00:00Anthropic just removed their most powerful model, Claude Mythos or Fable 5. Just a few days after
00:05its public release, Google just took content creators to next level with the launch of Search
00:10Profile, Luma launched their new model, GenSpark launched Skills, and many more mind-blowing
00:15things happened. And we will be covering all these updates, so stay tuned. Let's start with
00:20the biggest drama of the AI field. On June 9th, Anthropic launched Claude Fable 5. Now, for those
00:26who don't know, Fable 5 is basically the public version of their top secret and most powerful
00:31Mythos model. The one they said was so powerful that they weren't even letting regular people
00:36near it. Fable 5 was Anthropic's most capable model ever released to the public. It could
00:41handle massive coding tasks, think through complex problems, and people who tried it said it literally
00:46felt like a different level of AI. It was so powerful that during a test, it created an
00:51entire Pokemon game and beat it in a single shot. But here's where it gets crazy. Just
00:56three days later, on June 12th, the U.S. government sent Anthropic an emergency directive. The
01:02reason? A security jailbreak was discovered, meaning someone found a way to bypass Fable
01:075 safety filters and access capabilities that were supposed to be locked away. And a China-linked
01:12group was suspected of having already used that exploit. Anthropic had 90 minutes to comply,
01:17and they did the only thing possible. They pulled the model for everyone, worldwide. So as of today,
01:23Fable 5 and Mythos 5 are still offline. There's no confirmed return date. If you were using it,
01:28Anthropic recommends switching to Claude Opus 4.8 for now. The wild part? The day before the ban,
01:34Anthropic CEO Dario Amodei published an essay arguing that governments should have the power
01:39to shut down dangerous AI models. And then 24 hours later, the U.S. government did exactly that to him.
01:45Now moving from the sad Anthropic news, we have something exciting for creators and businesses,
01:50as OpenAI's Codex just got a massive upgrade. You might know Codex as a coding tool, but OpenAI is
01:56constantly pushing it to be an agentic workspace, to establish it as a strong competitor for Claude
02:01Code. For that, they are continuously adding new plugins to the platform, and this time they
02:06launch two plugins, which are game changers for non-developers. First, we have the Creative
02:11Production plugin. This is for marketers and creative teams. You just describe your campaign,
02:16and Codex helps you build the whole thing mood boards, ad variations, product images,
02:21campaign concepts. It connects with tools you already use, like Figma, Canva, and Shutterstock.
02:27So instead of jumping between five different apps, Codex ties your needs all together.
02:32Second, the Data Analytics plugin. This one is for anyone who works with data. You can ask questions
02:38about your business numbers in plain English, and it connects with tools like Snowflake and Tableau
02:42to pull reports and explain what's happening with your metrics. No coding needed. So, data
02:47professionals, your job just got a big enhancement. Next, we have something that every content creator
02:53needs to see. Google just launched Search Profiles. Think of it as your own dedicated page inside Google
02:59Search, where all your content from YouTube, your blog, your social media, all comes together in one
03:04place. Sort of like Linktree. People can follow you directly from that profile, and your content gets
03:10pushed higher in their Google Discover feed. This is a big deal because Google's AI summaries have
03:15been cutting into creators' traffic for a while now, and this feels like Google's way of giving
03:20something back. There's a catch though you currently need at least 100,000 followers to be eligible for
03:25this. And for now, it's only live in the US. But if you qualify, you can claim your profile through
03:31your
03:31Google Knowledge Panel or directly at profile.google.com. For creators who do qualify, this is basically a
03:38free visibility boost built into the world's biggest search engine. Now, let's talk about AI video as
03:43Luma just dropped something really exciting. Ray 3.2 has been released, and the headline feature is
03:49Frame Level Control. What does that mean? Basically, instead of just typing a prompt and hoping the AI
03:55makes something decent, you can now control exactly what happens in each frame of the video. You're not
04:00just prompting anymore, you're actually directing. It also supports HDR output and professional-grade
04:06file formats, which means filmmakers and production teams can actually use this in real workflows,
04:11not just for social media clips. And for the first time, all of Ray 3.2's controls are available
04:17through an API, so developers and creative platforms can integrate it directly into the
04:22tools they're already building with. Speaking of developers, this one is for them, although it's
04:27still worth knowing about even if you're not technical. Moonshot AI, the Chinese AI company behind
04:32Kimi, just launched Kimi K 2.7 Code, and it's a massive open-source coding model, and what makes
04:39it stand out is that it's genuinely multimodal for coding work. You can give it a screenshot, a diagram,
04:44or even a video of a user interface, and it'll generate code based on what it sees. It also has
04:50something called preserve thinking, which means it doesn't forget what it was reasoning about between
04:54conversation turns. Most open-source AI models start fresh each time, but Kimi K 2.7 Code actually
05:01remembers its own thought process, which makes it way better at complex and long coding sessions.
05:06And because it is open-source, developers can download and run it themselves, a strong challenger
05:12to cloud code and codex, in our opinion. After that, if you make any kind of visual content like
05:18thumbnails, posters, social graphics, Ideogram 4.0 is a massive deal. Ideogram has always been known as
05:25the best AI image tool for generating text inside images. Things like logos, titles, quotes on graphics,
05:33stuff that other AI image tools usually mess up completely. And now with version 4.0, they've gone
05:39even further. And the biggest news, Ideogram 4.0 is now open-weight. That means you can download the
05:45model and run it yourself. This is something Ideogram has never done before, and the users are very happy
05:51about it. After launching on June, it immediately became the number one-ranked open-weight image
05:56model in the world. Designers in Blind Tests actually preferred its output over more expensive
06:01closed models. If text inside images matters to your creative work, this is worth trying.
06:07And last but not least, Genspark just launched Skills. If you haven't heard of Genspark,
06:12it's basically an all-in-one AI workspace. It can research, build slide decks, generate images,
06:18make phone calls on your behalf, and a lot more. It's been growing really fast. And now with Genspark
06:24Skills, you can create and save your own reusable AI workflows like you can do on Claude. Think of a
06:30skill as a custom AI assistant built for one specific job. You can make it via a simple prompt or
06:36by
06:36uploading your existing process as a document. Once it's built, you can share it with your whole team,
06:42and you can also browse a community library of skills that others have already made.
06:46So instead of explaining what you want to the AI every single time, you build a skill once and
06:51reuse it forever. And that's the wrap on our big AI updates. From the Fable 5 drama to tools that
06:57are
06:57genuinely making creative and coding work easier, the pace of AI right now is fascinating. If you found
07:03this helpful, drop a like, comment your thoughts, and subscribe if you're new, and stay tuned for the
07:08next video. We also have other channels. Feel free to check them out. All links are in the
07:13description and this channel's homepage.
Comments