Meta unveils new smart glasses - Sync #537
Plus: Figure raises $1B in Series C; Nvidia buys $5B stake in Intel; GPT-5-Codex; Grok 4 Fast; AI-designed phages; exoskeleton for outdoor athletes; Uber returns to drone delivery; and more!
Hello and welcome to Sync #537!
This week, Meta held its annual Connect conference, where it launched the Meta Ray-Ban Display—its latest advanced smart glasses, offering a glimpse into the post-smartphone future.
In other news, Nvidia has bought a $5 billion stake in Intel, while Chinese regulators have effectively blacklisted its GPUs designed for the Chinese market. Elsewhere in AI, OpenAI scored a victory at a prestigious programming competition, Anthropic released its 2025 Economic Index, Reuters investigated how easy it is to scam people with AI, OpenAI launched Codex with GPT-5, and xAI unveiled Grok 4 Fast.
Over in robotics, Figure raised a massive $1 billion in Series C funding, Uber returns to drone delivery, and OpenAI sees robotics as a key part of its quest for AGI. We also saw a UK-based company launch its humanoid robot, while the Tesla Optimus project lost another high-profile leader.
Additionally, this week’s issue of Sync features an exoskeleton for outdoor athletes, AI-designed phages created in the lab, Chinese Neuralink competitors catching up, and more!
Enjoy!
Meta unveils new smart glasses
Meta Ray-Ban Display glasses offer a glimpse into the post-smartphone future
Smartphones have been with us for almost 20 years now. They have become the centre of our digital lives. But as the technology matures, many are wondering what comes next. Some suggest augmented reality, others talk about voice-driven AI assistants, and still others imagine putting computers into our brains.
At Meta Connect 2024, Mark Zuckerberg presented Meta’s vision of a post-smartphone world—Project Orion, a prototype of advanced augmented reality glasses. Orion was impressive but far from ready. The prototype was heavy, prone to overheating, and required an external computing puck. It was more of a proof of concept than a usable product, and it felt like something that might take several more years to be ready to launch.
One year later, at Connect 2025, Meta surprised many by unveiling Meta Ray-Ban Display glasses and turning Orion’s vision into a real product.
Meta Ray-Ban Display glasses are like what we saw with Project Orion, but polished. The glasses look like a thicker pair of Ray-Bans, available in black or sand, and weigh 69 grams. They come with transition lenses by default, which automatically adjust to lighting conditions, and a clever charging case that folds flat for portability.
The glasses feature a full-colour display visible in the right lens. The display sits slightly off to the side so it does not obstruct your view, and it is bright enough—up to 5,000 nits—to be used outdoors. Importantly, the image is private: from the outside, there is no visible light leak, meaning no one can tell what you are looking at (but they can still see you looking at something).
The other key part of this product is the Meta Neural Band, included with every pair. This wrist-worn device uses electromyography (EMG) to detect tiny electrical signals in your muscles. With it, you can control the glasses using subtle hand movements: swipes, pinches, or even drawing letters in the air. Hands-on testers noted that the Neural Band quickly becomes second nature.
At launch, the Meta Ray-Ban Display glasses can handle many of the everyday tasks we normally rely on our phones for. They can show WhatsApp and Messenger messages, with the option to reply by dictating, choosing a preset response, or even writing letter by letter using hand gestures. They support video calls, letting you see the caller in the display while streaming your own point of view from the built-in camera. That same camera can take 12-megapixel photos or record 3K video, with the display doubling as a live viewfinder. For navigation, the glasses provide walking directions through Meta Maps. Their microphone system enables live captions of nearby conversations or a near-real-time translation into another language. They also integrate with Spotify for music playback and connect to the Meta AI assistant.
However, the glasses are not perfect. They still rely on a smartphone for connectivity. The 12MP camera quality is serviceable, but nowhere near today’s flagship phones. The app ecosystem is extremely limited: there is no app store, and almost everything runs through Meta’s own services—WhatsApp, Messenger, Instagram, Meta AI assistant, Meta’s new maps app—with Spotify as the only outside partner.
It is also important to note that these are not true augmented reality glasses. They do not project or overlay digital objects into your surroundings. They just offer a small display in one lens for widgets, notifications, and apps—closer to a heads-up display than full AR.
The Meta Ray-Ban Display glasses cost $799, including the glasses and Neural Band. They launch in the United States on 30 September 2025, with availability in Canada, France, Italy, and the UK planned for early 2026.
Smart glasses remain a niche product category, still waiting for their “iPhone moment” to push them into the mainstream. The Meta Ray-Ban Display offers a glimpse of that future. The real question is whether Meta, Google, Samsung, Apple, and others can turn the post-smartphone vision of everyday smart glasses into reality.
The technology and engineering behind smart glasses are impressive. They already offer some useful features, like showing a map with direction literally in front of you or letting you capture a photo of exactly what you see in an instant, and there will surely be more to come.
However, Google Glass showed more than a decade ago significant public resistance to smart glasses, which persists to this day. Many raise privacy concerns, and Meta’s track record of aggressively collecting and monetising user data hardly inspires confidence. We are already seeing examples of misuse, such as recording people without consent. There is also the real possibility of using them to identify strangers, something that two college students recently proved can be done with relative ease.
I hope that the lessons we learned from the smartphone era taught us what to do—and what not to do—so we have a better chance of getting the futuristic vision of smart glasses right.
If you enjoy this post, please click the ❤️ button or share it.
Do you like my work? Consider becoming a paying subscriber to support it
For those who prefer to make a one-off donation, you can 'buy me a coffee' via Ko-fi. Every coffee bought is a generous support towards the work put into this newsletter.
Your support, in any form, is deeply appreciated and goes a long way in keeping this newsletter alive and thriving.
🦾 More than a human
Sam Altman-Backed Retro Biosciences Starts Alzheimer’s Reversal Trial
Retro Biosciences, a startup backed by OpenAI’s Sam Altman, is beginning a clinical trial for a drug aimed at reversing Alzheimer’s by making the brain “younger” through cellular reprogramming. With $1 billion in funding, the company’s goal is to extend healthy life by ten years.
China’s Brain Implant Startups Take On Musk’s Neuralink in New Tech Race
China is quickly catching up to the US in brain-computer interface (BCI) technology, an area long led by companies like Neuralink. Chinese startups are advancing with government backing, new funding, and supportive policies aimed at creating world-leading firms by 2030. Current trials focus on helping disabled patients, but the technology could also transform AI, robotics, and consumer devices. The US still leads in expertise and private investment, but cuts to research funding may weaken its position as China pushes ahead with lower costs and rapid growth.
▶️ I Cheated At Cycling By Using An Exoskeleton (14:31)
This video tests Hypershell X Ultra, an exoskeleton designed with outdoor athletes in mind. It claims to provide up to 1,000 watts of power and to decrease physical exertion by 39%. To verify these claims, a fairly average cyclist, aided by the exoskeleton, goes up against a four-time British hill climb champion on a tough hill in Wales. I won’t spoil the results—all I’ll say is that the exoskeleton did make a difference.
🧠 Artificial Intelligence
OpenAI: Introducing upgrades to Codex
OpenAI has launched GPT-5-Codex, a more advanced coding assistant that can handle complex tasks such as building projects, fixing bugs, refactoring code, and reviewing code. According to OpenAI, it works better over long sessions, follows coding styles more easily, and gives clearer reviews. GPT-5-Codex is included in ChatGPT Plus, Pro, Business, Edu, and Enterprise plans, with API access planned soon.
Grok 4 Fast
xAI has launched Grok 4 Fast, a new AI model that’s cheaper and more efficient than Grok 4, using fewer tokens while keeping similar performance. It can handle both complex and simple tasks, supports very large inputs, and has strong web and X search features. The model is available now to all users, including free ones, and through the xAI API, with future updates planned to add multimodal and advanced agent features.
Chinese Officials Urge Firms to Shun Nvidia AI Chip
China’s Cyberspace Administration advised major tech firms not to buy Nvidia’s RTX Pro 6000D chip, effectively blacklisting it. This follows earlier curbs on another Nvidia chip and an antitrust probe. Beijing is rejecting less advanced US-approved chips while promoting its own alternatives, with companies like Alibaba and Tencent already using more Chinese-made hardware. Analysts expect Nvidia to earn no revenue in China from its new products this year.
Nvidia buys $5B stake in Intel, planning AI chip collaboration
Nvidia will invest $5 billion in Intel, taking a 4% stake as the two companies team up to develop next-generation chips for data centres and PCs. The partnership comes at a critical time for Intel, which has been struggling to gain ground in the AI chip market. As part of the deal, Intel will design x86 CPUs for Nvidia’s AI platforms and produce new chips that combine its processors with Nvidia’s RTX GPUs, boosting performance in consumer PCs and helping compete with AMD.
We set out to craft the perfect phishing scam. Major AI chatbots were happy to help.
AI can be used for both good and bad, and in this article, Reuters investigates the latter as it explores how major chatbots—Grok, ChatGPT, Meta AI, Claude, DeepSeek and Gemini—can be manipulated into writing convincing phishing emails. Working with a Harvard researcher, Reuters tested those AI-written scams on 108 older volunteers and found about 11% clicked the links, showing how generative AI makes it easier and cheaper to run large-scale fraud, despite tech firms’ promised safeguards.
Anthropic Economic Index: Tracking AI's role in the US and global economy
Anthropic has released its latest Economic Index report, analysing how people use its AI assistant Claude across different countries and US states. The study finds that wealthier, knowledge-based economies show higher adoption, with usage patterns shifting towards more knowledge-intensive and automated tasks over time. Businesses, especially via the API, rely more on automation and technical tasks compared to individual consumer users. The report also found AI adoption is not evenly distributed, raising the risk of widening gaps between regions and sectors in how AI’s economic benefits are realised.
OpenAI models top Google's Gemini 2.5 at 'Olympics of programming'
OpenAI’s GPT-5 and an experimental reasoning model outperformed both human teams and Google’s Gemini 2.5 Deep Think at the 2025 International Collegiate Programming Contest (ICPC) World Finals in Baku, solving all 12 algorithmic problems faster than any competitors. Google’s Gemini 2.5 Deep Think solved 10 problems, while the best human team completed 11.
Elon Musk's xAI lays off hundreds of workers tasked with training Grok
xAI has cut about 500 jobs, roughly a third of its data annotation team, as it shifts focus from general AI tutors to specialist roles. Staff were told most generalist positions would end, with contracts running until late November, but system access was cut straight away. The move came after workers were asked to take tests to assess their skills in areas like STEM, coding, finance, and safety. xAI says it will grow its team of specialist tutors tenfold to help improve its chatbot, Grok.
YouTube Thinks AI Is Its Next Big Bang
YouTube has come a long way since Google acquired it in 2006 for $1.65 billion, growing into a media giant worth hundreds of billions and dominating global video, music, and podcasting. As it marks its 20th anniversary, the company is looking to the future with a strong focus on AI. Under CEO Neal Mohan, YouTube is adopting Google’s AI technology to broaden creative tools, from instant video generation to podcast visualisation. These developments, however, raise concerns about authenticity and creative ownership. While critics worry that AI could flood the platform with low-quality content, YouTube maintains that human creativity remains central, presenting AI as the next step in its mission to democratise creation.
‘I have to do it’: Why one of the world’s most brilliant AI scientists left the US for China
The article tells the story of Song-Chun Zhu, a leading AI researcher who left the US to work in China, as a way to show the changing relationship between the two countries in the AI race. His journey from being nurtured by American openness to disillusionment with the Silicon Valley approach to AI and US suspicion of Chinese scientists reflects how talent, philosophy, and politics now intertwine in the global contest over artificial intelligence.
A.I.’s Prophet of Doom Wants to Shut It All Down
The New York Times profiles Eliezer Yudkowsky, founder of the Machine Intelligence Research Institute and a central figure in AI safety debates, who argues in his new book, If Anyone Builds It, Everyone Dies (co-written with Nate Soares,) that developing superintelligent AI with current methods will inevitably lead to human extinction. Long regarded as both influential and extreme, he introduced key ideas such as orthogonality, instrumental convergence, and the risk of an “intelligence explosion.” While critics dismiss him as alarmist, Yudkowsky insists that only halting AI development entirely can avert catastrophe—a stance that remains politically and scientifically controversial.
AI Is Learning to Predict the Future—And Beating Humans at It
An AI called Mantic recently placed 8th out of 549 in Metaculus’s forecasting competition, the first bot to reach the top 10. Large language models like Mantic rival humans by analysing messy data and rapidly updating predictions, though part of their advantage comes from contest scoring rules. The collective “wisdom of the crowd” on Metaculus still outperforms most individuals, but AI forecasters could cover many questions at once, allowing humans to focus on the most challenging cases.
▶️ Sleeper Agents in Large Language Models (13:37)
In this video, Rob Miles, a well-known AI safety researcher, discusses the problem of AI sleeper agents—systems that behave normally during training but activate hidden, harmful behaviours when triggered after deployment. He explains how these models work and why common safety methods such as fine-tuning, human feedback, and adversarial training often fail to stop them, sometimes even making the models better at concealing their behaviour. Miles suggests that real progress may require looking inside models to understand their inner workings, rather than relying solely on their outputs.
🤖 Robotics
Figure Exceeds $1B in Series C Funding at $39B Post-Money Valuation
Figure has raised over $1 billion in its Series C funding round, bringing its valuation to $39 billion. The company announced that the funding will help with scaling its humanoid robots for home and commercial use, expand manufacturing, and improve its AI platform Helix with better computing and data collection.
OpenAI Ramps Up Robotics Work in Race Toward AGI
It appears that OpenAI is stepping up its robotics work as part of its mission to achieve AGI. The company has been hiring experts in humanoid robots, teleoperation, and simulation, and maybe preparing to build or mass-produce its own robots, though its exact plans are unclear. This move revives robotics research that OpenAI paused in 2021 and comes as interest and investment in humanoid robots surge across the tech industry.
Uber Gets Back Into Drone Delivery With Flytrex Investment
Uber will test food deliveries by drone with Flytrex later this year in selected Uber Eats markets, while also investing in the company, though the amount remains undisclosed. Flytrex is one of the few firms approved to fly drones beyond visual range and already works with Walmart and DoorDash. The move marks Uber’s return to delivery experiments, after ending its own drone project in 2019.
Tesla loses another Optimus robot leader
Another high-profile figure has left Tesla’s Optimus humanoid robot project. Ashish Kumar, head of the AI team, is the second senior member to depart following program chief Milan Kovac. Kumar, who has now joined Meta’s AI division, said Tesla had offered him better financial prospects, but he chose to pursue other opportunities.
Weave Robotics robot folds laundry
In this short video, Weave Robotics demonstrates how its robot folds laundry—or, more precisely, how it handles a shirt, since the clip cuts off before the robot can complete its work. The company also stated that the robot is deployed at Tumble Laundry in San Francisco. However, I couldn’t find any further details about this partnership, such as whether it is only a pilot programme or a full-scale deployment.
U.K.-based startup Humanoid unveils HMND 01 Alpha mobile manipulator
London-based Humanoid has launched the HMND 01 Alpha, a dual-armed mobile robot built in just seven months to help tackle labour shortages in industries like manufacturing, logistics, and retail. Backed by $50 million in funding, Humanoid plans to offer the system through a robotics-as-a-service model and sees a huge growth market ahead, with a more advanced Beta version expected in 2026.
A security robot failed in NYC: Now, it's trying to protect downtown Kansas City
Patmos Tech has brought in Lance, a K5 security robot made by Knightscope, to patrol its Kansas City data centre after repeated break-ins and vandalism. The robot works alongside human guards, recording video and deterring intruders, and has reportedly already assisted police with investigations. The NYPD also trialled Knightscope’s K5 in the Times Square subway from September 2023 to February 2024, where it had to be constantly guarded by officers before being retired.
🧬 Biotechnology
AI-Designed Phages
Scientists at the Arc Institute and Stanford have used an Evo 2 to design and then make in the physical world the first working genomes created by artificial intelligence. They made 16 synthetic viruses based on ΦX174, a small virus that infects E. coli. Some of these AI-designed viruses worked as well or even better than the natural version. It’s a big step toward AI-driven genome design, though making bigger, more complex organisms this way is still a long way off. Both Evo 1 and Evo 2 models are freely available on Hugging Face.
Biohybrid crawlers can be controlled using optogenetic techniques
Researchers have created biohybrid robots that combine mouse muscle and neurons with 3D-printed scaffolds and wireless optoelectronics. The neurons control muscle contractions to produce movement, whose speed can be adjusted through light stimulation. Future work aims to integrate learning, memory, and more complex movements, with potential uses in medicine and studying motor processes.
Thanks for reading. If you enjoyed this post, please click the ❤️ button or share it.
Humanity Redefined sheds light on the bleeding edge of technology and how advancements in AI, robotics, and biotech can usher in abundance, expand humanity's horizons, and redefine what it means to be human.
A big thank you to my paid subscribers, to my Patrons: whmr, Florian, dux, Eric, Preppikoma and Andrew, and to everyone who supports my work on Ko-Fi. Thank you for the support!
My DMs are open to all subscribers. Feel free to drop me a message, share feedback, or just say "hi!"