Sync #539
Plus: Claude Sonnet 4.5; Thinking Machines Lab releases its first product; SB 53 is now law in California; Zoox is expanding to Washington; scientists turned human skin cells into eggs; and more!
Hello and welcome to Sync #539!
It’s been an eventful week in tech, with new releases and announcements happening left and right. One highlight was the release of Sora 2, OpenAI’s upgraded video model, alongside the launch of the TikTok-like Sora app for AI-generated video clips. The latter came as a bit of a surprise, but it fits into OpenAI’s strategy to build moats and move towards profitability. You can read more about it here.
OpenAI's Push for Moats and Profitability
AI breakthroughs don’t pay for themselves—and OpenAI knows it. With losses piling up and rivals close behind, the company has to do more than release ever-better models. It has to launch products on top of those models and convert as many as possible of its 700 million weekly active users into revenue.
Elsewhere in AI, Anthropic released Claude Sonnet 4.5. Meanwhile, Mira Murati’s Thinking Machines Lab launched its first product; former OpenAI and DeepMind employees unveiled a startup aiming to build “AI scientists”; an AI actress has sparked controversy in Hollywood; and SB 53 has officially become law in California.
Over in robotics, Zoox is expanding to Washington, and DoorDash has unveiled Dot, its first commercial autonomous delivery robot. We also have a conversation with Rodney Brooks, a pioneer in robotics, who shares a grounded perspective on humanoid robots, AI hype, and the future of automation.
In addition, this week’s issue of Sync features scientists who created human eggs from skin cells and used them to form embryos, Microsoft’s discovery that AI can generate “zero-day” threats in biology, and Retro Biosciences’ ambition to add ten healthy years to the human lifespan—and much more.
Enjoy!
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
Retro Biosciences wants to add 10 healthy years to your life
In this short interview, Joe Betts-LaCroix discusses his journey from tech innovator to founder of Retro Biosciences, a company aiming to extend healthy human life by 10 years. Backed by $180 million from Sam Altman, Retro is developing treatments such as an autophagy-based drug for Alzheimer’s and a method to rejuvenate the immune system using reprogrammed “age-zero” blood stem cells. Betts-LaCroix believes combining biotech with artificial intelligence can unlock solutions to ageing and hopes to have several effective treatments available within the next decade.
China’s oyster-inspired ‘bone glue’ bonds fractures, can replace metal in surgery
Chinese researchers have developed an oyster-inspired bone adhesive called “Bone-02” that can quickly repair fractures without metal plates or major surgery. The injectable glue bonds bones in just two to three minutes, even in blood-rich areas, and is strong, biodegradable, and reduces infection risks. Early tests suggest it could make fracture treatment faster, less invasive, and more effective, especially for complex injuries.
🧠 Artificial Intelligence
Introducing Claude Sonnet 4.5
Anthropic has released Claude Sonnet 4.5, which it calls the best model yet for coding, building agents, and using computers. According to Anthropic, the new model delivers state-of-the-art coding performance and is capable of building production-ready applications, handling long and complex tasks for over 30 hours. It leads on key benchmarks, such as SWE-Bench Verified, and has demonstrated the ability to autonomously set up databases, purchase domains, and conduct security audits. Claude 4.5 includes improved coding tools, integrated file and code execution in the Claude app, and a new Agent SDK for developers, all at the same API pricing as its predecessor. Anthropic highlights enhanced safety and alignment, with reduced sycophancy and stronger defences against prompt-injection attacks. In addition, the company has launched an experimental Imagine with Claude feature for real-time software generation. For a good independent overview of Claude Sonnet 4.5, I recommend checking out
Mira Murati’s Stealth AI Lab Launches Its First Product
Thinking Machines Lab, a startup led by former OpenAI employees, has launched its first product—Tinker, a tool that automates much of the infrastructure needed to fine-tune frontier AI models such as Llama and Qwen. Tinker aims to democratise access to powerful AI by enabling researchers, companies and even hobbyists to fine-tune models using simple API calls for supervised or reinforcement learning, and then deploy them anywhere. While Tinker is an interesting tool, it is not a groundbreaking AI model that, on its own, justifies the company’s $2 billion in funding and $12 billion valuation. Tinker is now in private beta for researchers and developers, and those interested can apply for access through a waitlist.
Top A.I. Researchers Leave OpenAI, Google and Meta for New Start-Up
Periodic Labs, founded by ex-OpenAI and DeepMind researchers, has raised $300 million from investors like a16z, Nvidia, Jeff Bezos, and Eric Schmidt to build “AI scientists.” Their vision is to create systems paired with autonomous labs that will run real-world experiments to accelerate discoveries in physics, chemistry, and materials science. Its first focus is finding better superconductors and other advanced materials, aiming to generate fresh data and partner with industry to accelerate innovation.
SB 53, the landmark AI transparency bill, is now law in California
California Governor Gavin Newsom has signed Senate Bill 53, the “Transparency in Frontier Artificial Intelligence Act,” into law, introducing new transparency and safety rules for major AI developers. The bill, a revised version of last year’s vetoed SB 1047, requires large AI companies to publish their safety frameworks online, update them within 30 days of any changes, and report serious safety risks to the state. It also protects whistleblowers and includes penalties for companies that don’t comply. The law will be updated yearly to keep up with new technology. The bill has split the AI industry: Anthropic supported it after negotiations, while others, such as Meta and OpenAI, opposed it, warning it could harm innovation and drive companies out of California.
Hollywood celebrities outraged over new ‘AI actor’ Tilly Norwood
An AI actor named Tilly Norwood has sparked controversy after reportedly speaking with talent agencies. Presented online as a friendly, aspiring actress, Norwood has shared comedy sketches and mock appearances, but the actors’ union SAG-AFTRA and several Hollywood stars have criticised it, saying AI cannot replace human experience and threatens actors’ jobs. Norwood’s creator insists it is a creative project, not a substitute for real performers, but the debate has reignited concerns about AI in the film industry.
Apple Builds a ChatGPT-Like App to Help Test the Revamped Siri
Apple has reportedly built an internal iPhone app called Veritas to test new features for a major Siri upgrade, expected as early as March next year. The app lets engineers try new features, such as searching personal data, controlling apps, and holding longer conversations. Powered by new AI technology called Linwood, the revamped Siri aims to be smarter and more useful across Apple devices.
CoreWeave stock closes up 12% after company lands $14 billion deal with Meta
CoreWeave’s shares jumped nearly 12% after it agreed to supply Meta with $14.2 billion worth of AI cloud infrastructure. The deal comes soon after CoreWeave expanded its partnership with OpenAI to $22.4 billion. Meta, which can increase the deal through 2032, is investing heavily in AI and building huge data centres like its planned Hyperion supercluster to boost its computing power.
Cerebras Systems Raises $1.1 Billion Series G at $8.1 Billion Valuation
Cerebras has raised $1.1 billion in a Series G funding round, bringing its post-money valuation to $8.1 billion. The company plans to use the funds to expand its AI hardware and systems capabilities, scale US manufacturing and data centre infrastructure, and meet demand as its inference platform claims to outperform Nvidia GPUs by over 20 times in many models.
Musk’s xAI sues OpenAI, alleging theft of trade secrets
xAI has sued OpenAI in California, accusing it of stealing trade secrets by luring away key employees. The company claims OpenAI targeted staff to gain access to its technology and methods, including the rapid construction of its “Colossus” data centre. xAI argues that OpenAI acted unlawfully because it felt threatened by Grok. OpenAI denies the allegations, calling the lawsuit part of Musk’s ongoing harassment and insisting it has no interest in competitors’ secrets.
Introducing the Perplexity Search API
Perplexity has launched a new Search API that gives developers access to the same search infrastructure used by their public answer engine, complete with structured, fine-grained results, an SDK, open-source evaluation tools, and real-time indexing optimised for AI workloads.
Google reveals its Gemini-powered smart home lineup and AI strategy
Google continues to integrate Gemini into more of its products and services, now bringing it to Google Home and Nest. Many of Google’s 800 million connected devices will gain Gemini’s features, including natural conversations, smarter automation, and improved camera summaries, without the need for new hardware. Additionally, Google is opening Gemini to partners and introducing tools like “Ask Home,” aimed at making smart homes more accessible, intuitive, and helpful.
After Disney Cease-and-Desist Letter, Character.AI Says It Removed Media Company’s Characters From Its AI Chatbot Platform
Disney sent a cease-and-desist letter to Character.AI, accusing the company of using Disney characters in its chatbots without permission and misleading users into thinking the content was official. Disney said some chatbots were inappropriate and could harm its reputation. Character.AI removed the characters and said it wants to work with rights-holders to create official experiences.
DeepSeek releases ‘sparse attention’ model that cuts API costs in half
DeepSeek has launched V3.2-exp, an AI model designed to lower costs for long-context tasks using a new “Sparse Attention” system. It focuses only on the most relevant parts of the input, reducing server demands and potentially halving API costs. The model is available on Hugging Face. Additionally, DeepSeek has published a paper that explains the model and Sparse Attention in more detail.
Meta plans to buy chip startup Rivos to boost semiconductor efforts
Meta announced plans to acquire Rivos, a Santa Clara-based chip startup specialising in RISC-V technology, to boost its in-house AI chip development. Rivos has worked closely with Meta and brings expertise in full-stack AI systems. The deal, with undisclosed terms, will help Meta advance its custom Meta Training and Inference Accelerator (MTIA) chips and reduce reliance on Nvidia.
🤖 Robotics
▶️ Rodney Brooks: The Truth About Humanoid Robots and AI Hype (46:00)
Rodney Brooks, a veteran roboticist and founder of influential companies such as iRobot and Rethink Robotics, shares in this conversation a grounded view on humanoid robots, arguing that current systems excel at staged demos but lack the reliable manipulation, safe locomotion and practical deployment needed for real-world use. He emphasises that AGI isn’t necessary for useful robots—which he believes are centuries away—and that progress will come from pragmatic, collaborative machines that work alongside humans, like warehouse carts and power-assisted tools, rather than from hype-driven humanoid robots.
Zoox is expanding to Washington, D.C.
Amazon-owned Zoox is expanding to Washington, D.C., its eighth US test city, starting with mapping the streets using retrofitted SUVs before moving to autonomous trials with safety drivers later this year. The move follows the launch of a robotaxi service in Las Vegas and ongoing efforts to gain approvals for wider commercial deployment.
DoorDash unveils Dot, its first commercial autonomous delivery robot
After years of development and testing in real-world conditions, DoorDash has launched Dot, its first autonomous delivery robot for local deliveries. About one-tenth the size of a car and reaching speeds up to 20 mph (32.1 km/h), Dot is built to reduce congestion and handle small, quick deliveries. It can carry up to 30 lb (13.6 kg) and fit through most doorways. Dot will first roll out in Tempe and Mesa, Arizona.
▶️ Drone Delivery Is Taking Off (31:56)
In this video, Asianometry visits Manna, a Dublin-based drone delivery startup, to see how drone delivery actually works after a decade of failed promises from tech giants like Amazon and Google. Manna focuses on small, quick-to-serve items like coffee and fast food, using simple but reliable drones that drop packages by string. The drones fly at high speeds, bypass traffic, and deliver hot food in minutes, often faster and cheaper than car-based couriers. With over 250,000 flights completed and new regulations opening up the US market, Manna shows that drone delivery may finally be ready to scale, despite challenges like noise concerns, local rules and growing competition.
Two Amazon delivery drones crash into crane in commercial area of Tolleson
Two Amazon delivery drones crashed into a crane in a commercial area of Tolleson, Arizona. Amazon says it has found no problems with the drones or their technology and will restart deliveries. In response to the accident, Amazon said it will be adding new visual checks to spot moving obstacles, such as cranes.
Why the World Needs a Flying Robot Baby
Meet iRonCub, a jet-powered humanoid robot that can both walk and fly. Developed by Daniele Pucci’s team at the Italian Institute of Technology in Genoa, it achieved its first stable lift-off last summer after nearly a decade of development. Designed for disaster response, iRonCub can fly over obstacles to reach emergencies and then walk to save energy while clearing debris or opening doors. Although technical challenges remain, the project continues to inspire and attract students and researchers with its ambitious vision.
🧬 Biotechnology
Scientists Made Human Eggs From Skin Cells and Used Them to Form Embryos
Scientists have made human eggs from skin cells using a method called “mitomeiosis”. Some eggs were fertilised and grew into early embryos, but all had genetic problems and none were implanted. The technique could one day improve IVF or help people who cannot produce eggs—such as older women, cancer survivors, or same-sex couples—by creating eggs from their own cells. However, it is still inefficient, raises ethical concerns, and is likely many years away from clinical use.
Microsoft says AI can create “zero day” threats in biology
Microsoft researchers discovered a “zero-day” flaw in DNA biosecurity screening by using generative AI to redesign toxic proteins so they could bypass safeguards while remaining dangerous. The study was done entirely digitally, with no physical viruses created, and authorities and software vendors were warned before publication, though the fixes are still incomplete.
World’s first ‘behavior transplant’ between species achieved
Scientists at Nagoya University have, for the first time, transferred a behaviour from one species to another by activating a single gene. They switched on a gene in specific brain cells of Drosophila melanogaster, changing its usual courtship song, produced by wing vibrations, into the gift-giving behaviour seen in D. subobscura. This discovery shows that new behaviours can arise by rewiring existing brain circuits rather than creating new ones, suggesting that some species may carry hidden behaviours that can be switched on through simple genetic changes.
World’s first mushroom-powered waterless toilet appears in botanical garden
MycoToilet is a waterless toilet powered by mushroom mycelia that turns waste into compost without water, electricity or chemicals. Designed for parks and remote areas, it’s modular, wheelchair-accessible and built from timber with a green roof. A six-week pilot will test its performance, and it’s expected to produce 600 litres of soil and 2,000 litres of liquid fertiliser each year, offering a clean, eco-friendly waste solution.
💡Tangents
Apple Shelves Vision Headset Revamp to Prioritize Meta-Like AI Glasses
Mark Gurman reports that Apple has paused its planned overhaul of the Vision Pro headset to focus on developing smart glasses to compete with those offered by Meta. According to his sources, Apple is working on at least two types of smart glasses: one that will pair with an iPhone and lack its own display, and another with a built-in display that could rival Meta’s recently released Ray-Ban Display. The first model is rumoured to be unveiled as early as next year, while the more advanced smart glasses are planned for 2028, although Apple aims to release them sooner.
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!"