the tiny corp

@__tinygrad__84k followers

We make tinygrad; sell tinybox for the GPU middle class. Our mission is to commoditize the petaflop.

Open on X

Top outlier

9.1×

Best organic post

Peak views

88k

Single post

Bookmarks

1.4k

Across 15 organic posts

Last scan

Oct 16

Community cache

Posts 15 posts

#	Date	Post	↓ Outlier	Views	Likes	Reposts	Bookmarks
1	Oct 10	AMD ended up sending us the boxes and giving us a $2M contract. Their stock went from $97.10 to $608.10 and the software is quite usable now. Are you a chip maker who wants to turn	9.1×	88k	1.7k	63	250
2	Oct 11	The future depends on the direction this trend takes.	8.3×	80k	1.1k	107	374
3	Oct 8	Meet the new tinybox. Configurable. Starting at $7,000.	7.1×	68k	1.2k	49	178
4	Oct 13	Ours is $19k with 192GB of VRAM	3.7×	36k	657	12	138
5	Oct 16	It is sad that @finkd @alexandr_wang @Meta felt the need to nerf their open release so badly. I mean, it's fine as a statement to the wannabe intelligence rent seekers, but they ha	2.6×	26k	420	17	49
6	Oct 13	Sorry, I got the price wrong for the GLM-5.3 flash machine. 250 tok/s should be doable with decent software. Buy today!	2.3×	22k	345	5	70
7	Oct 9	If you run an AI or robotics startup, why are you having your employees use bad computers that can't do experiments? Buy them a bunch of tinyboxes!	2.2×	21k	357	11	67
8	Oct 9	Last call for the tinybox v2. We have 1 green left and 2 reds. Pro prices will also go up (slightly) when we add it to the new selector, so if you want a pro it's a great time to o	1.9×	18k	116	3	4
9	Oct 10	@intel @Qualcomm @tenstorrent Our mission is to to commoditize the petaflop; we want to see many competitors to @nvidia. AMD is now a serious competitor. Who will be next? Oh I sho	1.9×	18k	236	3	11
10	Oct 14	Simple specs can give rise to huge complexity. Buffers are initialized with 0. What's in result after this runs?	1.5×	14k	172	2	39
11	Oct 15	By the end of week 5 in 10-411, you'll have written an MNIST trainer and Qwen3 runner using your compiler. Week 6-10 will focus on GPU usage and performance. Guess the line count?	1.4×	14k	266	6	66
12	Oct 12	Who's ready for 10-411?	1.4×	13k	342	5	115
13	Oct 10	@intel @Qualcomm @tenstorrent @nvidia @Google @Meta @JeffBezos Intel really needs the most help. I think this might be the cheapest 768GB of RAM in the world, but Intel lowered the	1.3×	12k	226	1	27
14	Oct 13	We shipped the simple configuration for the most commonly chosen builds + the brand new $19k tinybox red flash. Compare the GPU RAM and GPU bandwidth to the NVIDIA version, AMD is	1.2×	12k	156	2	28
15	Oct 16	Accelerator companies focusing on inference and tok/s at the expense of fundamentals is a serious mistake. LLMs are just a small part of all the things that will be built with AI i	1.2×	11k	268	7	27
16	Oct 10	Want to deploy exaboxes?	1.1×	11k	164	1	15
17	Oct 16	The first tinybox red flash is up! Get yours today for $19k	1.0×	9.7k	182	6	17
18	Oct 13	These AMD GPUs have the same RAM and similar power to 2x RTX6000s but are $26,000 cheaper.	0.4×	3.7k	33	0	1
19	Oct 14	The fact that this gets 9 likes (on 2k views) but you complain about the 10-411 course not being open source really says something. Answering questions like this is what the course	0.4×	3.6k	35	0	1
20	Oct 16	It's so nice to finally have a usable model running on hardware we built in tinygrad. Works end to end in pi. Getting 300/53 tok/s prefill/decode, should be at 2000/100 soon. 5000/	0.3×	3k	36	2	1
21	Oct 16	In 6 months, when everyone is running GLM-5.3 flash sized models at home and the required box costs $50k, you are going to be mad you didn't buy one of these now.	0.2×	2.4k	31	1	4

Scan another creator

Run this on any account.

XOutlier builds this table for every public profile you open on x.com, with live data and every post, not just the top 50.

Add to Chrome Download .zip