After Jensen Huang’s departure, some people in the AI community have started stockpiling models like doomsday supplies.
DHH, the creator of Rails, posted a reminder on November 12: back up a few open-weight models in advance, and make sure you have devices capable of running them. No one knows what exactly has happened at NVIDIA, but the world has just lost one of the most important supporters of free and open AI.
This isn’t coming out of nowhere: in September, NVIDIA announced its roughly $12.9 billion acquisition of Hugging Face, home to more than 3 million models and 18 million developers. The deal is expected to close in the first half of 2027. Huang has also been one of the most vocal advocates for open models over the past year. In July, he reposted an open letter signed by 25 organizations calling for open weights not to be restricted.
Someone on Reddit has already shown off their “AI strategic reserve”: a 2 TB hard drive stocked with several model families, including Qwen3.8, Gemma 4, and DeepSeek V4 Flash, with quantized versions at different precisions all ready to go. They’ve also got speech, image generation, embeddings, and an offline Wikipedia, with the most important models copied to a NAS for a second backup.
Of course, there’s currently no evidence that NVIDIA is going to change its open-model policy. The concern is that its champion may be gone—not that the policy has already changed. I think the whole thing is funny, but the logic checks out: cloud services can change prices, be taken down, or cut off access at any time. Once the model weights are on your own hard drive, nobody can take them back. Models that run offline are the ones that truly belong to you.
#AI Frontier
$BTC
DHH, the creator of Rails, posted a reminder on November 12: back up a few open-weight models in advance, and make sure you have devices capable of running them. No one knows what exactly has happened at NVIDIA, but the world has just lost one of the most important supporters of free and open AI.
This isn’t coming out of nowhere: in September, NVIDIA announced its roughly $12.9 billion acquisition of Hugging Face, home to more than 3 million models and 18 million developers. The deal is expected to close in the first half of 2027. Huang has also been one of the most vocal advocates for open models over the past year. In July, he reposted an open letter signed by 25 organizations calling for open weights not to be restricted.
Someone on Reddit has already shown off their “AI strategic reserve”: a 2 TB hard drive stocked with several model families, including Qwen3.8, Gemma 4, and DeepSeek V4 Flash, with quantized versions at different precisions all ready to go. They’ve also got speech, image generation, embeddings, and an offline Wikipedia, with the most important models copied to a NAS for a second backup.
Of course, there’s currently no evidence that NVIDIA is going to change its open-model policy. The concern is that its champion may be gone—not that the policy has already changed. I think the whole thing is funny, but the logic checks out: cloud services can change prices, be taken down, or cut off access at any time. Once the model weights are on your own hard drive, nobody can take them back. Models that run offline are the ones that truly belong to you.
#AI Frontier
$BTC
