Siri’s Last Chance - Sync #532
Plus: Ozempic shows anti-ageing effects; Altman backs a Neuralink competitor; whispers of DeepSeek R2; Figure’s robot folds laundry; fight club for humanoid robots; Intel falls apart; and more!
Hello and welcome to Sync #532!
In this week’s issue of Sync, we take a closer look at how Apple plans to respond to the disastrous release of Apple Intelligence and its last chance to remain relevant in the AI revolution.
Elsewhere in AI, Nvidia unveils new world models, OpenAI wins the International Olympiad in Informatics, and we examine how AI has conquered the US economy. Meanwhile, whispers of DeepSeek R2 start to emerge, Anthropic releases a small update to Claude, and xAI’s co-founder is leaving the company.
Over in robotics, Figure shows how its robot folds laundry, Tesla secures a permit to run a ride-hailing service in Texas, and we explore the futuristic fight clubs where robots battle each other.
Additionally, Ozempic shows signs of anti-ageing effects, Sam Altman backs a competitor to Neuralink, Perplexity offers to buy Google Chrome, and Gamers Nexus documents Intel’s downfall.
Enjoy!
Siri’s Last Chance
The next year will decide Apple’s AI future. Can it fix Siri, build its own answer engine, and finally deliver on years of promises—or will it be left behind for good?
Apple’s venture into AI is off to a poor start. At WWDC 2024, the company finally unveiled Apple Intelligence—its big move into the generative AI race. The vision was compelling: Siri would get the long-overdue makeover that Apple fans had asked for years, AI features would be integrated across apps, and it would all be done with Apple’s focus on privacy.
That vision, however, did not materialise. Delays meant Apple Intelligence arrived late and in fragments, instead of a single cohesive update. The flagship features that wowed audiences in the WWDC demo were missing or watered down. What was shipped felt underpowered next to ChatGPT, Google’s Gemini, and Anthropic’s Claude. Instead of being the moment Apple proved it could lead in AI, Apple Intelligence risks becoming Apple’s biggest failure under Tim Cook.
Apple’s AI stumble is more than just a rough launch. With Apple Intelligence falling flat and rivals racing ahead, the company now needs to fix Siri, build a competitive AI search, and prove it can still lead in the next big shift in computing—or risk being left behind.
Apple Is Not Giving Up on AI
Despite the disastrous launch of Apple Intelligence, Apple is not giving up on AI. Following Apple’s latest earnings report—which beat Wall Street expectations—Tim Cook held a rare, company-wide all-hands meeting at Apple Park. According to Mark Gurman, Cook told employees that the AI revolution is “as big or bigger” than the internet, smartphones, cloud computing, and apps. “Apple must do this. Apple will do this,” he reportedly said.
Cook reminded in his pep talk that Apple has been late before, but it always caught up—and AI will be no different. Apple, he promised, will make the investments necessary to win and deliver an “amazing” pipeline of products.
Cook also urged teams to weave AI into their work more quickly, deploy new tools faster, and push managers to move with speed. With rivals already racing ahead, this was less a pep talk and more an admission that Apple needs to close the gap—and soon.
Apple Turns to AI Search
Part of Apple’s problems with AI is that Apple misread the market. From the start, Apple chose not to build its own chatbot, assuming consumer interest was limited. Instead, it made a deal with OpenAI to bake a minimal version of ChatGPT into Siri.
It turns out that assumption was wrong. ChatGPT’s success—alongside competitors like Gemini and Perplexity—proved people not only want chatbots, they’re using them daily to search, learn, and create. Even Apple’s own services chief, Eddy Cue, has admitted in court during Google’s antitrust trial that AI-powered search is the future.
Now Apple is scrambling to catch up. Earlier this year, it quietly formed the “Answers, Knowledge and Information” (AKI) team, led by former Siri head Robby Walker, to build an answer engine—essentially a stripped-down version of ChatGPT with strong search capabilities. According to reports, the project is still in the early stages. The team is working on infrastructure to power search features in future versions of Siri, Spotlight and Safari. Apple may also release a standalone Answers app.
Siri Reboot
Siri has been around for nearly 14 years. At launch, it felt like science fiction made real—an intelligent assistant in your pocket. But Apple never delivered on that promise. Instead, Siri stagnated. First, Alexa and Google Assistant exposed its shortcomings, then ChatGPT made them impossible to ignore.
Apple now has one final opportunity to turn Siri into the promised voice-based interface to the iPhone. This involves a complete rebuild under two parallel projects—one homegrown and one relying on outside models.
Linwood—internally known as LLM Siri and built by Apple’s Foundation Model team—is Apple’s large language model. It’s designed to understand natural language, handle complex queries, and maintain conversational context. Apple also wants Linwood to draw on a user’s data when answering questions or offering suggestions.
But Apple isn’t betting everything on Linwood. Glenwood is a competing project whose goal is to enable Siri to use third-party models. The final decision on which model to use has not yet been made, but according to reports, the Glenwood team has been testing Anthropic’s Claude while other teams have been using ChatGPT or Google Gemini to build and test features.
But conversational AI alone won’t save Siri. To make Siri truly useful, Apple is also working on an upgrade to App Intents, a framework that will let the assistant take direct control of apps using a voice-first interface. Imagine telling Siri to find a specific photo, edit it, and post it on Instagram, without ever touching the screen. That’s what Apple is working towards. The concept is similar to what other companies are promising with agentic AI, which can execute high-level tasks by navigating apps like a human would.
It’s an ambitious vision—and essential for Apple’s next wave of products, from a smart home hub to a tabletop robot. However, this too is not going smoothly. Reports suggest that engineers can’t yet make it work with a sufficient number of apps and reliably enough for high-stakes scenarios, like banking and healthcare apps, where mistakes are unacceptable.
Apple is reportedly planning to release a smarter and redesigned Siri in the spring of next year.
Apple is Racing Against Time
AI is moving fast. OpenAI and Anthropic ship major updates every few months, while Apple hasn’t shipped anything at all.
Apple can still catch up. Apple has the money, resources, and leadership commitment to go all in on AI. What Apple doesn’t have is time or the luxury of making mistakes. The next year will decide whether it joins the leaders or ends up a customer of theirs.
xAI showed that it is possible to build a world-class AI company with its own massive AI data centres from scratch in just two years. However, that requires billions of dollars of investment, top talent, and a leader with a clear vision and a sense of urgency matching that of Elon Musk.
The question is whether Apple can do the same. The signs point both ways.
On one hand, Apple has announced $500 billion in US investments over the next four years, with AI as a central focus. It also reshuffled executives earlier this year to bring Apple Intelligence back on track after months of delays. On the other hand, Meta is aggressively poaching top AI researchers, and Apple is not immune to raids on its talent pool. Several senior researchers from its Foundation Model team have already left for Meta, with more said to be considering the jump.
Turning Siri into what it was promised to be won’t be easy. If Apple gets it right, Siri will finally deliver on its original promise and secure Apple a place in the AI race. If it doesn’t, Apple risks being locked out of the most important revolution in computing since the iPhone—and this time, there will be no catching up.
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
Ozempic Shows Anti-Aging Effects in First Clinical Trial, Reversing Biological Age by 3.1 Years
A new study has found that Ozempic, a popular weight loss drug, may reverse ageing by over three years on average. In a 32-week trial, people taking weekly semaglutide injections showed improvements in brain health and the inflammatory system, where ageing was delayed by nearly five years, with extra benefits for the heart and kidneys. Scientists think this is due to better fat distribution and less inflammation, but warn it’s too early to use the drug widely for anti-ageing.
Sam Altman, OpenAI will reportedly back a startup that takes on Musk’s Neuralink
The Financial Times reports that Sam Altman is co-founding a new brain–computer interface company called Merge Labs, which may be valued at around $850 million and could receive funding from OpenAI’s ventures arm, though talks are still early. Merge Labs will compete with Elon Musk’s Neuralink, which is developing brain implants to help people with paralysis control devices. The project reflects Altman’s long-standing interest in merging humans with technology and adds a new chapter to his rivalry with Musk.
Can brain stimulation end addiction?
Here is an interview with neuroscientist Vaughn R. Steele, who is leading a summer study testing transcranial magnetic stimulation (TMS) as a treatment for alcohol and opioid addiction. The study compares different TMS protocols and hopes to secure funding for larger trials that could lead to FDA approval. Steele acknowledges results may or may not support TMS effectiveness—both outcomes inform future research.
Do neurotechnologies threaten our mental privacy?
The article looks at the rise of consumer neurotechnology—devices that can read or influence brain activity—and the challenges it brings. These advances could transform treatment for brain and mental health problems, but also risk misuse, such as exploiting personal brain data, manipulating thoughts, and increasing inequality. Experts call for “neurorights” and fair rules to protect people while still allowing innovation.
Apple’s new brain-controlled iPhone, iPad tech revealed in video
Earlier this year, Apple announced plans to support brain–computer interfaces (BCIs) on its devices. In this video, we can see this technology in action, with Mark Jackson, an ALS patient, using a Synchron BCI implanted in his brain’s motor cortex to control an iPad using only his thoughts.
🧠 Artificial Intelligence
Nvidia unveils new Cosmos world models, infra for robotics and physical uses
Nvidia unveiled at SIGGRAPH new models for robotics—Cosmos Reason and Cosmos Transfer-2. Cosmos Reason is a 7-billion-parameter vision-language model that helps robots “reason” using memory and physics understanding for tasks such as planning and data curation. Cosmos Transfer-2 enables faster synthetic data generation, complemented by an optimised distilled version, and is supported by new neural reconstruction libraries for realistic 3D simulation. Nvidia also announced updates to the Omniverse SDK, the RTX Pro Blackwell Server, and the cloud-based DGX Cloud platform.
OpenAI achieves gold medal at the International Olympiad in Informatics
After achieving a gold medal at the International Mathematical Olympiad three weeks ago, OpenAI set its sights on the International Olympiad in Informatics (IOI) and secured another gold medal, placing sixth out of 300 participants. The OpenAI team emphasised that its AI had the same five-hour time limit and 50-submission limit as human participants and, like the human contestants, competed without internet access or RAG, with only a basic terminal tool available. The X thread does not specify which model was used, stating only that they “competed with an ensemble of general-purpose reasoning models — we did not train any model specifically for the IOI.”
How AI Conquered the US Economy: A Visual FAQ
In this article,
Elon Musk threatens Apple with lawsuit over OpenAI, sparking Sam Altman feud
Elon Musk has accused Apple of favouring OpenAI in its App Store rankings and threatening legal action for his AI startup xAI. The spat on X marked another chapter in the ongoing feud between Musk and Sam Altman, with Musk alleging antitrust violations and Altman countering with claims that Musk manipulates X’s algorithms.
Alexa Got an A.I. Brain Transplant. How Smart Is It Now?
Amazon’s new Alexa+, announced in February 2025 and now rolling out, is a generative AI–powered upgrade to its voice assistant, aiming to combine conversational abilities with Alexa’s traditional smart-home functions. This article reviews Alexa+ and assesses whether it delivers on its promises, noting improvements such as more natural dialogue, multi-step task handling, and new features like story generation and restaurant booking, while also highlighting its bugs, slower responses, factual errors, and missing promised functions.
Staff at UK’s top AI institute complain to watchdog about its internal culture
Staff at the Alan Turing Institute (ATI) have filed a whistleblowing complaint alleging governance failures, poor leadership, and a toxic culture, warning of collapse if government funding is withdrawn. Under pressure from the UK technology secretary Peter Kyle to change leadership and refocus on defence and national security, ATI faces restructuring with 10% of staff at risk and key projects on online safety and AI ethics being cut.
Elon Musk’s xAI Releases Grok 4 For Free Globally, Challenges OpenAI’s GPT-5 Launch
xAI has made its new AI model, Grok 4, free for everyone worldwide to compete with OpenAI’s GPT-5, though the more powerful Grok 4 Heavy is still for paying subscribers. Grok 4 offers two modes: Auto Mode for quick answers and Expert Mode for more detailed ones. xAI also launched Grok Imagine, a free AI video tool in the US, which has faced backlash for being used to make explicit deepfakes. Musk says ads will soon appear in Grok to cover high running costs, as the company pushes for more users while dealing with challenges around money-making and responsible AI use.
Claude Sonnet 4 now supports 1M tokens of context
Anthropic has released an updated version of Claude Sonnet 4, which now supports up to 1 million tokens of context on the Anthropic API—a significant increase from the previous 200k token limit. The company has also updated its pricing for prompts exceeding 200k tokens, which will now cost $6 per million tokens for input and $22.50 per million tokens for output.
Huawei AI chip-powered DeepSeek R2 tipped to launch this month
Leaks suggest the DeepSeek R2 will launch globally between 15 and 30 August, though the exact date is uncertain. Designed to compete with GPT-5, it promises improved reasoning, answers, and efficiency, while remaining cost-effective, open-source, and making extensive use of Huawei’s Ascend 910B AI processors.
Igor Babuschkin, co-founder of Elon Musk’s xAI, departs the company
Igor Babuschkin, who co-founded Elon Musk’s AI startup xAI, is leaving the company to start Babuschkin Ventures, a firm that will fund AI safety research and startups aiming to benefit humanity. He helped grow xAI into a leading AI developer since 2023 but leaves after a series of controversies with its chatbot Grok, which produced biased, offensive, and inappropriate responses.
How Benchmaxxed is gpt-oss-120b?
OpenAI’s long-awaited open-weights model, gpt-oss, was released last week and scored well in benchmarks. But on closer inspection, some have raised concerns that the larger gpt-oss-120b variant may have been tuned to excel at those tests. I covered this in last week’s Sync, and here’s another sign that something shady may be going on with its benchmark results.
Inside the Summit Where China Pitched Its AI Agenda to the World
Wired takes us to the World Artificial Intelligence Conference in Shanghai, where China unveiled its Global AI Governance Action Plan, which emphasises global cooperation, UN involvement, and strict safety rules, and sharply contrasts with the Trump administration’s America-first, regulation-light approach. Both nations share concerns over AI risks, but while the US resists regulation, China actively sets and enforces standards, positioning itself as a potential global leader in AI governance despite its industry’s focus on growth over existential safety.
41% of all VC dollars deployed this year have gone to just 10 startups
PitchBook data shows that 41% of the $197.2 billion (or $81.3 billion) invested by US venture capitalists in 2025 has gone to just 10 startups—eight of them are AI companies. While backers cite the “winner-take-all” nature of the market, critics warn that such extreme capital concentration risks reduced diversification, inflated valuations, and a race driven more by FOMO than sound investing.
▶️ The AI Bandwidth Wall & Co-Packaged Optics (17:13)
The push for ever-greater compute power has hit bottlenecks in data transfer, particularly in optical IO, where growth has lagged far behind the rapid scaling of switch chips under Moore’s Law. In this video, Jon Y from Asianometry explores how Co-Packaged Optics (CPO) tackles these limits by moving photonics closer to silicon for better efficiency, lower power use, and space savings. Despite heat, coupling, and adoption challenges, the AI boom has accelerated uptake, with Nvidia, TSMC, Broadcom, and GlobalFoundries leading advances in faster, more efficient data centre infrastructure.
🤖 Robotics
▶️ Scaling Helix - Laundry (2:54)
Figure dropped a new video in which the company shows how good 02, its humanoid robot, is at folding laundry and showing human-like dexterity and movements.
Tesla Robotaxi scores permit to run ride-hailing service in Texas
Tesla has received a permit to operate a ride-hailing business across Texas, allowing it to compete with Uber and Lyft and paving the way for its autonomous “robotaxi” ambitions. The licence, valid until August 2026, enables Tesla to run a transportation network company without requiring human safety drivers.
Inside San Francisco's Robot Fight Club
In this article,
China’s industrial robot growth installation outpaces US, EU amid global market slump
China’s industrial robot installations rose 5% in 2024 to 290,000 units—54% of the world’s total—while other major markets, such as the US, Japan, and the EU, contracted. Production surged in early 2025, and robot density has nearly doubled in three years, with adoption expanding beyond electronics and automotive. China is also investing heavily in humanoid and embodied AI, signalling its ambition to lead the next era of global robotics.
🧬 Biotechnology
Scientists grow mini brain in a lab
Researchers have developed a “multi-region brain organoid” (MRBO)—a lab-grown miniature whole brain containing neural tissues from all major regions and rudimentary blood vessels. Resembling a 40-day-old human fetal brain, it can produce electrical activity and shows signs of forming an early blood–brain barrier. The MRBO provides a new and powerful human-cell model for studying disorders such as schizophrenia and Alzheimer’s.
How to Scale Proteomics
Meet Parallel Squared Technology Institute (PTI), a non-profit founded in 2023 to study the proteins inside single cells faster and more cheaply. They’ve created new barcoding and timing methods that let scientists analyse many more cells at once using existing machines. PTI is using these tools to track how proteins change in the brains of people with Alzheimer’s, hoping to spot the disease earlier. Their work could one day map every cell’s proteins in the human brain and help research many other diseases.
💡Tangents
▶️ Intel is Falling Apart (34:23)
Intel is not in a good spot. In this video, Gamers Nexus recaps Intel’s disastrous year, which includes failing to launch competitive CPUs; slowing, delaying, or cancelling fab construction projects; changing the CEO; and laying off tens of thousands of employees worldwide. Oh, and also Donald Trump called the current Intel CEO to step down. Intel promises to turn the ship around, but as Steve from Gamers Nexus argues, the company may be entering—or may have already entered—a downward spiral.
Perplexity offers to buy Chrome for billions more than it’s raised
Perplexity has offered $34.5 billion in cash to buy Google Chrome, promising to keep it open source, invest $3 billion in it, and leave Google as the default search engine. The bid comes after US regulators suggested Google should be forced to sell Chrome for breaking monopoly laws. Chrome has 68% of the browser market and may be worth over $50 billion, making the offer look cheap, though it’s far more than Perplexity’s current value. The move follows Perplexity’s recent launch of its own browser, Comet.
Self-cleaning glass uses electric field to remove dust particles within seconds
Chinese researchers have created a transparent, easy-to-produce self-cleaning glass that removes over 95% of dust in 10 seconds using an electric field, providing a sustainable, water-free alternative to conventional cleaning. It could keep windows, windscreens, and solar panels clear, and protect sensitive equipment in harsh or waterless environments on Earth and in space.
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!"
Any videos of robot fight club?
AI's effects will be limited until robotics progress accelerates. (I want to live in space but only after robots have built sophisticated habitats.) Please continue to provide lots of information on progress in robotics! Thank you for your work and reporting.