This week JD made it clear that America is stepping back from its traditional role as Europe's security guarantor to focus on rising threats from China.
This pivot challenges Europe to take greater responsibility for its own defense and values, amidst many concerns over democracy, free speech, industrial production, supply chain sovereignty and military readiness.
The narrative is just starting to play out, but it certainly feels very real and the "vibe shift" has already completely happened.
Today Trump tweeted that he's likely going to treat VAT on imports as a tariff, and will reciprocate with similar levels of tariffs. If we were all friends a few years ago, now that's definitely not the case anymore.
There is also a clear vibe shift in Europe itself, that almost like a kick in the balls mixes the surprise, the pain, the anger and the vengeance. But it's definitely not fully distributed yet.
We could write a very long piece here, but don't think we have too much to add to what Noahpinion and Draghi published this week, so suggest reading that below.
Keeping in topic, I'm gonna go see for myself and spend next week (24-27) in LA / El Segundo - ping me if you're around!
It's time for Europe to stand up
Noahpinion with a good overview on the US vs EU narrative this week.
Forget the US — Europe has successfully put tariffs on itself
None others than national Marione Draghi who after his massive report keeps telling us that "radical change" is needed in Europe.
But something is moving, in this case a new 96 MW hyperscale data center (with a designed capacity of 50 000 GPUs) set to go live in Mougins, France, by 2025 Around €4B total investment here
AI
1. The intelligence of an AI model roughly equals the log of the resources used to train and run it. Translated: (scaling laws are holding)
2. The cost to use a given level of AI falls about 10x every 12 months, and lower prices lead to much more use. Moore’s law changed the world at 2x every 18 months; this is unbelievably stronger.
3. The socioeconomic value of linearly increasing intelligence is super-exponential in nature. No reason for exponentially increasing investment to stop in the near future. (There are many more observations in the post TBH, worth a read).
Automating GPU Kernel Generation with DeepSeek-R1 and Inference Time Scaling
DeepSeek is designing better GPU Kernels than NVIDIA engineers, nothing to see here.
Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention
For the more technically inclined, DeepSeek released a new paper on their NSA (quite fun pun there) architecture. By integrating hierarchical token compression with blockwise token selection within a trainable architecture, our architecture achieves accelerated training and inference while maintaining Full Attention performance. NSA advances the state-of-the-art by demonstrating general benchmark performance matches full-attention baselines, exceeding modeling capability in long-context evaluations, and enhanced reasoning ability, all accompanied by measurable reductions in computational latency and achieving significant speedup.
Robotics
The Download: China’s EV to humanoid robot pivot, and voice clone censorship
Many Chinese electric vehicle manufacturers are pivoting into humanoid robotics, leveraging their established tech prowess and government support to potentially lead the global robotics market.
Why Robot Modularity Is So Evasive, But Also Inevitable
Modularity in robotics means enhanced adaptability, cost efficiency, and potentially sustainability. But for modularity to happen, you usually need to have standards, and potentially open source.
Chinese robot maker UBTech eyes mass production of industrial humanoids by year end
UBTech will deliver about 500 to 1,000 units of its Walker S Series industrial humanoid robots this year to customers and partners. Agibot also already produced more than 1000. Humanoids are coming, en masse, from China.
Muscle tissue meets mechanics in biohybrid hand breakthrough
MuMuTas: thin strands of muscle tissue grown in a culture medium and then rolled up into a bundle like a sushi roll to make each tendon. This means ensuring enough contractile force and length in the muscles to drive the hand’s large structure. The world of biohybrid science is here.
Science
See WiFi Signals with an ESP32 Antenna Array
A guy just casually dropped on Youtube a massive new innovation in defence / secruity. He developed a super cheap ESP32 antenna array that can visualize WiFi signals in real-time, mapping their propagation, reflections, and even tracking movement through walls. The system uses synchronized receivers, phase coherence, and AI-based channel charting, to enable passive radar, indoor localization, and high-resolution WiFi imaging. It is an incredibly powerful tool with myriads new applications.
AI Reveals Hidden Interior Design Rules of the Cell
ProtGPS deciphers hidden molecular signals that determine protein localization within cells and enables precise mapping of protein destinations. History is a slow realization of Laplace's deterministic dream.
Absolutely epic new way to create transparent aluminium, by just dropping some chemicals on it. Potentially game changing for a wide area of materials.
Quantum Encryption Breakthrough Uses Light and Color to Create an Unhackable Internet
It might just be my feeds, but it does feel like quantum research is progressing ridiculously fast. This new encryption technique uses light frequencies, i.e. colors, to encode quantum states.
Using AI to decode language from the brain and advance our understanding of human communication
By successfully decoding sentences from non-invasive brain recordings, they are not only enhancing our understanding of human intelligence but also offering hope for restoring communication abilities in those affected by brain injuries.
Other stuff we read
Unbundling the BPO: How AI Will Disrupt Outsourced Work
A16Z dropped a new thesis on AI capturing business process outsourcing (BPO). The move from outsourced to in-house AI powered will shift trillions in the economy if startups move faster than the incumbents can repurpose themselves (which is usually true)
This week Alex Karp's book come out. Will try to give it a read on the airplane and report back.
Cantos seems to be having enough of LPs putting them in the deeptech bucket. We tend to agree.
Turning Soldiers into Superheroes
Palmer Luckey is back (was he ever out?) "Whatever you are imagining, however crazy you imagine I am, multiply it by ten and then do it again. I am back, and I am only getting started." But he is definitely going to apply all of his genius combined with his VR/AR experience for military uses.
Healthspan Capital's 2025 Longevity tech report, worth a read if you're into this stuff.
A writing fellowship on scientific progress
Cool new writing fellowship by Works in Progress and Asimov Press (both very very good).
The Social Renaissance: Make Internet Social Again
Not something we know a lot of, but was good to learn more about the death of the feed and what AI means for consumer social. Tons of good examples too.
Quantum computing startup QuEra closes $230 million funding round
Quantum will happen. Google/Alphabet seems to be everywhere when you mention it.
KPMG partners see record year, rest of the company does not
KPMG's partners made a killing last year, and did so while: - only 1% revenue growth - 12000 employees on wage freezes - graduate hiring down 33% It is a new world, coming at us fast.
Lightspeed Venture Partners' Returns & Strategy
Newcomer has the scoop on Lightspeed returns (spoiler: they aren't that great). Behind a paywall, but if you search on twitter someone has spoilered a bit of it.
Conviction open sourced their LP letters, showing how their thinking has evolved in relation to the massive AI advancements of the last couple of years.
Perplexity Introduces Deep Research
The race to replace your analyst is on. Perplexity launches its own deep research at a fraction of the cost of ChatGPT Pro.
LLMs have started a new chapter in the gold rush for high-quality data.
What I Learned From My Optimizer
One exploration into how optimizers enhance discretionary investing by refining decision-making, mitigating biases, improving portfolio construction and risk management, revealing trade-offs in financial markets.
Thanks for taking the time to read this! We hope you enjoyed it, reach out if there's anything we missed!
