Sync #538
Plus: Gemini Robotics 1.5; Meta launches a super PAC to fight AI regulation; Waymo for Business; Oracle eyes a $20B deal with Meta; Stan Lee is back as an AI hologram; and more!
Hello and welcome to Sync #538!
Normally, I include an article that delves deeper into a notable event from the week. However, for this issue, I want to try something new.
The article originally intended for this week’s issue of Sync has been published as a standalone piece, released ahead of this newsletter. You can read it here.
The Dream of Abundant Intelligence
It’s been an eventful September for OpenAI. In the space of a fortnight, OpenAI made two megadeals: a $300 billion, five-year cloud contract with Oracle and a strategic partnership with Nvidia worth up to $100 billion.
Please let me know what you think of this format, either in the comments below or privately.
Now, let’s dive into this week’s news!
This week, Oracle is eyeing a $20 billion deal with Meta, while Meta is launching a super PAC to oppose AI regulation. Elsewhere in AI, Meta has launched an app for sharing AI-generated videos, OpenAI has introduced ChatGPT Pulse, and Stan Lee is set to appear at LA Comic Con as an AI hologram.
Over in robotics, Google DeepMind has released Gemini Robotics 1.5, bringing AI agents into the physical world. Meanwhile, Waymo has launched a new service for businesses, and Unitree has quirky new skills its humanoid robot has learned. Speaking of Unitree, cybersecurity researchers have discovered a vulnerability that could allow attackers to take control of its robots.
Additionally, Huntington’s disease has been successfully treated for the first time using a pioneering gene therapy; China has become the world leader in factory automation; and New Zealand is aiming to eliminate invasive predators by 2050 using a variety of tools, including AI and genetic engineering.
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.
🧠 Artificial Intelligence
Oracle eyes $20 billion AI cloud computing deal with Meta
Reuters reports that Oracle is in talks with Meta over a cloud computing deal worth about $20 billion. The deal would give Meta more power to train and run AI models, adding to its existing cloud providers. It follows Oracle’s large contract with OpenAI, worth $300 billion, and reflects the rising demand for computing capacity as the AI race intensifies.
Elon Musk’s xAI raising $10 billion at $200 billion valuation
xAI is reportedly raising $10 billion from investors, valuing the company at $200 billion, according to sources cited by CNBC. This comes shortly after Musk raised another $10 billion at a lower valuation of about $150 billion. Musk later denied the report, calling it “fake news.”
Executives at xAI Clashed With Musk Advisers Before Departing
According to The Wall Street Journal, several top executives, including co-founder Igor Babuschkin and former CFO Mike Liberatore, have left xAI after disputes over the company's management and concerns about its financial stability. Despite raising billions, xAI is under pressure from rivals like OpenAI and Anthropic and has faced setbacks with its Grok chatbot, while continuing to spend heavily on data centres and chips to grow quickly.
Meta launches super PAC to fight AI regulation
Last month, Meta launched a PAC in California, and now it has founded its first national super PAC, the American Technology Excellence Project, to push back against state-level tech and AI regulations it views as harmful. Backed by tens of millions of dollars, the group will support candidates from both parties who promote US tech leadership, advance AI progress, and give parents more control over children’s online experiences. This is one of several new tech-focused PACs launched recently, and I’ve written more about them here.
Google releases improved Gemini 2.5 Flash and Flash-Lite
Google has released preview versions of Gemini 2.5 Flash and Flash-Lite, which promise lower costs and better performance. According to Google, the new Flash-Lite reduces output tokens by 50% and improves instruction following, multimodal capabilities, and conciseness, while Flash offers enhanced tool-use abilities, better long-horizon reasoning, and a 5% gain on the SWE-Bench benchmark, all with greater efficiency. These preview versions are available in AI Studio and Vertex AI, and are designed to collect feedback and guide future stable releases.
Introducing ChatGPT Pulse
OpenAI is launching ChatGPT Pulse, a new feature that delivers daily personalised updates in a card-based feed, drawing on chat history, preferences, and connected apps like Gmail or Google Calendar. Pulse is designed to provide timely, relevant insights without constant prompting. Currently in preview and available only for Pro mobile users, Pulse aims to evolve from a reactive chatbot into a proactive AI assistant by integrating more apps and offering richer, context-aware actions in the future.
Detecting and reducing scheming in AI models
In a new paper, researchers from OpenAI and Apollo Research studied the problem of “scheming,” where AI systems appear to follow instructions but secretly behave in misleading ways. They tested several current AI models and found evidence of this behaviour. To counter it, they trained models to follow a clear set of anti-scheming rules before acting, which made them far less deceptive in tests. However, it is still unclear how well this approach will work in real-world settings. The researchers warn that scheming could become a greater risk as AI systems take on more complex and important tasks, making ongoing research and monitoring essential.
Apple working on MCP support to enable agentic AI on Mac, iPhone, and iPad
Apple’s new betas for iOS 26.1, iPadOS 26.1, and macOS Tahoe 26.1 hint at Apple adding support for Model Context Protocol (MCP), a standard that lets AI systems connect easily to apps and data. By linking it to App Intents, Apple could let AI assistants such as ChatGPT or Claude take actions directly in iPhone, iPad, and Mac apps. The feature is still early, but it points to Apple preparing for more powerful AI integration across its devices.
Introducing Vibes: A New Way to Discover and Create AI Videos
Meta is launching Vibes, a new feed of AI-generated short videos within the Meta AI app and on meta.ai, where users can generate clips from prompts or existing videos, remix others’ creations, and share them across Vibes, Instagram, and Facebook. Meta describes Vibes as being designed to inspire creativity and will become more personalised over time, with plans for deeper integration into Meta’s ecosystem.
It's been an extremely busy day for team Qwen
The team behind the Qwen family of open AI models has released a number of updates to their models, which this post neatly aggregates in one place. The updates include smaller FP8 versions of their Qwen3-Next models, a new tool for multilingual speech synthesis, an advanced multimodal model (Qwen3-Omni) that handles text, audio, and video, and an improved image-editing model that now supports multiple images.
Movie Studio Lionsgate is Struggling to Make AI-Generated Films With Runway
A year ago, Runway, an AI company specialising in video generation, partnered with Lionsgate to support movie production and potentially create full-length AI-generated films. However, as The Wrap reports, the partnership has struggled because the studio’s film library is too small to train effective models. The project also faces legal challenges over using actors’ likenesses without clear rules or payment. Despite the hype, AI video tools still fall short of professional quality, and even companies with much larger datasets, such as Disney, face similar limitations.
An AI Stan Lee Hologram Will Speak With Fans at L.A. Comic Con
Almost seven years after his death, Stan Lee will appear at L.A. Comic Con “resurrected” as an AI-powered hologram. Fans can visit the “Stan Lee Experience” booth to chat with him, take photos, or book short one-on-one conversations. Using decades of recordings, the hologram aims to stay true to Lee’s voice and personality, and may even show up as a surprise at panels during the convention, reports The Hollywood Reporter.
🤖 Robotics
Gemini Robotics 1.5 brings AI agents into the physical world
Google DeepMind has released Gemini Robotics 1.5 and Gemini Robotics-ER 1.5 models to help robots better understand their surroundings and handle complex tasks. Gemini Robotics-ER 1.5 serves as a high-level planner with state-of-the-art spatial reasoning and tool-use abilities, while Gemini Robotics 1.5 translates instructions into actions, thinks before acting and transfers skills across different robot types. Together, DeepMind says, they make robots more adaptable, transparent and safe, marking an important step towards more capable, general-purpose robots in the real world, and progress towards AGI in the physical world.
Waymo for Business
Waymo has introduced Waymo for Business, a new service that lets companies, universities, and event organisers offer people rides in its self-driving, all-electric cars. The service is in the early stages, and Waymo is inviting organisations to take part in the rollout.
There Are More Robots Working in China Than the Rest of the World Combined
China has become the world leader in factory automation, with over two million industrial robots in use and nearly 300,000 installed in 2024—more than the rest of the world combined. That year, almost 60% of the robots installed in China were domestically made, and the country’s share of global robot manufacturing rose to one-third. This growth has strengthened China’s dominance in global manufacturing, with the use of robots and AI improving efficiency. However, China still relies on other countries for some advanced components, such as sensors and semiconductors.
Exploit Allows for Takeover of Fleets of Unitree Robots
Cybersecurity researchers have discovered a critical vulnerability in the Bluetooth Low Energy (BLE) Wi-Fi setup of several Unitree robots, including the Go2 and B2 robot dogs and the G1 and H1 humanoid robots, allowing attackers to take full control and even spread the attack to other nearby robots. Despite being alerted to the problem in May 2025, Unitree has not yet fixed it. Experts warn that this could put users and the public at risk, especially as these robots are already in use, and urge manufacturers to take robot security more seriously.
▶️ Unitree G1 has mastered more quirky skills (1:33)
In this video, Unitree shows new skills its G1 humanoid robot has learned, like quickly getting up or being able to do multiple backflips one after another. What they showed is impressive and getting very close to a human-level performance.
Zoox asks federal regulators for exemption to launch a commercial robotaxi service
Zoox, the Amazon-owned self-driving car company, has asked US regulators for permission to launch its robotaxi service using vehicles without steering wheels or pedals. The National Highway Traffic Safety Administration (NHTSA) is reviewing the request, which would exempt Zoox from certain safety rules. If approved, the exemption would pave the way for Zoox to expand beyond testing and formally launch its commercial autonomous taxi service.
Rethink Robotics shuts down — again
Rethink Robotics, once a pioneer in collaborative robots with its Baxter and Sawyer models, has shut down for a second time. Founded in 2008 and acquired by Hahn Automation Group after its 2018 bankruptcy, the company was later relaunched in the US by United Robotics Group in 2024. However, delays in product readiness, poor sales, and withdrawn investor support led to its closure in August 2025.
▶️ HMND SERIES E01 | How to Build a Humanoid in 7 Months (22:45)
Last week, Humanoid, a London-based humanoid robotics company, released its first robot, HMND 01 Alpha. This video documents the culmination of seven months of work—bringing all the pieces together to assemble their first robot, along with all the challenges that came with it.
🧬 Biotechnology
Huntington's disease successfully treated for first time
Doctors say Huntington’s disease has been successfully treated for the first time using a pioneering gene therapy. In a small trial, the treatment slowed disease progression by 75% over three years, giving patients the chance to enjoy decades of life. Delivered in a single 12–18 hour brain surgery, the therapy uses a modified virus to silence the faulty gene that produces a toxic protein. While the approach appears safe and could last for life, it will be expensive and is not yet licensed, with approval in the US targeted for 2026.
How do you use a virtual cell to do something actually useful?
Virtual cells, simulators of real biological cells, promise to speed up research by reducing reliance on costly lab experiments. In this article, Abhishaike Mahajan from
To save its unique and rare birds, New Zealand is turning to AI and genetic research
New Zealand aims to eliminate invasive predators by 2050—the largest project of its kind in the world. To reach this goal, the country is using a wide range of tools, from traditional traps and predator-proof sanctuaries to advanced technologies like AI and genetic research, including gene drives. Although costly and complex, the effort has strong backing from the public and Māori communities and could serve as a global model for conservation.
Organoids Recreate Native Kidney Architecture to Model Complex Disease in Mice
Scientists have created advanced lab-grown kidney structures called “assembloids” that combine different cell types to act more like real kidneys. When transplanted into mice, these assembloids developed blood vessels and connective tissue, and showed functions such as filtering blood and producing early urine. This research is a major step toward the long-term goal of building functional synthetic kidneys for patients who need transplants.
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!"