Claude Opus 4.8 + Claude Code is Completely INSANE (BEST LLM FOR CODING?)

Okay, guys. So, huge news in the AI space today. This is Opus 4.8 basically just dropped. I think they're doing this because a lot of people are complaining about Opus 4.7. Like, all I ever saw was negative things about Opus 4.7. I've personally not had a single problem with Opus 4.7, but a lot of people have had issues with it. Um one of the main pieces of news that's dropped is that fast mode is three times cheaper than it was. And obviously, you know, the agenda, the benchmarks, the benchmark thing. We're all familiar with this, guys. I'm testing this in the background right now. I'm going to show you my tests in this video. And I want to actually know if this is good or bad. One of the most prominent improvements in Opus 4.8 is its honesty. We train all our models to be honest, for instance, to avoid making claims that they can't support. But a general problem with AI models is that they sometimes jump to conclusions, definitely confidently claiming they've made progress in their work despite the evidence being thin. Yeah, this is I mean, if they can improve this, then that would be a massive update. I I give you an example of this. Just the other day, uh we pushed an update to Harbor, like Harbor API, and I noticed that someone had generated 55 articles using the Harbor API when they had zero credits. So, I asked support to look at it. Claude said the code was perfect. I asked it to look at it again, this time more carefully, specifically pointing to the fact that the person had zero credits. So, how were they even generating articles? And it realized that it wasn't checking whether or not the person had credits before generating articles, right? So, this is a that's a perfect example of that. That was Opus 4.7. So, if they have improved that, I would love you know, I would love that, but I really really doubt that there's been a big improvement there. So, misaligned behavior, apparently, um Opus 4.8 is as good as Mythos uh preview. So, that's going to be interesting. Also launching today, dynamic workflows, which people are saying are really really strong. Um So, this new feature available in research preview preview Claude to take on even bigger tasks in Claude code, only if you have the usage. Claude can plan the work and then run hundreds of parallel sub agents in a single session. I mean, this is not useful to anyone except like enterprise users or people using the API because you're not going to be able to get much done with just the max plan. Uh so, I mean, this isn't This sounds interesting, but it's not going to be that useful for people, to be honest with you. Effort control, so you can use less usage by basically making it think less. And then the messages API now accepts system entries inside the messages array. Okay, you guys. So, right now I have this model running inside Claude code. By the way, guys, I'm a complete convert from terminal to uh ADE, I think people are calling them. So, agentic uh development environment. This is by far the best ADE, by the way, the Claude desktop app inside it, Claude code, right? Not I use code work if I need it to control my computer, but I use code if I need it to, you know, be programmatic or whatever. I think most people though are waiting for Mythos, like this big moment, but honestly, I've been saying this for a while, you don't need better models. You already have amazing models that can code for you. The other main thing, guys, is that the price hasn't changed. The only thing that's changed is the fast pricing, which I mean, I don't think many people are using fast mode anyway. So, let's take a little Wow, let's take a little look here. This is the classic test that I always do. You guys are probably very, very familiar with it. This looks amazing, I have to say. This is very, very high quality indeed, UI UX-wise. Um let's just see if everything works as intended. It looks like this is working perfectly as well. I mean, it's pretty standard that this benchmark that I do has been blown out of the water by the last few models, but this is very, very impressive. This is probably the best I've seen. The SVGs, amazing. Uh the SEO is all done. You can switch between Italian and English without it 404ing, which is actually an improvement because most of the time that doesn't work. We have one vertical block, a horizontal block, two vertical, horizontal three, four, five. So, we have five horizontal blocks. The prompt says five to seven. All of the SEO has been done, like I said. So, if you hover the meta title at the top here, you can see the meta title has been done really, really well. Uh this is honestly amazing. Um yeah, there's not really much to say except wow. Very, very good quality. I was expecting this, obviously. This is a marked improvement on 4.7, I would say though. 4.7 didn't do that good of a job, to be honest with you. Um yeah, I mean, it's done all the pages, as you can see. There's not a single 404 here, so. It even has like a little um little fact for each um place, which I've never seen any other model do. Like I said, the SVGs, amazing work. Um and honestly, I'm very, very impressed by this. Has the really uh really good FAQs right here. Uh it's it's it's very, very good, guys. The the the design is really, really premium as well. Okay, guys, just before going on to the next test, quick shout out from our sponsor me. This is Income Stream Surfer, my own website. If you need a website building, if you need a Shopify website building, if you need a rebuild or a second website, whether that's e-com, service-based website, anything, come and talk to us today. There's a link in the description of this video and in the pinned comment. I'm currently building a lot of websites and, you know, the website I'm about to show you is an example of the kind of website that we're building for people right now. I know a lot of people think that paying for a website right now is crazy because you can just build one on lovable or whatever, but a website you build on lovable is not going to be as good as a website that I build in terms of SEO and much more. So, there's a link here. You can just press book a call right here. Come and talk to us today, guys. Let's jump back into the video. Okay, so this is the website that I built using my Astro builder skill, which I'll leave a link to in the description of the video as well. It's an open-source skill that I use to build websites, right? What is going on here? There's just a big white space. Not sure about that. Let's see. Okay, so I'm not actually sure what's going on here. There's a huge white space here. It looks like something's not rendering properly, but that's fine. That can happen. This is actually the first time ever that the contact page has just been perfect first time. The quality of the design is okay. There's obviously some kind of issue on the homepage. So yeah, this is something I haven't seen before. But yeah, really really clean design like that when you just open the website. I do like that. But again, not not the best I've seen actually, to be honest with you. It is interesting the gap between American models and Chinese models is definitely definitely getting smaller. And that's never been more true, honestly. Like even today with a release like this, Quinn 3.6 max or 3.7 max, whichever one it was, could probably do as good a job as Opus here. But yeah, overall, I mean, pretty pretty good build. Let's see if it did the admin dashboard. It did. Let me get the password. Okay, so yes, the admin dashboard has also been created. If you didn't know, by the way, guys, we actually use Opus 4.7 inside Harbor. So this hasn't been released to the public yet, but I'm working on a website builder inside Harbor that basically uses this skill, which I've shown you today. Inside Harbor build to build Astro websites for people, SEO-ready websites. Just haven't got around to releasing it just yet, but it should be released very very soon. I'm going to go and test Opus 4.8 inside Harbor build instead of Opus 4.7 and see the difference between those two. If you didn't know guys, by the way, Harbor SEO.ai, which I'll also link leave a link in the description, too. It's my AI content generator. It does more than just AI content, by the way. It's a full-blown AEO and SEO tool at this point. But currently, we have 641 pages published, 380,000 impressions, and 3,300 clicks, which I'm super happy with, super proud of. We just released Harbor Dev or Harbor API, which means you can connect it to NA10 and many, many more things as well. You can basically connect it to Cloud Code as an MCP, whatever you guys want. We also have agency co-pilot, which means, you know, you don't even have to click buttons anymore. You can just say do this for me and it will do it for you. I've been generating content for clients recently using this method and it works extremely well. Okay, finally. So, this has been fixed. This is the rest of the home page, which was unfortunately hidden. Um I mean, it's good. It's It's actually not that crazy, to be honest with you. I was expecting more, but that might be the skill, right? So, the skill has a specific way to design websites. So, it's using that skill rather than its own um volition, its own like its own knowledge and stuff. Whereas, the first example I showed uh with the local with um the Rolls-Royce website was actually it using its own design. So, this is probably it just following the skill more than creating like bad design or whatever. But still, overall, extremely well done. Uh it had It actually hasn't created the individual pages, either, which is also unfortunately is going to have to be a minus uh in my book because this is the whole This is like 90% of the SEO of this website is is uh are these individual service pages, and it hasn't done them. So, yeah, I'm going to have to give minus points for that as well. Honestly, I'm I'm happy, but I mean, it's just a very, very slight improvement in my opinion. Look past the hype, guys. And yeah, I mean it I've been saying this for a while. In my opinion, we already have the models at a point where you can build anything, right? It's just you need to know how to build something. So, these slight upgrades or whatever, that I'm not really bothered about them anymore. Like, I used to be so hyped for a new model release, but now it's just like you know, whatever. I'll just I I would have Everything I've shown you in this video was already possible with Opus 4.7. Everything I'm already doing was already possible with Opus 4.7. So, like a small increase, great, thumbs up for me, but not really that important. Don't really care. And I'm just going to keep building, and I will be using Opus 4.8. Let's just see the usage. So, 5-hour limit, 12%, weekly, 3%, and I made two websites. I'm on the $200 plan. Guys, I'm going to leave the video there. Thank you so much for watching. If you are watching all the way to the end of the video, you're an absolute legend. Go and check out Harbor SEO.AI or if you need a website building, go and check out my website, Income Stream Surfer. Links in the description. Thanks for tuning in. Peace out.