Claude Mythos2 Preview is a general-purpose, unreleased frontier model that reveals a stark fact: AI models have reached a level of coding capability where they can surpass all but the most skilled humans at finding and exploiting software vulnerabilities. https://www.anthropic.com/glasswing
For those who still struggle with the meaning of AGI This is what misAnthropic is really telling you: We finna IPO and we need a lot of hype because capital is very limited. We gotta IPO now or else new government administration might not let us later. So lets hype our latest model as super power and very scary and all the idiots on the internet will repost, convincing everyone they will be super rich soon if they invest. Oh, we know the IPO price will crash because all the other billions of dollars already invested are going to cash out.
For several years, the definition of AGI has been continually redefined, downward. The term is almost useless now.
They literally don't have the computing power they need. It's why Sora was shut down (I know, different company).
I think Sora was a bit of a money pit. It started out somewhat popular, but the popularity started dwindling a bit and never ramped up to cover the costs of the infrastructure required to keep it a service. That latter part I think is basically what you're saying, but if people had been gung-ho about it, and they could keep infrastructure ramped, it could've stayed. But if it isn't making them money? This stuff is too expensive to maintain.
Versus all the comparisons to, like, Terminator movies, I wonder if a more probable worst-case scenario would have AIs causing havoc but not in a united fashion. Like, one rogue AI ****s up software system A, and another rogue AI shuts down the power grid on the east coast, while a third poorly-controlled AI creates a stock panic, etc. And these things could just start bubbling all the time.
Mark Zuckerberg reportedly working on AI clone of himself — Meta insiders claim 3D photoreal animated Zuck will be able to engage with employees on his behalf The AI clone was trained on the Meta CEO's public statements and strategies, and will respond in the CEO’s voice, including his mannerisms. The very top job at Meta may periodically be delegated to an AI replica of Mark Zuckerberg. Will anyone notice? The FT reports that Meta is developing 3D, photoreal, AI-powered characters that users can interact with in real time – and the project has recently pivoted to prioritize an AI clone of the company CEO, according to three unnamed insiders. Meta wishes to ‘dogfood’ its AI wares to gain a competitive advantage, and it shows confidence that this thrust extends to the very top echelons of the company. Making an AI-generated Zuck clone is something of a pivot, as we mentioned in the intro. The FT says that Meta was busy with a project in which it was building a ‘CEO agent’ to support top execs like Zuckerberg day-to-day. However, this CEO-cloning project is separate and has become a priority, according to the report. Perhaps the boss wants to go on an extended holiday soon? The training of the now-prioritized Zuckerberg AI character has been shepherded by the Meta CEO. “The Meta chief is personally involved in training and testing his animated AI,” reports the FT. And the character “could offer conversation and feedback to employees, according to one person.” As well as looking like the real Zuck, thanks to the 3D, photoreal, animated character that has been created, much deeper work is being done. The source report notes that the Zuck AI has been trained on his publicly available statements, his recent thinking on business strategy, and so on. That should provide a solid foundation for day-to-day Zuck-a-like reasoning. But the clone even replicates Zuckerberg’s mannerisms and tone, and will respond in the CEO’s voice, it is claimed. The above initiative is part of Meta’s multibillion-dollar personal superintelligence push, which is hoped to help it compete better with the likes of OpenAI and Google. Employees are also being encouraged to use AI tools and agentic systems based on things like OpenClaw. Meta’s prior AI character/chatbot work hasn’t been without issues and pratfalls. Users generating overtly sexualized characters prompted Meta to rein in access to its AI Studio character workshop at the start of 2026, for example.
Anthropic launches Claude Design, a new product for creating quick visuals Anthropic announced on Friday that it’s launching Claude Design, a new experimental product that lets users create visuals like prototypes, slides, one-pagers, and more using Claude. The company says Claude Design is intended to help people like founders and product managers without a design background share their ideas more easily. With Claude Design, users describe what they want, and Claude will create an initial version. From there, users can refine the visuals with direct edits or requests. For example, you could ask Claude to “prototype a serene mobile meditation app. It should have calming typography, subtle nature-inspired colors, and a clean layout.” You could then tweak the colors, the size of the typography, or ask Claude to add a dark mode toggle. While Claude Design may initially seem like it’s looking to compete with popular design app Canva, which has just expanded its own AI capabilities, Anthropic told TechCrunch that it’s intended to complement it rather than replace it. The company said its new product is built for people who aren’t starting from a design tool and need to get from an idea to something visual quickly. Once teams create presentation decks or prototypes, they can export them as PDFs, URLs, PPTX files, or send them to Canva. Once in Canva, they are fully editable and collaborative, Anthropic says. Claude Design can also apply a team’s design system to every project it creates so that the results are consistent with the company’s overall visual style. Anthropic says Claude Design is able to do this by reading a company’s codebase and design files. Additionally, teams can refine these components and maintain more than one design system. The new product is powered by Claude Opus 4.7 and is available in research preview for Claude Pro, Max, Team, and Enterprise subscribers. The launch highlights Anthropic’s ongoing push into the enterprise and prosumer categories, as competition intensifies around AI workplace tools. In January, Anthropic rolled out Claude Cowork, an agentic assistant built for complex tasks. A few weeks later, the company brought agentic plug-ins to Cowork that are designed to automate specialized tasks within a company’s various departments. Today’s announcement comes a few days after Bloomberg reported that VCs have been offering the company a preemptive funding round that would value it at $800 billion or more, which would almost match or even surpass its rival OpenAI. But so far, Anthropic isn’t interested in the latest offers, according to the report.
