Dark | Light
[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]

# ![@__tinygrad__ Avatar](https://lunarcrush.com/gi/w:26/cr:twitter::1674187388825006082.png) @__tinygrad__ the tiny corp

the tiny corp posts on X about gpu, chips, money, stack the most. They currently have XXXXXX followers and XXX posts still getting attention that total XXXXXXX engagements in the last XX hours.

### Engagements: XXXXXXX [#](/creator/twitter::1674187388825006082/interactions)
![Engagements Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1674187388825006082/c:line/m:interactions.svg)

- X Week XXXXXXX +367%
- X Month XXXXXXX -XX%
- X Months XXXXXXXXX -XX%
- X Year XXXXXXXXX -XX%

### Mentions: XX [#](/creator/twitter::1674187388825006082/posts_active)
![Mentions Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1674187388825006082/c:line/m:posts_active.svg)

- X Week XX +233%
- X Month XX -XX%
- X Months XXX +2.70%
- X Year XXX +36%

### Followers: XXXXXX [#](/creator/twitter::1674187388825006082/followers)
![Followers Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1674187388825006082/c:line/m:followers.svg)

- X Week XXXXXX +0.57%
- X Month XXXXXX +0.98%
- X Months XXXXXX +12%
- X Year XXXXXX +39%

### CreatorRank: XXXXXX [#](/creator/twitter::1674187388825006082/influencer_rank)
![CreatorRank Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1674187388825006082/c:line/m:influencer_rank.svg)

### Social Influence [#](/creator/twitter::1674187388825006082/influence)
---

**Social category influence**
[technology brands](/list/technology-brands)  #5150 [stocks](/list/stocks)  #1191 [finance](/list/finance)  [celebrities](/list/celebrities)  [social networks](/list/social-networks) 

**Social topic influence**
[gpu](/topic/gpu) #15, [chips](/topic/chips), [money](/topic/money), [stack](/topic/stack) #45, [$qcom](/topic/$qcom), [a very](/topic/a-very), [usb](/topic/usb), [open ai](/topic/open-ai), [the open](/topic/the-open), [rtx](/topic/rtx)

**Top assets mentioned**
[QUALCOMM, Inc. (QCOM)](/topic/$qcom)
### Top Social Posts [#](/creator/twitter::1674187388825006082/posts)
---
Top posts by engagements in the last XX hours

"@sytelus while it's great that the nanochat README includes the important metric of line count they forgot to count the lines in their dependencies 🙃 @karpathy"  
[X Link](https://x.com/__tinygrad__/status/1978017792470200645) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-14T08:39Z 59.8K followers, 17.2K engagements


"Playing with my AMD Hawk Point NPU the docs/examples are quite good You might already have one of these in your computer. It feels very @tenstorrent"  
[X Link](https://x.com/__tinygrad__/status/1979519693934465410) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-18T12:07Z 59.8K followers, 20.4K engagements


"On Omarchy. Get amd/xdna-driver installed then the stuff from this guide works"  
[X Link](https://x.com/__tinygrad__/status/1979520023233503368) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-18T12:08Z 59.8K followers, 4873 engagements


"The insight in their own words. From: The beauty of embedding this in tinygrad is that it will immediately work on all of tinygrad's GPU backends. AMD CUDA OpenCL Metal and QCOM"  
[X Link](https://x.com/__tinygrad__/status/1976085676119687607) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-09T00:41Z 59.4K followers, 7038 engagements


"@beaversteever what we have found is that it's best to work with chips that are already taped out and have a buy it now button like NVIDIA/AMD/Apple/Qualcomm GPUs. are you at that stage"  
[X Link](https://x.com/__tinygrad__/status/1978271734336758128) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-15T01:28Z 59.4K followers, 3308 engagements


"Shout out to ThunderKittens for writing simple yet very performant GPU code. We're working on "tinykittens" which uses the same insight but in tinygrad's language. The insight is that GPU "registers" are the wrong primitive and TK's "register tile" is a lot more sensible"  
[X Link](https://x.com/__tinygrad__/status/1976084605141909845) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-09T00:37Z 59.8K followers, 95.4K engagements


"@jeremyphoward They are trending in the right direction at least. MI355X gets XX% (up from 60%) while GB200 only gets XX% (down from 80%). CDNA4 is pretty nice. Dedicated acc registers and TCs that are 16x16x32 running at the same speed as the 16x16x16 on CDNA3"  
[X Link](https://x.com/__tinygrad__/status/1980088578811281632) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-20T01:48Z 59.8K followers, 1650 engagements


"AKA don't buy a DGX Spark buy a 5090 instead for half the price and 5x the performance. Would it help if you could use it from your Mac over USB"  
[X Link](https://x.com/__tinygrad__/status/1978376854416286143) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-15T08:26Z 59.8K followers, 89.8K engagements


"8xAMD MI300X mostly out of the box on nanochat (which is really an amazing repo and I'm excited for my custom chatbot). This is without PYTORCH_TUNABLEOP_ENABLED which I was too impatient for"  
[X Link](https://x.com/__tinygrad__/status/1978473112267317380) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-15T14:48Z 59.8K followers, 18.7K engagements


"@monstercameron What do you mean by "run cuda" How much would you pay We don't sell software but we are open to open source contracts if you have a specific target in mind"  
[X Link](https://x.com/__tinygrad__/status/1980105371986612356) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-20T02:54Z 59.8K followers, 1211 engagements


"@HotAisle I guess most people don't care but AMD should improve this"  
[X Link](https://x.com/__tinygrad__/status/1979097525102461201) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-17T08:09Z 59.6K followers, XXX engagements


"err pick up tinygrad and learn. once you have the framework finished the chip is easy"  
[X Link](https://x.com/__tinygrad__/status/1977988005857284458) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-14T06:41Z 59.8K followers, 100.3K engagements


"Here are the open bounties. If you know how to make things fast there's a lot of money on the table. Also some driver work MLPerf models and two new backends on offer"  
[X Link](https://x.com/__tinygrad__/status/1978017067879018630) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-14T08:36Z 59.8K followers, 29.3K engagements


"You can see from this chart why we also think the RTX Pro 6000 is not worth the money. 5x the price for X performance gain over a 5090 (do you really want llama 70B at XX tok/s). This is why tinyboxes are chock full of 5090s"  
[X Link](https://x.com/__tinygrad__/status/1978378080872755545) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-15T08:31Z 59.8K followers, 9303 engagements


"Thanks to sirhcm tinygrad now supports all the backends in Mesa by rendering to NIR. One of the Mesa backends is NAK and with it we can compile to SASS. An NVIDIA free stack"  
[X Link](https://x.com/__tinygrad__/status/1978402535535993210) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-15T10:08Z 59.8K followers, 11K engagements


"We have both AMD and NVIDIA GPUs working over USB4 on Apple M chips. Sadly it needs a XXX line dext to map the BARs into user space. We are waiting on Apple to give us the DriverKit entitlement. Do you think they will"  
[X Link](https://x.com/__tinygrad__/status/1978648251558690843) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-16T02:24Z 59.8K followers, 43.6K engagements


"It's not tiny but it's the easiest way to get a machine capable of running gpt-oss-120b and nanochat (at reasonable speeds) in your house. It has 15x the FP16 FLOPS and 26x the RAM bandwidth for just 7x the price of DGX Spark. Buy today"  
[X Link](https://x.com/__tinygrad__/status/1978984754461499428) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-17T00:41Z 59.8K followers, 42.9K engagements


"5090s are up in price on eBay. Is $25k too cheap for the tinybox green v2 Should we raise the price Only X boxes left in stock and I don't know when we'll get more 5090s"  
[X Link](https://x.com/__tinygrad__/status/1979218591548608918) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-17T16:10Z 59.8K followers, 12.7K engagements


"Hmm Grok thinks it can solve a bounty. Do we believe it"  
[X Link](https://x.com/__tinygrad__/status/1979745494588182741) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-19T03:04Z 59.8K followers, 49.2K engagements


"NVIDIA over USB4 on MacBook is ready to try * ADT-UT3G dock + any 30/40/50 series GPU * Disable SIP * Install driver extra/usbgpu/tbgpu * Install NVK compiler brew install tinymesa * Test with: DEBUG=2 NV_NAK=1 NV=1 python3 test/test_tiny.py TestTiny.test_plus"  
[X Link](https://x.com/__tinygrad__/status/1980082660920918045) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-20T01:24Z 59.8K followers, 144.3K engagements


"@TheZachMueller Is this a sponsored post by NVIDIA You can buy one 6000 Pro or you can buy 4x 5090s for the same price"  
[X Link](https://x.com/__tinygrad__/status/1980083368558809397) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-20T01:27Z 59.8K followers, XXX engagements


"@kommentlezz lol we'll settle for NVIDIA fixing the XX series reset issue and Apple approving our DriverKit cert"  
[X Link](https://x.com/__tinygrad__/status/1980085457892573357) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-20T01:35Z 59.8K followers, 2971 engagements


"@mfastbanger In theory any external GPU dock should work. Try it"  
[X Link](https://x.com/__tinygrad__/status/1980188240327172175) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-20T08:24Z 59.8K followers, 1125 engagements


"Our GPU stack for both NVIDIA and AMD aside from minimal pieces of signed firmware is XXX% open source and pure Python except for the compiler. It's not using vendor drivers frameworks or libraries. That's why it's so easy to make it work on Mac. For compilers on AMD we use upstream LLVM and on NVIDIA we use the NAK compiler from the MESA project. We plan to replace the compiler with pure tinygrad in a year or two as well. With RANGEIFY merged our lowering stuff now matches the state of the art TVM style. We're studying ThunderKittens and TileLang for speed at that level and should have all"  
[X Link](https://x.com/__tinygrad__/status/1980193031019258109) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-20T08:43Z 59.8K followers, 43K engagements


"Meeting time In our Discord. Everyone can listen"  
[X Link](https://x.com/__tinygrad__/status/1980257840599298534) [@__tinygrad__](/creator/x/__tinygrad__) 2025-10-20T13:00Z 59.8K followers, 7058 engagements

[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]

@tinygrad Avatar @tinygrad the tiny corp

the tiny corp posts on X about gpu, chips, money, stack the most. They currently have XXXXXX followers and XXX posts still getting attention that total XXXXXXX engagements in the last XX hours.

Engagements: XXXXXXX #

Engagements Line Chart

  • X Week XXXXXXX +367%
  • X Month XXXXXXX -XX%
  • X Months XXXXXXXXX -XX%
  • X Year XXXXXXXXX -XX%

Mentions: XX #

Mentions Line Chart

  • X Week XX +233%
  • X Month XX -XX%
  • X Months XXX +2.70%
  • X Year XXX +36%

Followers: XXXXXX #

Followers Line Chart

  • X Week XXXXXX +0.57%
  • X Month XXXXXX +0.98%
  • X Months XXXXXX +12%
  • X Year XXXXXX +39%

CreatorRank: XXXXXX #

CreatorRank Line Chart

Social Influence #


Social category influence technology brands #5150 stocks #1191 finance celebrities social networks

Social topic influence gpu #15, chips, money, stack #45, $qcom, a very, usb, open ai, the open, rtx

Top assets mentioned QUALCOMM, Inc. (QCOM)

Top Social Posts #


Top posts by engagements in the last XX hours

"@sytelus while it's great that the nanochat README includes the important metric of line count they forgot to count the lines in their dependencies 🙃 @karpathy"
X Link @tinygrad 2025-10-14T08:39Z 59.8K followers, 17.2K engagements

"Playing with my AMD Hawk Point NPU the docs/examples are quite good You might already have one of these in your computer. It feels very @tenstorrent"
X Link @tinygrad 2025-10-18T12:07Z 59.8K followers, 20.4K engagements

"On Omarchy. Get amd/xdna-driver installed then the stuff from this guide works"
X Link @tinygrad 2025-10-18T12:08Z 59.8K followers, 4873 engagements

"The insight in their own words. From: The beauty of embedding this in tinygrad is that it will immediately work on all of tinygrad's GPU backends. AMD CUDA OpenCL Metal and QCOM"
X Link @tinygrad 2025-10-09T00:41Z 59.4K followers, 7038 engagements

"@beaversteever what we have found is that it's best to work with chips that are already taped out and have a buy it now button like NVIDIA/AMD/Apple/Qualcomm GPUs. are you at that stage"
X Link @tinygrad 2025-10-15T01:28Z 59.4K followers, 3308 engagements

"Shout out to ThunderKittens for writing simple yet very performant GPU code. We're working on "tinykittens" which uses the same insight but in tinygrad's language. The insight is that GPU "registers" are the wrong primitive and TK's "register tile" is a lot more sensible"
X Link @tinygrad 2025-10-09T00:37Z 59.8K followers, 95.4K engagements

"@jeremyphoward They are trending in the right direction at least. MI355X gets XX% (up from 60%) while GB200 only gets XX% (down from 80%). CDNA4 is pretty nice. Dedicated acc registers and TCs that are 16x16x32 running at the same speed as the 16x16x16 on CDNA3"
X Link @tinygrad 2025-10-20T01:48Z 59.8K followers, 1650 engagements

"AKA don't buy a DGX Spark buy a 5090 instead for half the price and 5x the performance. Would it help if you could use it from your Mac over USB"
X Link @tinygrad 2025-10-15T08:26Z 59.8K followers, 89.8K engagements

"8xAMD MI300X mostly out of the box on nanochat (which is really an amazing repo and I'm excited for my custom chatbot). This is without PYTORCH_TUNABLEOP_ENABLED which I was too impatient for"
X Link @tinygrad 2025-10-15T14:48Z 59.8K followers, 18.7K engagements

"@monstercameron What do you mean by "run cuda" How much would you pay We don't sell software but we are open to open source contracts if you have a specific target in mind"
X Link @tinygrad 2025-10-20T02:54Z 59.8K followers, 1211 engagements

"@HotAisle I guess most people don't care but AMD should improve this"
X Link @tinygrad 2025-10-17T08:09Z 59.6K followers, XXX engagements

"err pick up tinygrad and learn. once you have the framework finished the chip is easy"
X Link @tinygrad 2025-10-14T06:41Z 59.8K followers, 100.3K engagements

"Here are the open bounties. If you know how to make things fast there's a lot of money on the table. Also some driver work MLPerf models and two new backends on offer"
X Link @tinygrad 2025-10-14T08:36Z 59.8K followers, 29.3K engagements

"You can see from this chart why we also think the RTX Pro 6000 is not worth the money. 5x the price for X performance gain over a 5090 (do you really want llama 70B at XX tok/s). This is why tinyboxes are chock full of 5090s"
X Link @tinygrad 2025-10-15T08:31Z 59.8K followers, 9303 engagements

"Thanks to sirhcm tinygrad now supports all the backends in Mesa by rendering to NIR. One of the Mesa backends is NAK and with it we can compile to SASS. An NVIDIA free stack"
X Link @tinygrad 2025-10-15T10:08Z 59.8K followers, 11K engagements

"We have both AMD and NVIDIA GPUs working over USB4 on Apple M chips. Sadly it needs a XXX line dext to map the BARs into user space. We are waiting on Apple to give us the DriverKit entitlement. Do you think they will"
X Link @tinygrad 2025-10-16T02:24Z 59.8K followers, 43.6K engagements

"It's not tiny but it's the easiest way to get a machine capable of running gpt-oss-120b and nanochat (at reasonable speeds) in your house. It has 15x the FP16 FLOPS and 26x the RAM bandwidth for just 7x the price of DGX Spark. Buy today"
X Link @tinygrad 2025-10-17T00:41Z 59.8K followers, 42.9K engagements

"5090s are up in price on eBay. Is $25k too cheap for the tinybox green v2 Should we raise the price Only X boxes left in stock and I don't know when we'll get more 5090s"
X Link @tinygrad 2025-10-17T16:10Z 59.8K followers, 12.7K engagements

"Hmm Grok thinks it can solve a bounty. Do we believe it"
X Link @tinygrad 2025-10-19T03:04Z 59.8K followers, 49.2K engagements

"NVIDIA over USB4 on MacBook is ready to try * ADT-UT3G dock + any 30/40/50 series GPU * Disable SIP * Install driver extra/usbgpu/tbgpu * Install NVK compiler brew install tinymesa * Test with: DEBUG=2 NV_NAK=1 NV=1 python3 test/test_tiny.py TestTiny.test_plus"
X Link @tinygrad 2025-10-20T01:24Z 59.8K followers, 144.3K engagements

"@TheZachMueller Is this a sponsored post by NVIDIA You can buy one 6000 Pro or you can buy 4x 5090s for the same price"
X Link @tinygrad 2025-10-20T01:27Z 59.8K followers, XXX engagements

"@kommentlezz lol we'll settle for NVIDIA fixing the XX series reset issue and Apple approving our DriverKit cert"
X Link @tinygrad 2025-10-20T01:35Z 59.8K followers, 2971 engagements

"@mfastbanger In theory any external GPU dock should work. Try it"
X Link @tinygrad 2025-10-20T08:24Z 59.8K followers, 1125 engagements

"Our GPU stack for both NVIDIA and AMD aside from minimal pieces of signed firmware is XXX% open source and pure Python except for the compiler. It's not using vendor drivers frameworks or libraries. That's why it's so easy to make it work on Mac. For compilers on AMD we use upstream LLVM and on NVIDIA we use the NAK compiler from the MESA project. We plan to replace the compiler with pure tinygrad in a year or two as well. With RANGEIFY merged our lowering stuff now matches the state of the art TVM style. We're studying ThunderKittens and TileLang for speed at that level and should have all"
X Link @tinygrad 2025-10-20T08:43Z 59.8K followers, 43K engagements

"Meeting time In our Discord. Everyone can listen"
X Link @tinygrad 2025-10-20T13:00Z 59.8K followers, 7058 engagements

creator/x::__tinygrad__
/creator/x::__tinygrad__