Open-weight models can be run locally, fine-tuned with proprietary data, and used entirely offline — this is a real benefit to US companies and breeds innovation independent of frontier labs.
CJ explains that open-weight models allow companies to fine-tune internally with trade secrets and run offline, reducing dependence on OpenAI, Anthropic, or Google. ✦ AI generated
CJ · Syntax · 2026-07-30 · original ↗
starts at this moment · 12:07
You can essentially take a model and fine-tune it about specific data about your organization like how your company works, your org chart, maybe specific trade secrets that you don't want to be hosting on some third-party service. And you can do all of that internally with open-weight models. And that's really one of the main benefits and one of the main misconceptions I want to clear up because this is a real benefit to US-based companies. If we can take these open-weight models and modify them so that we can better do our business or whatever that may be and not necessarily have to depend on a large AI company like OpenAI, Anthropic or Google really just do it ourselves. That also breeds innovation. We don't necessarily have to wait for these AI labs to make their own progress in terms of releasing newer and better models. We can take existing open models and then make them work really well for one specific purpose or one specific business case.
verbatim transcript · starts at 12:07
11:47you can change the portion size. You're completely in control. And this is one of the other things that open weight models give you. Not only can we run them ourselves or access them via some thirdparty provider like Microsoft, you can actually fine-tune them. That is tweak the numbers inside of them to make it more purpose-built for what you're trying to do. And we actually see a lot
12:07of large companies or enterprises doing this. So instead of having to be dependent on some other corporation like Microsoft, they spin up their own infrastructure internally and then they run these openweight models and they can fine-tune them so they can better work for their organization. You can essentially take a model and fine-tune it about specific data about your organization like how your company works, your org chart, maybe specific
12:29trade secrets that you don't want to be hosting on some thirdparty service. And you can do all of that internally with openweight models. And that's really one of the main benefits and one of the main misconceptions I want to clear up because this is a real benefit to US-based companies. If we can take these openweight models and modify them so that we can better do our business or
12:48whatever that may be and not necessarily have to depend on a large AI company like OpenAI, Anthropic or Google really just do it ourselves. That also breeds innovation. We don't necessarily have to wait for these AI labs to make their own progress in terms of releasing newer and better models. We can take existing open models and then make them work really well for one specific purpose or one
13:09specific business case. And one of the last benefits of this do-it-yourself type of model is the fact that regular plain old people like you and me can actually run these models ourselves inside of our homes. And so at that point, you don't even have to be connected to the internet. And if you have the right hardware, typically it's very powerful gaming machines, you can actually run these local models inside
13:30of your own home and never even touch the internet. And so if we think about the different ways to access these models, as you can see, there are pros and cons to all of them, but it's not all bad. And that's really what I'm trying to get across here. If you're accessing openw weight models from a Chinese provider, then yes, your data is traveling to China. And if you have