Google escapes breakup in antitrust ruling - Sync #535
Plus: Anthropic raises $13B at $183B valuation; DeepSeek preps AI agents; Tesla’s next-gen Optimus revealed; where are all trillion dollar biotech companies; genetically engineered horses; and more!
Hello and welcome to Sync #535!
This week, Google avoided being broken up, but a federal judge imposed new restrictions designed to curb its monopoly—which we will take a closer look at in this week’s write-up.
Elsewhere in AI, Anthropic has raised a $13 billion Series F round, valuing the company at $183 billion. Some of that money will probably go to authors, as Anthropic agreed to pay $1.5 billion to settle a lawsuit from writers who claimed the company trained its AI on pirated books without permission. Mistral also closed a new funding round, bringing in €2 billion (about $2.35 billion), giving it a €12 billion ($14 billion) valuation. Meanwhile, OpenAI announced an AI-powered rival to LinkedIn job boards, DeepSeek is preparing to release AI agents, and Anthropic and OpenAI tested each other’s models for safety.
Over in robotics, we’ve got a glimpse of the next-gen Tesla Optimus, while Figure showed how its humanoid robot can load a dishwasher. We also have a humanoid robot playing table tennis and Agility Robotics explaining how they built a motor cortex for their robots.
Elsewhere, this week’s issue of Sync also covers Neuralink expanding to Canada with successful brain implants in spinal injury patients, an article asking why there are no trillion-dollar biotech companies, genetically engineered Argentinian horses, and more!
Enjoy!
Google escapes breakup in antitrust ruling
Google was spared from being broken up but faces new limits on its search monopoly
Google has avoided the worst outcome in the biggest US antitrust trial of the digital age, but the company must change how it operates its search business after a federal judge imposed new restrictions designed to curb its monopoly power.
In a decision last week, US District Judge Amit Mehta rejected calls from the Department of Justice to force Google to sell Chrome or Android. Instead, he ordered some changes intended to open the online search market to greater competition.
No breakup, but…
Judge Mehta ruled last year that Google had acted illegally to maintain its near-90% share of the search market, relying on default placements to lock out rivals. Yet he stopped short of imposing structural remedies, calling divestiture “a poor fit” and warning it could disrupt products billions rely on.
Instead, the judge barred Google from entering into or maintaining exclusive distribution deals for Search, Chrome, Google Assistant, and its Gemini AI app. Google may still pay for default placement, but manufacturers are free to preload or promote alternatives alongside Google’s products or services.
Google must also provide “qualified competitors” with access to parts of its search index and user-interaction data. The aim is to help rivals like DuckDuckGo or Bing compete with Google’s quality of results. The DOJ had sought broader access, but Mehta limited the scope.
To enforce compliance, a technical committee will oversee the remedies, which last six years and take effect 60 days after final judgment.
Mixed response
The DOJ welcomed the ruling as progress but said it was weighing “whether the ordered relief goes far enough.” DuckDuckGo’s founder Gabriel Weinberg called it inadequate, warning that consumers would “continue to suffer.”
Google, by contrast, framed the outcome as vindication. It highlighted the role of AI in reshaping search and argued that competition is already “intense.” Still, it raised concerns about privacy risks from mandated data-sharing.
Investors sided with Google: shares of its parent company Alphabet jumped more than 7% after the ruling, while Apple’s stock also rose. Apple, which earns more than $20 billion annually from Google to keep its search engine as the iPhone’s default, benefits from the continuation of those payments, now without exclusivity.
How AI influenced the ruling
Artificial intelligence played an unusually prominent role in the ruling. Judge Mehta noted that ChatGPT, Perplexity, and Google’s own Gemini are transforming how people find information online.
Mehta described the court as having to “gaze into a crystal ball and look to the future,” acknowledging that the rapid pace of AI complicates traditional antitrust remedies. Forcing a breakup, he suggested, might be less effective when the search landscape itself is being reshaped by new technologies.
This view shaped his decision to avoid structural remedies and instead focus on limiting exclusivity and requiring data-sharing. By doing so, he signalled that the market may evolve away from Google’s dominance organically, provided rivals have a fairer shot at building competitive products.
The ruling could pave the way for deeper collaboration between Google and Apple, as it is rumoured that Gemini will be integrated into Siri. At the same time, AI start-ups like Perplexity, which even floated a bid to buy Chrome for $34.5 billion in cash (almost twice its current valuation), may benefit from access to Google’s data to helps them scale.
What’s next
The legal battle is far from over. Google has vowed to appeal the finding of liability, a process that could stretch into 2027 or beyond. At the same time, it faces another major trial over its digital advertising business, where a federal judge has already ruled it holds illegal monopolies. Remedies in that case will be debated later this month.
Google may have escaped being broken apart, but with AI-powered answer engines transforming how people find information, the company’s era of search supremacy could be drawing to a close.
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
Neuralink expands to Canada with successful brain implants in spinal injury patients
Neuralink has completed its first Canadian brain-computer implant surgeries at Toronto’s University Health Network, after recently starting trials in the UK. Two patients with spinal cord injuries received the device through robotic surgery as part of the CAN-PRIME Study, which is testing whether people with paralysis can safely use their thoughts to control things like a computer cursor or robotic arm.
Pig Lung Transplanted Into a Human in Major Scientific First
Doctors in China have transplanted a genetically modified pig lung into a brain-dead patient, where it worked for nine days before rejection set in and the experiment was ended. The lung came from a pig with six gene edits to reduce immune response, and although it avoided immediate rejection, complications like swelling and tissue damage soon developed. While not a permanent solution, the study offers important insights into the difficulties of xenotransplantation.
🧠 Artificial Intelligence
Anthropic raises $13B Series F at $183B post-money valuation
Anthropic has raised $13 billion in a Series F round led by ICONIQ, valuing the company at $183 billion. The funding comes as demand for its AI platform, Claude, has surged—helping revenue grow from $1 billion to over $5 billion in just eight months. With more than 300,000 business customers and the growing popularity of its tools, such as Claude Code, Anthropic is one of the fastest-growing tech companies ever.
Mistral Set for $14 Billion Valuation With New Funding Round
French startup Mistral AI is close to securing €2 billion (about $2.35 billion) in new funding, giving it a €12 billion ($14 billion) valuation and making it one of Europe’s most valuable tech firms. Founded in 2023 by former DeepMind and Meta researchers, Mistral builds open-source language models, a European-focused chatbot called Le Chat, and AI cloud services with Nvidia.
Anthropic Agrees to Pay Authors at Least $1.5 Billion in AI Copyright Settlement
Anthropic has agreed to pay $1.5 billion to settle a lawsuit from authors who said the company trained its AI on pirated books without permission. This is the first major US settlement over AI and copyright, and it could shape future rules for how creative work is used in AI training. The payout is far less than what a trial might have cost, but it sets a clear precedent that companies must compensate copyright owners. Anthropic denies any wrongdoing, and the company still faces other lawsuits, including one from major music labels.
China’s DeepSeek Preps AI Agent for End-2025 to Rival OpenAI
Bloomberg reports that Chinese startup DeepSeek is building a new AI model with advanced agent features to compete with US rivals like OpenAI, aiming to launch it by the end of the year. Led by founder Liang Wenfeng, the new model is designed to handle multi-step tasks on its own and improve through experience. The move follows DeepSeek’s earlier success with its R1 model and comes as global tech firms race to develop AI agents that can carry out more complex tasks with less human guidance.
Head of UK’s beleaguered Alan Turing Institute resigns
Jean Innes is stepping down as chief executive of the Alan Turing Institute after staff unrest and government calls for change. Technology secretary Peter Kyle wrote to the institute in July, saying it should shift its focus to defence and national security and ensure it had the right leadership for that move. Innes, who became chief executive in 2023, had overseen a major restructuring but faced growing discontent among staff, job cuts, and fears over the loss of £100m in government funding. While the board has agreed to follow the government’s direction, many employees worry the institute is moving away from its wider mission to tackle societal challenges.
Mistral adds custom MCP connectors and memory to Le Chat
Mistral has introduced two major updates to its AI assistant Le Chat: custom MCP connectors, allowing users to integrate with over 20 enterprise platforms or build their own for tailored workflows, and memories, enabling the assistant to retain useful details such as preferences and facts for more personalised responses. Both features are available even on the free tier, with full user control to add, edit, or delete memories, plus options to import data from ChatGPT.
OpenAI announces AI-powered hiring platform to take on LinkedIn
OpenAI plans to launch the OpenAI Jobs Platform in 2026, an AI-powered hiring service that will connect businesses with workers, competing directly with LinkedIn. The company will also offer AI fluency certifications through its OpenAI Academy, starting a pilot in 2025 with Walmart and aiming to certify 10 million Americans by 2030. While OpenAI admits AI could replace many jobs, it says these programmes will help people learn AI skills and find new opportunities.
Apple’s rumored AI search tool for Siri could rely on Google
Apple is creating a new AI-powered search feature for Siri, called “World Knowledge Answers,” which can give users summaries with text, images, and videos, Bloomberg reports. To build it, Apple may use Google’s Gemini model, though it is also testing other AI systems. The updated Siri will be able to search the web, use personal data, and organise results into clear summaries, with the launch expected as soon as March next year.
OpenAI to launch its first AI chip in 2026 with Broadcom
OpenAI is reportedly working with US chipmaker Broadcom to produce its first custom AI chip next year, which it will use for its own systems rather than selling to others. The move is aimed at reducing reliance on Nvidia and cutting costs, following similar strategies by Google, Amazon and Meta as demand for AI computing grows. Broadcom’s CEO said the company expects strong AI revenue growth after securing major new orders, with OpenAI thought to be one of the new clients.
Alibaba Creates AI Chip to Help China Fill Nvidia Void
A report from The Wall Street Journal says Chinese tech companies such as Alibaba, MetaX, Cambricon and Huawei are racing to develop their own chips after US restrictions limited access to advanced semiconductors. Alibaba has created a new AI chip made in China and compatible with Nvidia’s system, while MetaX and Cambricon are offering alternatives to restricted Nvidia products. Although China is making progress, it still struggles to match US technology for training AI, though heavy investment and innovation may help close the gap.
Amazon launches Lens Live, an AI-powered shopping tool for use in the real world
Amazon has launched Lens Live, a new AI feature in its shopping app that lets users point their phone camera at real-world objects and instantly see similar products on Amazon. The tool also connects with Amazon’s AI assistant, Rufus, which provides quick product details and answers questions to help with buying decisions.
OpenAI co-founder calls for AI labs to safety-test rival models
OpenAI and Anthropic briefly worked together to test each other’s models for safety issues. The study showed that Anthropic’s models were more cautious, often refusing to answer, while OpenAI’s gave more answers but made more mistakes. Both companies raised concerns about “sycophancy,” where AI can encourage harmful behaviour. Despite competition and some tensions, both sides said they hope to keep collaborating on safety as AI becomes more widely used.
Atlassian to buy Arc developer The Browser Company for $610M
Atlassian, best known for its project management software Jira, has announced plans to acquire The Browser Company—the maker of Arc and Dia browsers—for $610 million in cash. Atlassian CEO Mike Cannon-Brookes said the deal aims to create an AI-powered browser tailored for knowledge work and SaaS-heavy workflows, while The Browser Company’s CEO Josh Miller confirmed the startup will remain independent and continue developing Dia, an AI-first browser.
In a first, Google has released data on how much energy an AI prompt uses
Google has published a report showing how much energy, water, and carbon its Gemini AI uses per text query. A typical prompt consumes 0.24 watt-hours of electricity (like running a microwave for one second), 0.26 millilitres of water (about five drops), and produces 0.03 grams of CO₂. The report also explains where the energy goes, with chips, CPUs, backup systems, and data centres all playing a part, and notes that efficiency has improved greatly—using 33 times less energy in 2025 than in 2024. While this is the clearest picture yet from a major AI company, experts say more data and standard measures are needed to compare AI systems fairly.
Chinese ‘Virtual Human’ Salespeople Are Outperforming Their Real Human Counterparts
The article explores how AI-powered “virtual human” livestreamers are changing online shopping in China. Companies like Brother use these avatars to sell products nonstop on platforms such as Taobao, often boosting sales and cutting costs compared to human hosts. It also discusses the risks and glitches of the technology, how it might affect real livestreamers and influencers, and the possibility of it spreading to global markets.
🤖 Robotics
▶️ Scaling Helix - Dishes (1:07)
Figure drops another video of its humanoid robot in action, this time showing how it can take dishes and place them into a dishwasher.
Tesla’s next-gen Optimus revealed
Marc Benioff, the CEO of Salesforce, shows in this video the latest version of Tesla’s Optimus humanoid robot. This version has a golden body and is equipped with AI to understand commands. However, it takes Optimus some time to process instructions and respond, and the overall demo feels rough.
UBTECH secures $1B in financing for Middle East humanoid production
UBTECH Robotics has secured up to $1 billion in funding from Infini Capital to boost production and marketing, with plans to build a superfactory, research centre, and regional base in the Middle East. Founded in 2012 in Shenzhen, the company makes humanoid robots for industries like manufacturing, logistics, and healthcare, working with partners such as BYD, Geely, and Foxconn.
Agility Robotics: Training a Whole-Body Control Foundation Model
In this blog post, Agility Robotics shares how the company developed a “motor cortex” for its humanoid robot, Digit, to help it balance, walk, and handle objects safely in human environments. The system is a small LSTM neural network with under one million parameters, trained fully in simulation and able to work in the real world without extra tuning. It allows Digit to lift heavy objects, recover from disturbances, and perform tasks more flexibly by using task-space prompting and position-based training, representing a step toward safe and versatile humanoid robots.
Are we truly on the verge of the humanoid robot revolution?
UC Berkeley’s Ken Goldberg says humanoid robots are not close to matching the progress of AI chatbots. Unlike chatbots, which learn from massive amounts of online text, robots lack enough data to master basic tasks like picking up objects. Experts disagree on whether progress will come from more data or traditional engineering, but current advances mostly come from robots learning on the job, like self-driving cars or warehouse machines. Goldberg believes blue-collar jobs are safe for now, while routine office work is more at risk, and that widespread humanoid robots are still many years away.
▶️ HITTER: A HumanoId Table TEnnis Robot via Hierarchical Planning and Learning (2:40)
Meet HITTER, a research project aimed at training a humanoid robot (in this case, the Unitree G1) to play table tennis. Anyone who has ever played the game knows it is not an easy task, as it requires excellent perception and coordination. The researchers succeeded in their endeavour, training the robot to play effectively against both humans and other humanoid robots.
🧬 Biotechnology
Where are all the trillion dollar biotechs?
The article argues that biotech faces a fundamental economic crisis: despite massive spending, drug R&D has become less productive, and only a few medicines have ever reached huge blockbuster sales. Strategies like using human genetics, repurposing old drugs, and applying AI have offered hope but have mostly failed to deliver reliable success, especially in treating age-related diseases such as Alzheimer’s and heart disease, which are the biggest markets. The path to reversing biotech’s decline lies not in incremental efficiency but in tackling age-related diseases, with new therapeutic modalities, adaptive regulations, and tools capable of unlocking fundamentally new biology.
World’s first gene-edited horses are shaking up the genteel sport of polo
Kheiron Biotech in Argentina has produced the world’s first genetically edited horses, designed for greater speed using CRISPR. While cloning is accepted in polo, the sport’s authorities and breeders have blocked these foals, fearing the loss of tradition and risks to the industry. Supporters believe the technology could transform breeding, but critics see it as going too far. For now, the gene-edited horses cannot compete, and their future in polo remains uncertain.
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!"
Great roundup of stories. Especially love the videos. Amazing stuff.