After the fallout of Llama 4, Zuckerberg isn't sparing as much as a penny for his new "Meta Superintelligence Labs" as he onboards some of the most high-profile hires in the industry.
Our other major takeaways as follows:
- Microsoft debuts an AI healthcare system that it claims will pave the "path to medical superintelligence"
- Cursor's newly released web app lets developers manage their AI agents in the browser
- Apple is considering OpenAI or Anthropic to be the voice behind AI Siri as in-house solutions struggle
Join us at AI Tangle as we untangle this week's happenings in AI!
|
Mark Zuckerberg's aggressive push for the industry's best is beginning to take shape as an internal company memo made it clear that Meta is moving all of its AI efforts under a new name, reports Bloomberg. Dubbed "Meta Superintelligence Labs," the group will be spearheaded by the recently acquired Alexandr Wang, former CEO of data labeling startup Scale AI, together with ex-GitHub CEO Nat Friedman overseeing Meta's AI products and applied research.
But its not just Wang and Friedman in the team
Along with Wang and Friedman in the helm, the Bloomberg report also revealed that Meta has, to date, managed to lure 11 new AI researchers from its competitors with its multimillion dollar offers, including names previously unknown. It had already poached three of OpenAI's AI researchers earlier last week, then another four later in the same week. However, the Bloomberg list also included Google DeepMind principal researcher Pei Sun, Gemini tech lead Jack Rae, Anthropic engineer Joel Pobar, and many more high-end talents. Only time will tell what Zuckerberg's new team will be capable of after the mixed reception to its previous Llama 4 models and reports of internal struggles.
|
Detect security design flaws at scale (however your code is generated).
Check out the latest from Endor Labs, with their AI Security Code Review: a multi-agent approach for detecting security design flaws at scale.
|
Led by Mustafa Suleyman, Microsoft's AI research arm recently teamed OpenAI's o3 model with what it is calling a bespoke "diagnostic orchestrator" agent, dubbed MAI-DxO, to tackle 300+ New England Journal of Medicine case studies. The combined system solved over 80% of these diagnostically complex challenges, of which practicing physicians managed just 20%, albeit unaided. Microsoft is now positioning MAI-DxO as a complement to clinicians, pushing its cost and efficiency gains in test ordering while downplaying job displacement in its "path to medical superintelligence," as it calls it.
Cursor has introduced a web app that lets developers manage its AI "background agents" directly from a browser, a step beyond its IDE and Slack integrations. Through the usual language prompts, users can delegate coding tasks, track agent activity, and merge completed changes into codebases. Just month, the company had announced that it had surpassed the $500 million ARR mark and reached an adoption rate of over 50% among Fortune 500 companies. With CEO Michael Truell expecting AI agents to handle "at least 20% of a software engineer's work by 2026," Cursor introduced its $200/mo plan earlier to meet demand.
A new report from Bloomberg's Mark Gurman alleges that Apple is considering letting either OpenAI or Anthropic power its AI Siri as its in‑house "LLM Siri" continues to struggle. The company has reportedly asked both to create LLM versions that could then be run and tested on Apple's private cloud infrastructure for evaluation. Apple's AI lead, Mike Rockwell, who was only recently appointed to the role, has previously had his team test Claude, ChatGPT, and even Gemini for handling basic requests in a competition against its own LLM - Anthropic seemed the most promising at the time.
Berlin's data protection commissioner, Meike Kamp, has ruled that DeepSeek "illegally sends user data to China," calling its transfer of German personal information "unlawful" under the EU's GDPR. Kamp also noted that DeepSeek failed to match EU-equivalent safeguards. As such, the watchdog has asked Apple and Google to consider carrying out a "timely review" of the app's compliance, a step that could essentially ban DeepSeek across the EU, not just Germany. Italy has already ordered its removal, and Ireland has sought details on its data handling.
Google is expanding its Classroom suite with a heap of free Gemini AI features for educators and, for the first time, opening its popular NotebookLM to users under 18. Teachers with Workspace accounts gain a new Gemini tab, be it for planning lessons, drafting another email, or helping craft test problems - those with Education Plus also get audio lessons. The update also includes a new tab in Analytics for tracking student performance trends. NotebookLM will offer interactive study guides and audio overviews from teacher‑uploaded materials, including curriculum "Gems," Google's take on mini AI agents.
|
Clarify - Cut out the busywork and focus on closing more deals with Clarify, an AI-powered, autonomous CRM that helps you sell more by taking care of what doesn't need your focus.
Markable - Markable is an AI-enhanced search engine for your bookmarks that quickly and accurately indexes your bookmarks and categorizes them to ensure fast retrieval and a smooth user experience.
Outerbase - Navigate your data with AI - View, edit, query, and visualize your data in one powerful and easy-to-use interface tool for your database.
Caramel - Turn your ideas into campaigns in seconds with Caramel, an AI-powered ad creation and optimization platform for Facebook, Google, Instagram, and more without marketing experience.
|
Can AI Run a Vending Machine? (10-min read)
How plausible is AI as a middle manager? Silicon Valley AI unicorn Anthropic, together with Andon Labs, put theirs to the test in running a vending machine in its own office, detailing where Claude excelled, where it didn't, why it ultimately went wrong, and what the team learned from their experiment.
AI & The Four-Day Workweek (4-min read)
As AI looks to make each individual more productive, many have begun to advocate for the four-day workweek once more. Where could the four-day workweek be applied more easily, and is it time for workers to reap the benefits of technology?
|
| What did you think of this edition? Let us know - your review helps us improve! |
|
|
|
|
|