Can you ACCESS that same AI that just got BLOCKED?! This leads us to a BIG QUESTION

You might be able to access AI so powerful that the US government just banned it. I think you all know what I'm talking about. But the reason why we're able to do it leads us to some questions that could change the way you see this space forever. I'm Gigaqian and Fable 5 is still off limits, but there might be a workaround. So of course I'm going to tell you about it. Rapid fire catch up. The US government got a bit spooked because there was a bit of a jailbreak and they thought foreign nationals might use tech to do bad things. So the Department of Commerce issued an export control directive under Secretary Launick. This is so strict that Andrej Karpathy, the expert who famously left OpenAI to go to Anthropic, can't even touch Fable 5. And Anthropic can't build a passport check for an API overnight. So they just shut the whole thing down globally. Within 48 hours, indie devs were already having serious success recapturing some of Fable 5's capabilities. That's because someone leaked the models' entire 120,000 character system prompts to GitHub. And if you've been following me, you probably have a guess as to who that might be. That's right, it's our friend Pliny. Now that repo, Claritas, was more about intellectual intrigue and research purposes until it wasn't. That means the magic wasn't all in the weights, it's also about the prompt. So the performance of one of the most hyped models of all time is largely reproducible by copy-pasting structural instructions into an older, cheaper engine. If all these claims hold up, we have three huge takeaways. First, system prompts are indeed the crown jewels. They're already being extracted at scale. Claritas already has you covered for Google, OpenAI, Meta, and Anthropic. Next, and more controversially, we have the ban itself. They knew to block us off from the weights, but no one considered to cut us off from the cognitive recipe that actually tells the model how to act. This is all easier said than done, of course. And 90% of the performance is speculative. It could also just be 90% of the vibe, cringes that may be. But here's the thing. If AI companies are shaping corporate power, the future of the world through hidden prompt scaffolds, then what would you call extracting those prompts? I know a lot of y'all are gonna call me AI or say this was written by AI, but it still has to be said. It's not hacking. It's journalism.