Anthropic keeps shooting themselves in the foot...

0:00And Anthropic keeps shooting themselves in the foot, and they have no one to blame but themselves. Look at this tweet they just put out. "Starting September 14th, we're permanently raising standard weekly limits in Claude code by 25% for Pro, Max, Team, and seat-based enterprise plans. Until then, the current 50% increase will be in place." This was a word salad to basically tell you they're reducing usage by 17% in a few weeks. And rightfully so, a community note immediately went up that says, "This is them trying to tell you they're raising limits, but in reality, they gave you really advanced math to tell you they're lowering limits." Now,

0:38to their credit, they did do a follow-up tweet where they're like, "Yeah, I I we are reducing limits here." But, they put it in the follow-up tweet, which always gets a fraction of that first tweet. I mean, it it feels like every time Anthropic makes a public announcement or their name is in the news, they're doing something that either makes them look bad or like the entire AI space look bad. I mean, this

0:59is one of many things they've done. They're already the lowest limit of any of the tools. They're the only company that's like putting limits on their best model that's like separate from their subscriptions. They famously were the only AI lab that said, "Hey, you can't use your subscription with OpenClaude" when OpenClaude was the hottest tool on the block. And I think the worst thing they've done is their marketing like

1:19invented the fear-based AI marketing. They've been saying for years, "Hey, our technology is going to eliminate all the jobs. All the jobs are going to be eliminated. Regulate us. Regulate us, please." And then when they get regulated, they're like, "No, no, no, not like that." I think there really is no proof at the moment AI is going to eliminate all jobs. So, going around telling people AI is going to take your job will only make Americans hate AI more. And And we're kind of in this battle for public opinion right now over AI in data centers. And a lot of Anthropic's rhetoric is the reason why

1:50the anti-AI, anti-data center team is winning right now, which is a really bad thing for America overall. Every time Anthropic says AI is going to take all your jobs, all of America's enemies cheer very loudly. But putting the marketing aside, which again I think is horrible, I believe they made some bad bets early on that are leading to a lot of these bad decisions. Dario, the CEO of Anthropic, famously said over the last couple years many times that he's not betting the house on compute and chips. He's bullish on AI but conservative on the data center buildout. And so while Open AI was openly going investing hundreds of

2:29billions of dollars into compute, and SpaceX AI has literally the biggest data center in the entire world in Memphis, Anthropic was saying, "Yeah, you guys enjoy spending all that money. We're just going to keep working on AGI." Well, it turns out that bet was not a good bet at all. It turns out to actually not only train incredible models, which Anthropic's done very good at, but serve those models to customers, you need an absolute crap ton of compute. You need memory, you need GPUs, you need everything. And Anthropic was like, "We'll be conservative when it comes to this stuff." Now they've done nothing but lower limits, have tons of

3:07outages and crashes, and not be able to really serve the best model out there right now, Fable 5, to their customers as much as the customers want to use it. All while at the same time, Open AI is resetting their limits literally every 5 minutes. And SpaceX AI put out Grok 4.6 and gave it just absolutely tremendous amounts of usage. It turns out compute was significantly more important than even the labs who were going all in on compute thought. Now this has led to basically where we are now, where Anthropic continuously having to lower limits, and then that's on top of their marketing that's pissing off a lot of

3:44people. Now I will give them credit, Fable 5 is the best model on planet Earth. It has been for months now, it has been since June. It is like in the age of AI being the top for 3 months is like basically being at the top for like 40 years. So, they clearly have a concentration of talent unlike any company on planet Earth. But, if they had the kind of customer relations OpenAI did and the marketing that like SpaceX AI has, they would be in the lead by far. People would love them and everyone would be using their technology. But, instead they go this route where they just continuously step

4:22on their customers toes. So, I think this means a few things for you. Number one, you need to know which models to use. Usage is going down. They already had low usage. You need to know which models to use. Right now, I think the best model out there is ChatGPT 56. So, the reason why I think it's the best is it's basically neck and neck with Grok 46, but I do believe that ChatGPT Harness, the desktop app, is the best in the business with ChatGPT voice and all that. If vibe coding isn't super important to you and you care more about like the agentic side of things, I think

4:54Grok bot is the best agent out there. You will have to spend a little bit more money on like that Grok super heavy plan, but you get probably the best agent in the biz and a very solid model. The issue with there is they they don't have a fantastic coding harness. You'll probably have to use Cursor. Cursor's good. Cursor's good. But, ChatGPT

5:14desktop app has so much functionality. So, it's really up to you. Do you care more about like the deep vibe coding side of things, then I lean ChatGPT. Do you care more the agent side of things, then I lead Grok. Should you still keep your Claude plan? Well, I still personally believe that Claude fable 5 is the best model on planet Earth. I think it's the smartest model on planet Earth. The issue is is you don't get a ton of usage. So, how do you use it? How do you use a model you don't get a ton of usage? Well, I use it as kind of my

5:41CEO advisor of my entire business. I go to it regularly and like bounce ideas off of it. It's great for that. And so, if you're willing to spend the extra $200, which actually in my opinion, I do think it's worth it to get access to Fable, I think you should do it because it's going to help you strategize on a level that I don't think any other models on planet Earth can do. So, I am keeping my Claude plan. I just probably won't be using their models much for vibe coding. I'll probably be using Grok and ChatGPT a lot more. And then outside of AI, I think you got to invest in

6:12bottlenecks. I think it's never been clearer we are in the biggest compute crisis in the history of mankind. Demand is only going to go up. We don't have enough compute to satiate all of these companies out there right now. This is not financial advice, but I'll tell you what's in my portfolio. Again, not financial advice. This is just the companies I'm investing in at the

6:31moment. I think Micron for memory. Micron's like the number one American memory company. It's basically the only American memory company and the president of the United States is pushing them very, very hard. SpaceX is another company I'm looking at that Elon has the most compute in the entire world. I think they're like 8% of all Nvidia revenue is SpaceX. So, they're going to be good on the compute side. I mean, Anthropic's literally leasing compute from them at the moment. And

6:57then obviously, there's the goat Nvidia. They make the absolute best GPUs in the world. GPUs basically run everything on planet Earth. I know they're at like a $5.6 trillion valuation. But I personally I'm going to keep investing until there are chips in absolutely everything and there's no more human beings fighting wars cuz it's all drones and robots. Until then, there's going to be a huge demand for GPUs. This is not financial advice. I just kind of wanted to give you my thoughts on the companies out there and

7:25the important ones in the bottlenecks. So, what are your thoughts? Which AI labs are you still going with? Which models you still using for which tasks? Let me know down below. Leave a like and subscribe if you learned anything at all and I'll see you in the next video.