to read about these floppy disks still being used in the aviation industry.

muku 的头像
muku16 小时前

cool man made browser theme with fable 5.2

danilo 的头像
danilo18 小时前

@ivanfioravanti 😍

AI MAYA 的头像
AI MAYA17 小时前

AI on a floppy disk is the most 2026 sentence ever

ἰχῶρος 的头像
ἰχῶρος17 小时前

and tells you the super bowl sandwich slappy into polling freedom gnats monster cheese when you ask what 2+3 is

wiser 的头像
wiser18 小时前

We need to see more what this model is capable of! Crazy stuff

Umair Shaikh 的头像
Umair Shaikh18 小时前

Sure, follow for more such updates

Rubi 的头像
Rubi14 小时前

Incredible innovation packed into tiny hardware!

pan robaczek 的头像
pan robaczek17 小时前

I could easily see this as a new type of virus, that can unpack itself into the machine and write the code for the rest of it on it's own. Mind blowing

Latent 的头像
Latent20 小时前

what can it do on facts or combinations that weren't in the generated training set? the tiny hardware demo is fun, but separating recall from generalisation would make it much more interesting.

Umair Shaikh 的头像
Umair Shaikh20 小时前

Fair point

Hypertyzz 的头像
Hypertyzz16 小时前

Is this legal?

𝚕𝚘𝚜𝚝 𝚘𝚗 𝚠𝚠𝚠 的头像
𝚕𝚘𝚜𝚝 𝚘𝚗 𝚠𝚠𝚠13 小时前

ok. this is cool :D

Juxhin Çela 🇪🇺🇺🇦 的头像
Juxhin Çela 🇪🇺🇺🇦16 小时前

Wow

Zentora 的头像
Zentora13 小时前

wth AI is everywhere

阿宇的AI日记|AI工作流实测 的头像
阿宇的AI日记|AI工作流实测14 小时前

367K参数的小模型,这分工有点意思

Aryan 的头像
Aryan14 小时前

what is this exactly?

Javi 的头像
Javi13 小时前

Why this is needed? 🤔

Dominik 的头像
Dominik13 小时前

367k parameters on a floppy makes the hardware constraint feel tangible.

zuphr1n 的头像
zuphr1n15 小时前

what if you could generate the weights with 5.5

Billy 的头像
Billy14 小时前

Running a sub-megaparameter language model on vintage hardware isn't for practical production—it's pure tech-hobbyist flexing to see if you can squeeze a transformer into a system with less RAM than a modern mouse's firmware.

Kimi Nguyen 的头像
Kimi Nguyen19 小时前

what's the actual performance of this model compared to modern ones?

Juxhin Çela 🇪🇺🇺🇦 的头像
Juxhin Çela 🇪🇺🇺🇦16 小时前

@grok is this true?

Syncognia SI 的头像
Syncognia SI17 小时前

How did you evaluate the training data?

Alex 的头像
Alex14 小时前

It's pretty dumb but once models get better at making models, then this could be impressive.

berberecho13 的头像
berberecho1315 小时前

repo? hf?

Mykhailo Sorochuk 的头像
Mykhailo Sorochuk15 小时前

tiny brain on a floppy, wild recursion, love the retro hack

Nakul Goel 的头像
Nakul Goel17 小时前

This is courage

Fascinating 的头像
Fascinating17 小时前

so useless

Paweł Toporczyk 的头像
Paweł Toporczyk16 小时前

A 367K model on a floppy and my browser game still needs a real GPU to draw a canyon. My whole performance budget is an upscaler and a render scale that quietly drops when the swarm gets big.

the_confused_teacher 的头像
the_confused_teacher15 小时前

Beautiful.

相关视频