Claude Fable 5 Is The MOST POWERFUL AI!! transcript

This game came from one prompt. This SaaS product came from one prompt. And this entire launch campaign also came from one prompt. I did not tell Claude Fable 5 what files to create. I did not tell it the tech stack. I did not walk it through the steps. I gave it three impossible goals and just let it work. Claude Fable 5 is being hyped as one of the most powerful AI models ever released. Benchmarks are cool, but I don't care if a model is 3% better on a chart.

I care if it can take a real goal and turn it into something I can actually use. So, today, that's what I'm doing. I'm testing Claude Fable 5 with three builds that should be way too big for one prompt. By the end, I want to know one thing. Is Claude Fable 5 actually the most powerful AI I've ever used? Or is this just another launch cycle with great marketing? Here's the rule for this video. I'm not allowed to break the project into tiny steps for Fable.

For each test, I can give Claude Fable 5 one big prompt. That's it. Because that is the real promise of this model is long-running tasks. Can it plan the job, build the thing, catch mistakes, and keep moving without me babysitting every decision it has to take? So, let's start with the dumbest and most unfair test first. Can Claude Fable 5 build a full playable game from one prompt? For the first test, I wanted something people could understand instantly.

So, I gave it this prompt. Build a complete playable browser game from one prompt. It should have a main character, enemies, quests, upgrades, NPC dialogue, a map, a final boss, and a clear win condition. Make it weird, funny, and impressive enough that someone would want to share it with their friend. That's it. Let's run it. All right. Looks like we are finally done uh here. It took 14 minutes to do it. Now, there's nothing more to do but to try it.

All right. Sock and Sorcery, The Quest for the Left Sock, a tale of loss, lint, and revenge. All right, let's uh let's enter this. Let's try it out. All right. Let's see. Where are we? What in the world is this game? Hi, this is me again from the world of post-editing. I just realized I forgot to tell you about the lore of this game. We're playing as Bartholomew, a knightly sock whose left sock has been lost to the laundry.

The sock is held by the dryer, devourer of left socks, who apparently has eaten over 4,000 left socks, and our job is to go through the bosses to reach the dryer, the final boss, kill it, and redeem our left sock. We can kill NPCs. The more we kill them, the more lint we get. The more lint we get, the more we can purchase upgrades. But, that being said, have fun watching me die a bunch of times. My left How did I just There was this option all along.

Oh my god. I have eaten 4,000 left socks. Do you know why only left ones? Neither do I. It is a compulsion. Give sock back. Come then. Let us stumble. Oh, no, I need to fight it. Oh my god. Oh my god. How do I That is very difficult. It's a lot more difficult than it seems. Oh my god. That is a very tough game, I'm not going to lie. Oh my god. All right, and Fable did give me a finished game, but it gave me something that felt like the first playable prototype of an actual idea.

I mean, when you think of AI creating games, I feel like our expectations veer towards it being able to create GTA 6 from scratch with like full >> [snorts] >> insane graphics, but the game it gave me um I mean, although fun and functional throughout the entire thing, which is very impressive, still lacking um sort of I don't know, that impressive look, I would say. That being said, I have actually built games with AI before.

I've not released that video yet, so stay tuned if you want to see that. And the first prototype it would give me was non-playable, full of bugs, uh a lot of missing pieces. So, for it to just one-shot it again with one prompt and have everything not broken or nothing broken is quite impressive in that sense. And by the way, if you'd like to try out this game for yourself, see if you can beat the final boss, I'll include the link in bio.

You'll be able to play the exact same game Claude Fable 5 just made. The next test, however, is one that really matters for me. Can Fable build something useful? For the second test, I want Claude Fable 5 to build a real product, not a landing page pretending to be a product, a working micro SaaS that solves a real problem I actually have. >> [snorts] >> So, I gave it this prompt. Build a working SaaS product from one prompt including a landing page, onboarding dashboard, database schema, user accounts, pricing page, admin panel, analytics, and a realistic demo flow.

The product should solve a real creator or business problem. The problem I wanted to solve is a few YouTube content creation. I would like this app to take YouTube ideas and generate titles, thumbnail concepts, intro hooks, competitor angles, retention risk, and a final recommendation for which version is strongest. You can use any tools at your disposal or any LLMs inside the website. Make sure the website is fully functional and outstanding before you hand it off to me.

You can get as creative as possible. I want this to be the best SaaS product for YouTubers. That's it. That's the prompt. Let's see what it does. Just like that, we have our second build ready. So, let's go ahead and look at it. Wow. Your video's good. Your packaging is why nobody clicked. Analyze my first idea for free. This is very nice by the way because something a lot of people struggle with, especially if you're new to marketing, you don't really have much experience.

This right here is this analyze my first idea for free is a strong selling point, especially at the hero section of a website. Cheaper than one bad upload. This is very good. Very good marketing. Paying for us paying us is cheaper than you uploading a bad video. This is amazing copy, essentially, is what it is. I [laughter] know you guys are watching me geek out over this, but >> [snorts] >> I mean, being someone who works in the marketing industry, being a marketer myself, building or having built a lot of similar products to this before using Opus, using Codex GPT 5.5.

This is amazing copy and I have not asked it to do any of this. So, let's uh let's try analyze my first idea for free. 1 second new analysis burn toast lab studio drop your video idea right here. I test out newest release Fable 5 by building three builds. All right. This is the video right here. 47 left this month on creator. All right. Uh what happens if we add the context? Who is this for? AI curious people like all of us watching this video right now including myself.

Anything else? Uh I'll just keep this as is. Analyze this idea. Building your package. All right. And it keeps past analyses visible. So, it has created a database. Just me seeing this means it has created a database which is insane because previously I'd have to verbally say, "Please build a database where you include blah blah blah." Okay, so it hooked Opus in. It's crazy because Opus used to hook Sonic and it used to see it as its inferior.

Now Fable is hooking Opus in. It's like, "Opus, he's my inferior." Just have this model this LLM run your website uh or your analyses. The idea I test out Claude's newest release Fable 5 is a 52. The idea is fine that but the packaging is sleepy. Six ways to say it. I gave Fable 5 an impossible task. This is a nice title. Claude Fable 5, three builds, one brutal verdict. Uh, I I don't know if I agree. But I see I see what it's done.

It's created a ranking system, which is nice to see. Now, it would have been helpful for me to give it uh maybe YouTube's algorithm. I mean, it could have just looked for it itself. It could have seen reviews online and um sort of tried to understand what works on YouTube, what is viral, what isn't. [snorts] Uh I don't think any of these would be things I would end up actually putting. So, minus one point for that. Number two, thumbnails, concepts that survive the brutal verdict.

So, this is the winner, the three-way split. Three mini app screenshots stacked diagonally, each with a tiny green check or red X. Your hand pointing at the worst one, bright Fable 5 logo glowing on top. This is a very nice thumbnail idea if we're going for this title. So, with this title, this is a great thumbnail idea. So, it wins one point for that. Why it stops the scroll, this is just empty. What the hell? Oh, why it stops the scroll is here.

Then why did it replicate it? I don't know. Anyway, your first 15 seconds. In the next 10 minutes, this brand new Claude model is going to build three apps from scratch. Two of them actually work, one of them broke in a way I did not expect. So, I gave it a task no one should be able to do this fast and watched what happened. Great great intro, great hook. I mean, you guys can also give your feedback right here. Am I being delusional, delulu?

Or is this actually good? What everyone else is doing, so it did a benchmark of the industry. Your lane, skip specs entirely, make it a live build challenge. How does it know me so well already? Where they will leave you, retention, so things to pay attention to where people might skip the video. Ship this package. Honest call, likely to match or slightly beat channel average. I mean, all all in all, this is solid. Let's go to dashboard.

Uh so, dashboard I can run new things, pricing that it create a pricing page. It did. Your current plan, switch to studio. What happens if I switch to studio? All right, it just switches. I don't think it's hooked to a Stripe account, although it does have access to my Stripe API. It could have done that. But, maybe this is because this is a demo. Can I actually sign up? That's what I really care about. Switch to creator.

So, if I do that, okay, it just switches automatically. There's no Stripe link to this. I have to say I am impressed with build number two for many reasons. Number one, because it understands marketing copy a lot better than any of the previous models I've used. It understands what makes a good hero website, a good landing page. What hooks people in, what incentivizes people to actually go from just seeing the website to using it.

Now, the parts where it loses points is this isn't an insane product. I mean, the idea's great, but I gave it the idea. I might be biased. Maybe the idea sucks, I don't know. You tell me. But, the actual implementation of the idea would this be a product I would personally pay for? If the idea was right, absolutely. I mean, I pay for somewhat of a similar product already. So, I'm the target audience here, not just the developer.

But, in its current implementation, this is not a finished product. This is not, "Oh my god, it just built a working SaaS product in one single prompt." Granted, we can go back and prompt it again and again and again. TLDR, is this is leaps ahead of Opus 4.8. Being someone who heavily uses Opus 4.8, I can see myself, my life quality improve drastically now that Fable 5 is out. But, we're still not at a point where you can just type in one prompt and boom, your life dreams and goals and aspirations are achieved and you are now a kajillionaire.

Unfortunately, we're not there just yet. By the way, while it was working on this build, I was playing the game from the first build and I still can't beat the final boss. It's just too damn difficult. But, building one product is just one part of making something valuable. >> [snorts] >> The next question is much bigger. Can Fable build the entire launch around that product? >> For this third test, I want to see if Claude Fable 5 can act like an entire startup team.

Not just an engineer, a analyst, a strategist, a copywriter, a product person, a launch team, and a content team all in one. So, I gave it this prompt. Create a full launch campaign for Thumb Stop, which is what it just built. Uh make the product positioning, the landing page copy, launch video script, X thread, five short-form scripts, email sequence, founder announcement, pricing strategy, FAQ, and launch dashboard, as well as a product launch video, high quality.

Make this a wow factor. Make it feel like a real startup launch. For this prompt, I did prompt it to ultra thing and use ask user questions until you reach clarity. I wanted to sort of give it the same type of prompt I would have given Opus. So, not an unfair comparison. This is how I personally use Claude. And I want to see how it does. So, I'm going to answer the questions and come back once the product is live. Okay, take back what I said.

I literally just completed the game. 28 minutes. I've been playing it while waiting for the other prompts to finish. This is fun and it's it's challenging. I I have to admit I I was wrong. Oh my god, it just finished right now. Literally, while I was playing the game, took exactly the same amount of time. Uh two rendered MP4s in code with promotion. The soundtrack is an original electronic track I composed. So, it actually made the music for the videos.

The metaplay, I ran the launch video's own idea through Thumb Stop. It scored it 74 out of 100. So, your YouTube upload uses the AI's own packaging. It's live on the site. All right, let's start with the founding member offer. So, it's made that. Let's open Thumb Stop. There it is. Yeah. The first 200 creators lock creator at $19 per month forever. Zero out of 200 claimed. Very neat. Claim a founding member and then you'd need to sign up.

All right, let's go to admin. Okay, nice. Nice. Okay, I didn't know that. We have a mission control, an admin mission control that shows us our sign-ups, our users, our monthly recurring revenue. This is a demo billing, obviously. No one has subscribed to that yet. Uh the creator's studio, so on which packages are people, top niches. This is very nice. I've built websites that serve and cater to this audience before.

Uh very similar offerings, albeit not for YouTube. And I can tell you it's taken me multiple prompts to just reach this stage with Opus. Opus is insane. It's not like Opus sucks. Opus is amazing. This is on a whole other level. Now, this is not, you know, product complete, that's it, we call it a day, we go uh uh toast to a good life, but this is leaps ahead on prompt number one from what Opus and where Opus was. It said to try {slash} launch.

Okay, and we can track our launch as well, now that we have a campaign running. So, that being said, let's look at the wow factor, or what is supposed to be the wow factor. It said we had two videos. Uh so, this one and a vertical one. Where can I find them? Where can I find both videos? Can you open them for me? So, we have the first video here, which is the the the full one, uh 1 minute, but I want the vertical one open as well.

There it is. It literally just opened it for us. All right, let's start with this one. Let me center it. Really? All right, let's watch it. >> [music] [music] >> Oh my god. >> [music] [music] >> That is an insane hook. And that is an insane tagline. I did not give it this. Win the click before you film. How bad ass is is that? I'm I'm at a loss for words because I've also used Opus so many other AI agents to create launch videos.

The fact it just made this from one very small prompt that wasn't just focused on the launch video. >> [snorts] >> The prompt was really it was small and it was focused on the entire campaign. It just made this from the first prompt. There's a lot we can fix, I know, but what just happened? This is insane. This is very good. It made the music. It's it's made the creative uh vision for this video and then it executed.

Let's look at the um the vertical one. So, it's 30 seconds long. Okay, it starts with the pain point of the creator, it was the package. [music] There you go, and then it reaches resolution very quickly. It demos the product. It repackaged the long form into a 30-second short form. One third it. >> [music] >> When the click before you sell. Now, I love this. This is insanely good value for what we're getting. I don't think I've ever seen an AI model do this this quickly and this well.

I don't know. You guys comment, what do you think of this? Am I insane or is this actually very good? So, now we have to answer the actual question. Is Claude Fable 5 the most powerful AI I've ever used? My answer is yes. But with a very specific caveat. Claude Fable 5 is not the most useful model for every task. If you need a quick answer, this is obviously overkill. If you need a simple rewrite, this is overkill. But if your goal is so big, messy, and hard to break into clean steps, this is the most powerful AI experience I have ever had so far.

The reason is not that every output was perfect. I mean, the game had rough parts, the SaaS needed real engineering, uh the launch campaign still needed human taste, but Fable did something that felt different. It took a huge vague goal, planned the system, built the first version, audited its own weaknesses, improved the output, and it kept the whole thing aligned with the original objective. That is a massive jump.

Before the hard part was getting AI to produce useful work without it falling apart. Now, the hard part is giving it a goal big enough to be worth the compute. This is why this model feels different. The scary part is not that Claude Fable 5 built an app. The scary part is that it built a game, a product, and a launch system from goals that were barely more than a few sentences. And then it understood the work well enough to make it better.

This is why I think Claude Fable 5 is the most powerful AI I've ever used. If you'd like to watch more videos like this, make sure to subscribe to my channel. Oh, and would you look at that? The YouTube algorithm gods think you'll love this video next. I'll see you there.