IT News Review by Control F5 Software: The social network that only works 3 hours a day: seven39

This '90s-style social network only runs three hours a day

seven39 is a new social network available only three hours a day. If you try to access it at any other time, you'll be greeted by a blank page. The Verge's journalist Victoria Song describes it as "a social network with a curfew". The basic idea? Limiting online presence rather than encouraging constant engagement.

No advertising, no algorithms. Just one common stream of posts from all users, a minimalist design, a purple background and a 200-character limit. You can upload photos, reply to posts, rate content, follow other users - familiar and simple functionality reminiscent of Twitter, but without the extra elements.

The content is pretty standard: jokes, pet photos, drawings, personal notes. But thanks to daily refreshing, the platform often gives the feeling of stepping into a new mini-forum full of strangers. Many users opt for anonymous usernames, adding a nostalgic-internet touch to the Internet of yesteryear.

Bill Gates published the original Microsoft source code

If you want to see the original source code behind Microsoft, Bill Gates is now making it available. The Microsoft co-founder published it on his blog, Gates Notes, reminiscing about the company's beginnings on its 50th anniversary. Gates has written a lot of code in the past five decades, but he called this code "the coolest code I've ever written."

Distributed a photo of him holding a huge stack of sheets on which the code is printed, Gates wrote that he was inspired by the January 1975 issue of Popular Electronics magazine. The magazine's cover featured an Altair 8800, a revolutionary personal computer created by a small company called MITS.

Apple is preparing "a major change" for its 20th anniversary iPhone

Apple is preparing "a major change" for the iPhone's 20th anniversary in 2027, according to Bloomberg's Mark Gurman.

The overhaul is expected to include a foldable iPhone, a project that's been in development for years, and a new Pro model that will make "more extensive use of the screen." On the other hand, consumers probably won't see radical changes to the iPhone before that time.

Gurman predicts that the iPhone 17 Pro "will look a lot more like the 16 Pro than anticipated." He adds that the front of the iPhone 17 will look similar to its predecessor, though the rear camera will be positioned on "a new panel that spans the full width of the phone."

Microsoft has created an AI-generated version of Quake

Microsoft introduced the Xbox AI era earlier this year, with a new Muse AI model that can generate gameplay. While it looked like Muse was still an experimental project from Microsoft Research, the Xbox maker is now allowing Copilot users to try out Muse through an AI-generated version of Quake II.

The tech demo is part of Microsoft's Copilot for Gaming initiative and includes an AI-generated replica of Quake II that can be played in the browser. The Quake II level is very simple and includes blurry enemies and interactions, and Microsoft limits the amount of time you can play this demo.

Meta's meta tags for its new AI models are a bit misleading

One of Meta's new artificial intelligence models released Saturday, Maverick, ranks second in the LM Arena, a test in which human evaluators compare the models' answers and pick their favorite. But it turns out that the version of Maverick that Meta uploaded to the LM Arena differs from the version made widely available to developers.

As several AI researchers pointed out on X, Meta noted in its announcement that the Maverick in LM Arena is an "experimental chat version." A graphic on the official Llama website also shows that Meta's LM Arena testing of Meta's LM Arena was conducted using "Llama 4 Maverick optimized for conversations."

Microsoft brings Copilot Vision to Windows and mobile for real-world AI assistance

Microsoft is bringing Copilot Vision to Windows and mobile, extending it beyond the web. Copilot Vision was a big part of last year's Copilot redesign, but until now it's been limited to Edge pages to guide users through what they're seeing.

Copilot Vision will now be able to analyze real-time video content from a camera on a mobile phone, allowing the AI assistant to offer tips for plant maintenance or home decorating ideas. A similar feature will soon be available on Windows.

BBC calls for more prominent mention in Apple and Google News

The British Broadcasting Corporation has complained to the UK competition authority, claiming that aggregators such as Apple News and Google News minimize credit for the stories they feature.

Apple Insider reports that the BBC is asking the UK's Competition and Markets Authority to require Apple and Google to provide more prominent mention of news sources. And while the authority's decision would theoretically only target UK publications, any changes imposed on aggregators could affect other publishers.

Exposed AI database reveals misuse of explicit image generator

A massive data leak has exposed the dark side of generative AI technology, revealing how some users are exploiting it to create potentially illegal content. Tens of thousands of AI-generated images, including what appears to be child sexual abuse material (CSAM), were discovered in an unsecured database with links to a South Korean AI company, according to recent research analyzed by WIRED.

The database, which contained more than 95,000 files and 45 GB of data, included AI-generated sexualized images of celebrities including Ariana Grande, the Kardashians and Beyoncé - many of whom were apparently rejuvenated to resemble children.

How Microsoft's AI chief is measuring Copilot's market entry among consumers

As Microsoft CEOs past and present gathered to celebrate the company's 50th anniversary, one of the leaders said he's looking to improve a certain metric to guide his AI strategy.

Mustafa Suleyman, chief executive of Microsoft AI, said its consumer and research division is tracking the usual adoption indicators for the company's AI assistant, called Copilot. These include daily and weekly active users, distribution and usage intensity for Copilot's consumer offerings, he said.

Hackers target Australia's largest superannuation funds in coordinated attacks

Hackers targeting Australia's major superannuation funds in a series of coordinated attacks have stolen members' savings from the largest fund and compromised more than 20,000 accounts, according to a source familiar with the situation.

Cybersecurity Coordinator Michelle McGuinness said in a statement that she is aware that "cybercriminals" are targeting accounts in Australia's A$4.2 trillion superannuation savings sector and is organizing a response across government, regulators and industry.

A new study suggests OpenAI models have memorized copyrighted material

A recent study fuels concerns that OpenAI may have trained its AI models without permission using copyrighted content.

OpenAI is currently facing several lawsuits from authors, developers and other rights holders who allege that the firm used their copyrighted works - including books and code - to build its models. OpenAI has argued that its actions fall under "fair use," but critics point out that US copyright law does not clearly allow the use of copyrighted works for AI training purposes.

UN warns AI could affect 40% of jobs and deepen global inequality

Artificial intelligence is on track to become a $4.8 trillion industry by 2033 - roughly the size of Germany's economy - but its benefits are far from evenly distributed, according to a new report by the United Nations Trade and Development Agency (UNCTAD).

Although AI promises major productivity gains and is spurring digital innovation across sectors, the UN agency is sounding the alarm about the risks it poses to the global workforce. The report, published on Thursday, warns that AI could affect up to 40% of jobs worldwide, particularly through automation.

Smarter and greener AI: How Super-Turing technology mimics the human brain to reduce energy consumption

Artificial intelligence is powerful - but energy-intensive. Modern AI systems, from search engines to chatbots like ChatGPT, require huge computing resources, relying on sprawling data centers that consume gigawatts of electricity. By comparison, the human brain runs on just 20 watts - less than a standard light bulb.

What if AI could be as energy efficient as a brain?

This is what Suin Yi, an assistant professor at Texas A&M University, and his research team are trying to realize. They've developed a revolutionary concept called Super-Turing AI, designed to function more like a human brain - learning and memorizing simultaneously, just like we do.

Meta launches two new Llama 4 models: Scout and Maverick

Meta has officially launched Llama 4, the newest generation of its family of AI models, which now powers the Meta AI assistant on platforms like WhatsApp, Messenger, Instagram and the web.

This launch includes two models:

• Llama 4 Scout: A compact, high-performance model that can run on a single Nvidia H100 card.
• Llama 4 Maverick: A more advanced model, designed to rival OpenAI's GPT-4o and Google's Gemini 2.0 Flash in performance and efficiency.

Both models are available for download directly from Meta or through Hugging Face.

AI startup SandboxAQ attracts Nvidia and Google as investors and raises another $150 million

SandboxAQ, a startup that relies on quantum computing techniques to develop quantitative artificial intelligence models for companies, announced that it has raised $150 million from new investors including Google, Nvidia and BNP Paribas.

The investment increased SandboxAQ's Series E round to $450 million, valuing the startup at $5.75 billion. This brought SandboxAQ's total funding to $950 million, with investors including T. Rowe Price Associates and Breyer Capital.

Papa Johns wants AI to transform the way we order pizza

Papa John's International wants to integrate artificial intelligence technology into the way its customers order pizza through an expanded partnership with Google Cloud, the company announced Thursday.

The company's plan is to add personalized scoring - based on past customer behavior and other contextual factors - to push notifications, marketing emails and loyalty program offers. It will also add a new online chatbot and the ability to order through virtual assistants.

GitHub Copilot introduces new limits and fees for "premium" AI models

GitHub Copilot, the AI code assistant owned by Microsoft, could become more expensive for some users.

Friday, GitHub announced "premium applications" for GitHub Copilot, a new system that imposes usage limits when users switch to AI models other than the core one for tasks such as agile coding and multi-file edits. GitHub Copilot subscribers can still perform an unlimited number of actions with the base model (GPT-4o from OpenAI), but actions with newer models, such as Sonnet 3.7 from Anthropic, will now be limited.

Customers on the Copilot Pro plan ($20/month) will receive 300 premium requests per month starting May 5, GitHub said in a blog post. As for Copilot Business and Copilot Enterprise users, they will receive 300 and 1,000 premium requests per month, respectively, starting May 12-19.

Midjourney launches V7, first new AI image generation model in nearly a year

Midjourney, one of the first AI image generation services on the web, has launched its first new model in almost a year.

Named V7, the model started going live in alpha around midnight ET on Thursday, a week after OpenAI released a new image generator in ChatGPT that went viral for its ability to create Ghibli-style images.

The Midjourney model is not officially optimized for the Ghibli style - at least not officially - but it can still generate aesthetically pleasing work.

Summary made with the help of a monitoring feed provided by Control F5 Software.