Superintelligence has partnered with Perplexity to give 50 readers early access to Comet, a new browser that can browse the web for you, fill in forms, manage emails and calendars, search your history, and help with research — see Comet in action — click the button below to enter!
In partnership with
Dear Readers,
How often do you see a mix of political pressure, technical genius, and pure improvisation in the tech world? China's DeepSeek is right at that point: between the dream of complete chip independence and the harsh reality that Nvidia hardware is still the gold standard. Behind the scenes, a race is unfolding that reveals more about the future of global AI power relations than any press release ever could.
In this issue, we take you inside the workshops of the AI revolution: from DeepSeek's risky bet on Huawei chips to groundbreaking research that uses AI to predict biological age from microbiome data, to a peptide that reactivates epigenetic “youth signals” in mouse brains. Also: new leaderboard surprises, leaks from the AI scene, and a question that affects the geopolitical balance. Dive in—the future is being written right now.
In Today’s Issue:
Why China's most advanced AI models are still secretly built on American chips
Your health records can now be used by AI to predict how fast you're aging
The bacteria in your gut might be a better predictor of your age than your ID
Could the "love hormone" be a key to reversing cognitive decline?
And more AI goodness…
All the best,
DeepSeek r2 release expected this month – here are the reasons for the delay
The Takeaway
👉 H20 dependency remains China's Achilles heel: Despite political pressure, all major Chinese AI companies – from DeepSeek to Alibaba – continue to use Nvidia's H20 chips because they are technically superior and have enabled China's greatest AI successes.
👉 Beijing tightens control over chip imports: The new requirement to justify H20 orders signals tighter monitoring of American technology dependencies while promoting domestic alternatives.
👉 Innovation through necessity: Pressure to use domestic technologies is forcing Chinese companies to come up with creative solutions and could accelerate the development of independent tech ecosystems in the long term.
👉 Geopolitical AI fragmentation accelerates: The forced separation of US technologies will create separate AI development paths that could permanently alter the global tech landscape.
The story behind DeepSeek's R2 development reads like a tech thriller: a Chinese AI company struggles with domestic chips, revealing both the strengths and weaknesses of China's path to technological independence.
DeepSeek was encouraged by the authorities to use Huawei's Ascend processor instead of Nvidia systems after releasing its R1 model in January. But the reality was sobering: despite a team of Huawei engineers working on site, DeepSeek was unable to complete a successful training run with the Ascend chip. The solution? A clever compromise: Nvidia for training, Huawei for inference.
This development is much more than just a technical problem. It reveals the complex dynamics between political pressure and technical reality in global AI development. Ironically, DeepSeek's R1 success – which had shaken the entire AI industry – was largely developed using Nvidia's H20 chips, the very chips that China wants to move away from. And it's not just DeepSeek: ByteDance, Tencent, and Alibaba – all of China's top players – continue to rely on H20 chips.
This is where it gets politically explosive: Beijing is now requiring local developers to justify their H20 orders, while the US only allowed the sale of these chips again in July. Industry insiders report stability issues, slower chip-to-chip connectivity, and inferior software in Chinese chips compared to Nvidia products.
But here's where it gets interesting: Latest reports suggest that DeepSeek R2 could launch this month with Huawei Ascend training and 1.2 trillion parameters. What does this mean for the future of AI development? Perhaps we are witnessing the moment when China is learning to go its own way – with all the ups and downs that come with it.
Why it matters: DeepSeek's challenges with Chinese chips illustrate China's paradoxical situation: while the country's greatest AI successes are based on American H20 chips, political pressure for technological independence is forcing companies to settle for suboptimal solutions. This tension between innovation and ideology will have a lasting impact on global AI development.
Sources:
Ad
Love Hacker News but don’t have the time to read it every day? Try TLDR’s free daily newsletter.
TLDR covers the best tech, startup, and coding stories in a quick email that takes 5 minutes to read.
No politics, sports, or weather (we promise). And it's read by over 1,250,000 people!
Subscribe for free now and you'll get our next newsletter tomorrow morning.
In The News
Runway's Aleph Reimagines Video
Runway has introduced Aleph, a new AI tool that allows creators to precisely replace, retexture, or completely reimagine specific parts of a video using simple text commands.
Gemini's New AI Tutor
Google for Education has launched "Guided Learning" in the Gemini app, a new personal AI learning companion designed to guide students step-by-step toward a deeper understanding of any subject.
Manage Jira From Google Chat
The newly enhanced Jira app for Google Chat helps teams stay on top of projects by letting them track progress, get real-time updates, and take action on issues, all without leaving their chats.
Graph of the Day
Open Model Leaderboard Update. Qwen-3 ranks first place, closely followed by Kimi-K2
Unexpected ability of large language models: predicting aging status
A Nature Medicine team shows that large language models can predict a robust “aging scale” from unstructured health data that correlates with morbidity and outcomes. New: LLMs are evolving from text generators to biomarker engines. Relevance: Scalable risk profiles directly from EHRs. Future: personalized prevention and study endpoints without expensive lab tests.
Chronological age estimation from human microbiomes with transformer-based Robust Principal Component Analysis
A transformer approach (TRPCA) predicts the age of skin, oral, and gut microbiomes up to 28% more accurately—and remains interpretable (SHAP). New: state-of-the-art AI + robust PCA for clinical explainability. Relevance: The microbiome as a controllable lever. Future: Targeted nutrition/probiotic interventions with measurable “age delta.”
Oxytocin Enhances Demethylation Through TET Enzyme Expression in Neurons of Aged Mice: Oxytocin as a Potential Antiaging Peptide
Intranasal oxytocin restores TET demethylation, 5hmC, and mitochondrial markers in aged mouse neurons and normalizes inflammation. New: A peptide triggers epigenetic “youth signals” in the brain. Relevance: a possible, safe path to cognitive preservation. Future: rapid translation via existing OXT application, initially as adjuvant gerotherapy.
Put your AI-driven Longevity Breakthrough in front of 200,000+ People
Working at the intersection of AI and human longevity? We’re inviting researchers and innovators to submit their work to be featured in Superintelligence, the leading AI newsletter with 200k+ readers.
Whether you’re using AI to extend lifespan, improve healthspan, or accelerate biomedical breakthroughs, email your arXiv.org link or a summary to [email protected] with the subject line “Longevity Submission”. If selected, we’ll contact you for a potential feature.
Question of the Day
Do you think China will overtake the US in AI?
Tweet of the Day
the ai skeptic’s most clever device is the score ceiling benchmark
performance always feels logarithmic on tests with 0 - 100% scoring
but when you look at no-ceiling benchmarks, we see a very different curve…
— #Aidan McLaughlin (#@aidan_mclau)
7:00 PM • Aug 10, 2025
Ad
Fact-based news without bias awaits. Make 1440 your choice today.
Overwhelmed by biased news? Cut through the clutter and get straight facts with your daily 1440 digest. From politics to sports, join millions who start their day informed.
Rumours, Leaks, and Dustups
Today was my last day at xAI, the company that I helped start with Elon Musk in 2023. I still remember the day I first met Elon, we talked for hours about AI and what the future might hold. We both felt that a new AI company with a different kind of mission was needed.
Building
— #Igor Babuschkin (#@ibab)
9:22 PM • Aug 13, 2025
Igor Babuschkin, co-founder of xAI is leaving the company.