GPT-6 Astra is trash don’t use it

GPT-6 Astra is exactly why you need to be building a local model A.S.A.P. That's right. I'm talking about AI sovereignty, and if you want to know more about this topic, go ahead and click the save button so you can come back to this. For months, I've been saying that this is the cheapest we will ever pay for AI, and you're probably wondering why I say that. A few months ago, I saw a stat that only 0.3% of the world's population had a paid subscription to something like Claude or ChatGPT. Now, that was roughly four months ago, and so the number's probably increased, but to consider that it's increased beyond a few percent is still pretty crazy. So, why is that important? Well, it means that still a relatively small portion of the world's population is using a paid subscription to Claude or ChatGPT, meaning demand is still relatively low. And that's important because it means because demand is low and supply is high, we should expect a change in this economic trend. Because economic principles tell us that as demand increases and supply decreases, we will see an increase in cost of that good or service. So how does this pertain to OpenAI and Astra 6? Well, last night I prompted Astra to do a code review on my AI Chief of Staff project, specifically to identify vulnerabilities, to identify inconsistencies that might lead to overconsumption of tokens, and for any stale components that might exist. In just 30 minutes, I had already consumed my entire five-hour usage limit, which immediately just blew my mind that it could consume that much tokens that quickly. So I thought it was a great idea to run the experiment again. So I used one of the resets that OpenAI gave us. And again, in about 30 minutes, I had consumed my entire five-hour usage limit. Astra is consuming massive amounts of tokens. And it's honestly a hidden cost lever that most people won't notice because it's not directly attached to a dollar figure. So the question becomes, why is Astra consuming more tokens? A friend of mine, some of you know Northrup Human First, was explaining that it's not like the underlying math that breaks up language into tokens is any different. The more powerful models may have deeper training or specialized data sets, but is that enough to explain consuming an entire five-hour usage limit in 30 minutes? Twice? Or is having these models consume tokens faster how they get everyone to upgrade from Pro to Max or Max to 5x Max to 20x Max? And that's just on the individual scale. On the corporate scale, this could become the main profitability lever for these frontier models. For anyone that's used GPT6 Astra so far, I'd love to hear your experience in the comments. Some of us have a Discord where we collaborate on this kind of thing, and pretty much everybody else in the Discord is telling me the exact same thing, that it's basically unusable because it consumes your entire usage in a very short amount of time. I'm keeping an open mind, but there's a limited amount of things I can use Astra for at the moment, just given the amount of tokens that it consumes so quickly.