Hi there, this is your daily ☕️ Techpresso.
In today's Techpresso:
👓 Apple shelves Vision Pro overhaul to focus on AI glasses
📈 OpenAI is now the world's most valuable private company at $500 billion
🛠️ Mira Murati's Thinking Machines launches AI tool Tinker
🎵 Major music labels are close to AI deals
🚀 New spacecraft aims for one hour global delivery
🎤 Instagram head denies using your microphone to listen
🎁 + 18 other news you might like
🔮 + 5 handpicked research papers and tools
👓 Apple shelves Vision Pro overhaul to focus on AI glasses LINK
Apple is reportedly sidelining plans for a cheaper and lighter Vision Pro, moving employees from that project to instead focus on developing new AI-powered smart glasses to compete with Meta.
The company is creating a first model of smart glasses, nicknamed N50, which will pair with an iPhone without its own display and is planned for a 2027 release.
A second version with its own display is also being developed to challenge the Meta Ray-Ban Display, and Apple is now working to expedite its original 2028 release date.
📈 OpenAI is now the world's most valuable private company at $500 billion LINK
OpenAI finalized a $6.6 billion secondary share sale that establishes its valuation at a record $500 billion, officially surpassing SpaceX to become the world’s most valuable private company.
The transaction provides employees with liquidity by letting them sell stock, a critical retention tool for OpenAI amid an intense AI talent war and aggressive hiring campaigns from rivals like Meta.
Fueled by strong investor demand after the company reported huge growth, the tender offer saw employees sell only two-thirds of the authorized shares, signaling confidence in OpenAI’s long-term prospects.
🛠️ Mira Murati's Thinking Machines launches AI tool Tinker LINK
Mira Murati's Thinking Machines Lab has launched its first product, a flexible API called Tinker that is designed to help researchers and developers build their own custom AI models.
The tool is a managed service granting users low-level control over algorithms and data, while it handles the complex infrastructure required for distributed training on various open-weight models.
To support the platform, the company also released an open-source library called the Tinker Cookbook and has opened a public waitlist after a private beta with several research institutions.
🎵 Major music labels are close to AI deals LINK
Universal Music and Warner Music are reportedly weeks from striking licensing deals with Google, Spotify, and AI startups over how their song catalogs are used by the tech industry.
These negotiations center on how the labels will license their songs for training AI models and for generating completely new music with artificial intelligence tools from various firms.
For compensation, the music giants are pushing for a system like streaming that gives a micropayment per play, requiring AI companies to build software for tracking this specific usage.
🚀 New spacecraft aims for one hour global delivery LINK
Inversion revealed its Arc spacecraft, a new on-demand vehicle designed for the US military to deliver 500 pounds of supplies almost anywhere on the globe in less than one hour.
The company's mission involves pre-positioning Arcs on orbit for up to five years, ready to autonomously deorbit and land with cargo at a desired location within sixty minutes.
To prove its technology, Inversion launched a small spacecraft named "Ray" in January to test its in-house subsystems and perform a deorbit burn using its bipropellant rocket engine.
🎤 Instagram head denies using your microphone to listen LINK
Instagram head Adam Mosseri posted on his account to deny the theory that Meta secretly records your private conversations with a phone’s microphone to show you more relevant advertisements.
Mosseri explained that the powerful recommendation system works by using data from advertisers and showing ads to people based on what users with similar interests also like, not from audio.
Coinciding with this denial, Meta’s upcoming privacy policy will allow it to target ads using data from your interactions with its AI products, creating an even more powerful signal.
Other news you might like
- Musk becomes world's first half-trillionaireLINK
- Amazon’s ‘model factory’ is training the next generation of AI on the tech giant’s own businessLINK
- Microsoft CEO Satya Nadella shakes up his own job, taps veteran exec as CEO of commercial businessLINK
- Once valued at $100 billion, AOL is about to be sold by Yahoo for just $1 billionLINK
- Microsoft launches AI and productivity software bundle for consumersLINK
- Jeff Bezos’ Blue Origin Wins Contract to Take NASA Rover to the MoonLINK
- Intel in early talks to add AMD as foundry customerLINK
- OpenAI staff grapples with the company’s social media pushLINK
- 'Tokenization is Going to Eat the Entire Financial System' Says Robinhood CEOLINK
- Red Hat confirms security incident after hackers claim GitHub breachLINK
- M5-powered iPad Pro breaks cover in GeekBench, scoring 4,133 in single-threaded tests — matches M4 Max and beats every single-core PC chip scoreLINK
- Hulu isn’t going away after all — it’s going globalLINK
- Many workers are using unapproved AI tools at work - and sharing a lot of private data they really shouldn'tLINK
- Japan is running out of its favorite beer after ransomware attackLINK
- Oracle customers being bombarded with emails claiming widespread data theftLINK
- OpenAI’s Sora 2 Copyright Infringement Machine Features Nazi SpongeBobs and Criminal PikachusLINK
- T-Mobile's Starlink satellite service now works with a handful of appsLINK
- Brave browser surpasses the 100 million active monthly users markLINK
Latest research and tools
Immich v2.0.0: a self-hosted photo and video backup solution that allows you to privately manage your media library, which has now reached its first stable release.LINK
Edge264: a minimalist, high-performance software decoder for the H.264/AVC video format that supports resolutions up to 8K.LINK
The Missing Link Between the Transformer and Models of the Brain: this paper proposes that the attention mechanism in Transformer AI models is functionally similar to how the brain retrieves memories to provide context for current information.LINK
Efficient LLM:Bandwidth, Compute, Synchronization, and Capacity are all you need: this paper identifies four main system bottlenecks—bandwidth, compute, synchronization, and capacity—that limit the performance of large language models.LINK
Implementing OpenMP for Zig to enable its use in HPC context: this paper details the work of adding OpenMP support to the Zig programming language for high-performance computing applications.LINK
Want to get the latest news differently? Find us on:
See you tomorrow for a new dose of ☕️ Techpresso!