Meta attempted to acquire AI video startup Runway. But it failed, before turning to Scale AI – a development that underlines just how aggressively the company is moving to reshape its artificial intelligence strategy.

According to CNBC and Bloomberg, Meta held early-stage talks with Runway, known for its generative video tools, but the discussions did not progress and ultimately dissolved without a formal offer. While Meta declined to comment, sources familiar with the matter said the talks centered on Runway’s technology and talent—an area where Meta is making significant and urgent moves.

Runway, which recently earned a spot on CNBC’s 2024 Disruptor 50 list, has become a prominent player in the AI video space, enabling users to generate and edit video content using simple text prompts. Its tools have found popularity among creators, startups, and media teams seeking faster, AI-driven video production.

While the Runway acquisition fell through, it’s just one in a string of AI-related moves by Meta. In June, the company invested $14.3 billion in Scale AI, a leading data-labeling and AI infrastructure startup. Meta acquired a 49% stake in Scale AI and also hired its founder, Alexandr Wang, along with a small group of engineers.

The investment marks Meta’s most significant capital commitment to AI to date and highlights its shift toward partnering with or absorbing outside companies to bolster internal AI development.

Additionally, Meta has reportedly approached other AI startups this year, including Safe Superintelligence Inc. and Perplexity AI, as part of its effort to keep pace with rivals like OpenAI, Google, Anthropic, and Amazon.

Although Meta was unable to acquire Safe Superintelligence outright, it’s moving to hire the company’s cofounders—Daniel Gross and Nat Friedman—to join Meta’s AI division. Gross and Friedman are expected to work in product development under Wang, further strengthening Meta’s technical talent base in artificial intelligence.

Zuckerberg’s Broader AI Vision

CEO Mark Zuckerberg has been outspoken in recent months about Meta’s strategic focus on AI. The company is actively integrating AI tools across its product portfolio, from content recommendations on Instagram and Facebook to chat assistants in WhatsApp and Messenger.

Meta has also been investing heavily in infrastructure, including building out its own custom silicon and expanding compute capacity to train large-scale AI models, such as its open-source LLaMA (Large Language Model Meta AI) family.

The company’s broader AI ambitions have led to aggressive hiring and infrastructure upgrades. Zuckerberg has stated that Meta intends to lead in both open-source AI development and real-time AI-powered experiences for consumers.

Industry Context

Meta’s increased AI investments come as major tech players scramble for dominance in artificial intelligence. Google, OpenAI, Amazon, Apple, and Microsoft are all pursuing their own distinct strategies—ranging from building foundational models to acquiring startups and forming cloud AI partnerships.

The brief discussions with Runway also reflect a wider trend of AI startups with narrow, specialized capabilities—such as video generation or conversational search—becoming targets of acquisition talks or large-scale investments. Other startups drawing attention and funding include Hugging Face, Mistral, Anthropic, Character.ai, and Inflection AI.

Meta’s talks to acquire Runway may have ended without a deal, but they are part of a clear and deliberate pattern that indicates the company is expanding its reach in AI by targeting both infrastructure providers and niche technology developers.

In a year defined by breakneck AI funding and ethical uncertainty, Cluely—a viral San Francisco startup that built its brand on helping users “cheat at everything”—has raised $15 million in a Series A round led by Andreessen Horowitz (a16z).

The fresh capital, announced Friday by cofounder and CEO Chungin “Roy” Lee, marks a bold endorsement of a product that toes the line between AI innovation and outright manipulation.

Cluely’s technology offers “undetectable” real-time AI assistance by seeing a user’s screen and feeding them contextual answers. Originally designed to help software engineers game job interviews, the tool has since expanded and sanitized its branding after Lee was suspended by Columbia University over its earliest version, known as “Interview Coder.”

Though references to cheating have been removed from Cluely’s website, the company has retained its provocative core identity—and now, a wave of high-profile investors.

“We backed Roy early because he brings a rare mix of vision and fearlessness,” said Bryan Kim, a general partner at Andreessen Horowitz. “He’s a founder with the boldness to rethink what’s possible.”

Profitable and Viral—Despite Controversy

Lee claims that Cluely is already profitable, generating “millions of dollars in revenue” since its launch earlier this year. The company’s valuation has reportedly soared to around $120 million following this latest funding round. Previous investors Abstract Ventures and Susa Ventures also participated in the raise, adding to a $5.3 million seed round earlier this year.

The startup is now preparing for a massive marketing push. Lee told Business Insider that his goal is to rack up 1 billion views across platforms.

“We’ll do pretty much whatever it takes to do that,” he said. To achieve it, Cluely is hiring 50 “growth interns” who will be required to post at least four TikToks a day.

In typical viral fashion, the company debuted with a tongue-in-cheek video showing Lee using Cluely to impress a date—an attempt that was both humorous and unsuccessful. But this shock-based marketing strategy is proving effective in a market where attention often trumps traditional credibility.

Performance Under Scrutiny

Despite its growing user base and strong early revenue, Cluely’s product has drawn skepticism from users and analysts. Several testers have noted significant flaws in its performance: laggy responses, hallucinated facts, and answers that don’t quite fit natural conversation.

Some reviewers describe Slick as a little creepy, and not quite ready for your next meeting.

But as it stands, investors appear more interested in potential than perfection—especially if a product taps into a viral niche and gains traction among younger, tech-savvy users.

Part of a Broader Niche-AI Boom

Cluely is only the latest example of a broader trend reshaping the AI startup landscape. As major players like OpenAI, Meta, and Google dominate foundation model development, nimble startups are capturing investor attention by focusing on highly specialized use cases:

• Hippocratic AI raised $65 million to develop AI agents tailored for healthcare support staff like nurses and insurance processors.
• Synthesia, which focuses on AI-generated corporate training videos, has raised more than $100 million.
• Rasa, a company building AI customer service infrastructure, recently secured $30 million.
• Harvey, an AI tool for legal professionals built on GPT-4, has attracted investment from Sequoia Capital and the OpenAI Startup Fund.

Many of these companies, including Cluely, are capitalizing on a sentiment spreading in venture capital: that the next great AI wave will come not from general-purpose models, but from tools solving narrow, practical problems—or, in Cluely’s case, pushing controversial boundaries with a Gen-Z edge.

The AI frenzy isn’t without backlash. Cluely’s initial pitch—helping people cheat on interviews—prompted an uproar in academic circles and even led to Lee’s temporary suspension from Columbia University. The startup has since rebranded its use cases, but the lingering questions about ethics haven’t gone away.

In that climate, startups like Cluely are thriving. Their formula? A niche product, a loud message, and just enough momentum to grab a16z’s attention.

What’s Next for Cluely?

With the new funds, Cluely is expected to double down on improving its product and aggressively expanding its user base. There’s also speculation that the company may pursue enterprise clients, such as onboarding tools or test preparation services, though Lee hasn’t confirmed such plans.

Time will tell whether Cluely evolves into a serious business platform or remains a flashy, meme-driven tool. However, one thing is certain: in 2025, AI startups that strike a cultural nerve—however controversially—are getting the capital they need to take off. For now, investors seem more than willing to bet that cheating the system might just be the next billion-dollar idea.

Microsoft has officially confirmed that System Restore points in Windows 11 will now be automatically deleted after 60 days, setting a clearer timeline for users relying on the feature to safeguard their PCs from software or configuration mishaps.

System Restore is a built-in recovery tool in Windows that lets users roll back system files, registry settings, and installed drivers to a previous, functional state—essentially offering a safety net when updates, driver installations, or configuration changes go wrong. Personal files such as documents and photos are not affected by the process.

However, the lifespan of those restore points has been a matter of confusion, with inconsistent behavior reported by users since Windows 11 debuted in 2021. While Windows 10 documentation suggested restore points could last up to 90 days, actual retention in Windows 11 was found to be inconsistent—often defaulting to just 10 days, depending on the system configuration and available disk space.

Now, in a new support document tied to the June 2025 security update for Windows 11 version 24H2, Microsoft has finally codified the rule: “After installing the June 2025 Windows security update, Windows 11, version 24H2 will retain system restore points for up to 60 days… Restore points older than 60 days are not available. This 60-day limit will also apply to future versions of Windows 11, version 24H2.”

Why It Matters

System Restore has long been considered a lightweight and quick recovery option, especially compared to full system backups or reinstalls. But its usefulness has always hinged on how long those recovery points are available.

Ten days is too short a window for many users to notice issues caused by bad drivers or failed updates—especially those who don’t use their computers daily. The new 60-day retention window offers a more reasonable timeframe, balancing data recovery flexibility with disk space management.

Windows will still automatically delete older restore points once the limit is reached—either due to the 60-day window or storage capacity constraints.

According to Windows Latest, which first reported on the updated retention policy, the new approach offers more clarity than past behavior: “This will give you multiple snapshots, but Windows will still delete the oldest ones once they exceed the retention window (now 60 days on Windows 11 24H2 by default).”

How to Manually Create a System Restore Point

While Windows does create restore points automatically before certain system changes (like major updates or driver installations), it’s good practice to manually create them as well—especially before installing new software, making system tweaks, or updating drivers.

Here’s how to manually create a restore point:

• Open the Start menu and search for “Create a restore point.”
• This will open the System Protection tab under System Properties.
• Under Protection Settings, select the drive where Windows is installed (typically C:), and click Configure.
• Ensure system protection is turned on for that drive.
• Click Create, name the restore point for easy identification (e.g., Before the driver update), and follow the prompts.

Your manually created restore point will remain available for 60 days unless storage limits require earlier deletion. Windows typically allocates a small percentage of disk space (often around 1-3%) for storing restore points, and once that fills up, older points are removed to make room for new ones.

Now that the 60-day expiration rule is official, experts recommend creating restore points every few weeks, especially for users who frequently install software, tweak settings, or test new configurations. For users with limited storage space, adjusting how much disk space is allocated for restore points may help preserve more recovery options.

Also worth noting: System Restore is disabled by default on some Windows 11 devices, particularly those with limited storage or OEM restrictions. Users should verify it’s enabled under System Protection.

With this clarification, Microsoft brings more transparency to one of its longest-standing recovery features. Though not as robust as full system backups, System Restore remains a vital part of Windows’ defense against unexpected issues—and now, with a known 60-day window, users can plan their protection strategies more effectively.

YouTube is about to dive even deeper into the AI age—and not everyone is excited about it. During his keynote at the Cannes Lions International Festival of Creativity on Wednesday, YouTube CEO Neal Mohan revealed that the platform will soon introduce a new tool that uses Google’s Veo 3 generative AI to create YouTube Shorts from scratch.

In essence, it’s an AI-powered factory line for vertical video content, capable of assembling fully generated Shorts with little to no human input. Google’s Veo 3, revealed earlier this year at Google I/O 2024, is capable of producing high-fidelity, 1080p video up to one minute long from a simple text prompt. The tool uses multimodal deep learning models that allow it to generate scenes, characters, motion, and even stylized visual effects—capabilities that are now being integrated into YouTube’s Shorts format.

“The possibilities with AI are limitless,” Mohan said during the keynote. “A lot can change in a generation. Entertainment itself has changed more in the last two decades than any other time in history. Creators led this revolution.”

While YouTube presents this as a groundbreaking innovation for creators, critics say it opens the floodgates to what many are calling a coming wave of “AI slop”—content so disconnected from real human storytelling that the lines between authenticity and fabrication will become even more blurred.

But Mohan struck an optimistic tone during his speech, per the Hollywood Reporter.

“Communities will continue to surprise us with the power of their collective fandom. And cutting-edge AI technology will push the limits of human creativity,” he said. “My biggest bet is that YouTube will continue to be the stage where it all happens.”

“YouTube has always been the home for creative experimentation,” he added. “Now, we’re giving more people the tools to create content at the speed of imagination.”

However, outside the Cannes auditorium, the sentiment is not as utopian. Many believe that far from empowering creators, tools like these could dilute the creative ecosystem by flooding it with synthetic, low-quality, and indistinguishable content—most of it likely designed to chase views, game algorithms, or mimic trending formats without any human nuance.

The unveiling comes amid ongoing industry tension over the role of AI in media. YouTube has already inked a deal with the Creative Artists Agency (CAA), allowing certain high-profile artists and athletes to control how their likeness is used by AI. This move comes as a partial response to growing concern across Hollywood that generative AI tools could be used to digitally clone actors and performers without consent.

This concern came to a head during the 2023 SAG-AFTRA and Writers Guild of America strikes, where one of the key demands was the regulation of AI in content creation, particularly regarding digital likeness rights and compensation. Though those strikes concluded with concessions on AI clauses, they did not stop the momentum of generative tools being rolled out in commercial platforms like YouTube, Meta, and TikTok.

YouTube’s AI tool announcement also follows broader Google efforts to push generative AI across its services. Veo 3 is the successor to Google’s earlier video generation models like Phenaki and Imagen Video and is part of a suite of new generative tools launched under Google DeepMind and Google Labs, which also include music generation, text-to-image synthesis, and large multimodal models.

For creators who depend on YouTube for income and audience engagement, the tool could be a double-edged sword. On one hand, it lowers the barrier to entry. On the other, it threatens to saturate the platform with auto-generated content that could make discovery harder and widen the gap between top-earning creators and new entrants.

YouTube’s new AI Shorts tool isn’t live yet, but the company confirmed it is currently being tested internally and will be offered to select creators before a wider rollout later this year. It will likely be integrated into YouTube Studio and the Shorts creation suite on mobile.

Many consider this not just a new feature but a signal—a sign that the platform once built by people with cameras and stories is evolving into something harder to define, harder to trust, and possibly, harder to care about.