Europe’s Deeptech Ambition: Trust, Alliances and the Capacity to Act

An evening with Supernova Invest on trust, industrial alliances and the commitments that give Europe the capacity to act.

Journal

all →
Nov 5, 2026 FIELD
Europe’s Deeptech Ambition: Trust, Alliances and the Capacity to Act
An evening with Supernova Invest on trust, industrial alliances and the commitments that give Europe the capacity to act. read → 16 minLONG READ
#deeptech #trust #sovereignty #Europe #investment
Oct 26, 2026 ESSAY
Can We Control AI Without Understanding It?
What AI safety research tells us about the limits of control, the risks of delegation, and the pressure to trade oversight for efficiency. read → 22 minLONG READ
#AI #agents #safety #alignment #governance #cybersecurity
Oct 19, 2026 FIELD
The French AI Report 2026: from models to medicine and defence
Notes from our evening at Google France: the state of France’s AI startup ecosystem, growing tumours in the lab, bringing AI into hospitals, and building defence companies. read → 23 minLONG READ
#AI #France #VC #health #biology #defence #robotics

Deep Dives

all →
Long-form essays and multi-part series: strategy, finance, technology.
Current series
How Machines Learn the World How far can general models go in specialised fields, and where do they still need experiments and domain expertise? · Series started Oct 2026 · 1 of 3 sections published
Part 0 · Introduction 14 min Oct 2026
Part 1 · What Should a General Model Know? upcoming
Part 2 · The Last Bit Has to Come From the World upcoming
Part 3 · Why There Is No AWS for Biology upcoming
Series
Reading Nvidia: On strategy, defensibility, and the return of industrial AI Moats, systems, and the physical economy behind the AI build-out. · Series started Apr 2026 · 5 of 5 sections published
Part 0 · Series overview 5 min Apr 2026
Part 2 · The Moat That Cannot Be Coded 24 min LONG READ May 2026
Part 3 · Pareto Frontiers and Lock-In 25 min LONG READ May 2026
Part 4 · The Return of the Real World 18 min LONG READ Oct 2026
Recent standalones
AI Moats Are Different Oct 2025 Why traditional moats fail in AI, and a four-level framework for building defensible positions: technical foundation, product depth, market position, sustainability. · 16 min LONG READ
Market Sizing for Founders Oct 2025 A practical framework for sizing markets when you're building something new. · 23 min LONG READ
AI as Normal Technology Sep 2025 General-purpose technologies diffuse slowly and unevenly. AI will follow similar patterns. Value creation depends on when bottlenecks collapse, not benchmark scores. · 21 min LONG READ

Library

all →
Recently read
What We Owe the Future William MacAskill #philosophy #AI #ethics #longtermism ★★★★ my notes
The New Map: Energy, Climate, and the Clash of Nations Daniel Yergin #energy #geopolitics #AI ★★★★★ my notes
Aliénation et accélération Hartmut Rosa #philosophy ★★★★★ my notes
Development as Freedom Amartya Sen #philosophy #economics #development ★★★★★ my notes
Why Nations Fail: The Origins of Power, Prosperity, and Poverty Daron Acemoglu & James A. Robinson #economics #institutions #strategy ★★★★★ my notes
Steak Machine Geoffrey Le Guilcher #animals #food #journalism ★★★★ my notes
Changement de quart Bruno Colmant, Laurent Hublet & Marie Vancutsem #economics #geopolitics #AI ★★★★ my notes
Énergie et inégalités : une histoire politique Lucas Chancel #energy #economics #inequality ★★★★ my notes
Innovations : une économie pour les temps à venir Dominique Foray #economics #innovation #policy ★★★★★ my notes
Multinationales : une histoire du monde contemporain Olivier Petitjean & Ivan du Roy #economics #history #corporate-power ★★★★★ my notes
La planification écologique Mathilde Viennot #economics #ecology #France #policy ★★★★★ my notes
What Is Life?: Evolution as Computation Blaise Agüera y Arcas #AI #tech #biology #philosophy ★★★★★ my notes
Energy and Civilization: A History Vaclav Smil #energy #history ★★★★★ my notes
Technology and the Rise of Great Powers Jeffrey Ding #geopolitics #tech #strategy ★★★★★
Strategy: A History Lawrence Freedman #strategy #history ★★★★★ my notes
These Strange New Minds Christopher Summerfield #AI #philosophy ★★★★★ my notes
How To Think Like Socrates Donald Robertson #philosophy ★★★★★
Reading list
Atlas of AI Kate Crawford #AI #geopolitics
Chip War Chris Miller #tech #geopolitics #history
L'Évangile selon Pilate Éric-Emmanuel Schmitt #fiction #philosophy
L'Individu, fin de parcours ? Julien Gobin #philosophy #economics #France my notes
La mort et le printemps Mercè Rodoreda #fiction
La Peau de chagrin Honoré de Balzac #fiction #philosophy #France my notes
La Pharmacie de Platon Jacques Derrida #philosophy #France my notes
Phenomenology of Spirit G. W. F. Hegel #philosophy my notes
Selfpressionnisme : Et si l'IA nous rendait plus humains ? Marie Dollé #AI #philosophy #France #humanism my notes
Tentative d'épuisement d'un lieu parisien Georges Perec #literature #France my notes
The Investment Philosophers Erik Everett #finance #philosophy
Water: The Epic Struggle for Wealth, Power, and Civilization Steven Solomon #history #geopolitics
Top reads
What AI’s Mathematical Breakthroughs Mean for Specialised Models deep dive How far can general models go in specialised fields, and where do they still need experiments and domain expertise? · 14 min
AI Moats Are Different deep dive Why traditional moats fail in AI, and a four-level framework for building defensible positions: technical foundation, product depth, market position, sustainability. · 16 min LONG READ
On Systemic Errors journal A frequent critique of GenAI is that because it is probabilistic, it will inevitably make… · 9 min
Foundation Models in Biology Hit the Noise Floor journal A new benchmark of 600+ model configurations reveals that biological foundation models do… · 13 min
Market Sizing for Founders deep dive A practical framework for sizing markets when you're building something new. · 23 min LONG READ
Food for Thoughtall →
Exploring the investor landscape for venture capital
European Central Bank · 18 min · 20-11-2026 report
An ECB box maps who actually funds venture capital on each side of the Atlantic: US VC funds total about €930B against roughly €150B in the EU, invest around six times more, and keep 80% of their money at home, while more than half of EU-located funds' investments go to firms outside the EU. Governments make up about a third of LPs in EU funds versus 4% in the US, where pension funds and foundations dominate, so the authors argue Europe needs its own ERISA moment, with the EIF as the hub to crowd in institutional money.
Nvidia's interim CEO Colette Kress has taken over one of the most consequential companies on the planet; those who know her say she's up to the task
Yahoo Finance · 5 min · 18-11-2026 article
Yahoo Finance profiles Colette Kress, the low-key CFO who now runs Nvidia after Jensen Huang's resignation, and who just reported $115.8B of quarterly revenue on her first call as interim CEO. Colleagues from Ian Buck to Cisco's Chuck Robbins vouch for her product depth, while analysts like Bernstein's Stacy Rasgon stay on the sidelines, worried about talent retention, China and a permanent CEO search.
Superintelligence Needs the Physical World with CEO and Co-Founder of Arda Bob McGrew
No Priors · 70 min · 20-11-2026 podcast
OpenAI's former chief research officer Bob McGrew tells Sarah Guo and Elad Gil why he left to automate manufacturing rather than start another frontier lab: models are powerful, yet the physical world looks much as it did ten years ago. His case is that factory automation is bespoke and doesn't generalize from one product to the next, so Arda wants manufacturers to upload a design and reconfigure an automated line in seven days, the speed at which humans can be retrained.
The Rocky Road from Aggregate to Per Share Value: Anthropic's IPO update
Musings on Markets (Aswath Damodaran) · 17 min · 18-11-2026 article
On the eve of the listing, Damodaran updates his Anthropic valuation to about $1.35 trillion of equity, trimming it for smaller IPO proceeds, costlier data centers and a heavier regulatory drag after the midterms. The real lesson is about share count: depending on how you treat 1.8 billion options and a billion restricted shares, value per share ranges from $76 to $91, and his preferred option-adjusted figure of about $76 sits below the $83 offer price, with a warning about the power concentrated in the founders' single golden share.
5 key findings from Atomico's State of European Tech 2026
EU-Startups · 5 min · 18-11-2026 article
EU-Startups distils Atomico's State of European Tech 2026: investment is set to rebound 62% to €73B, the second-best year on record, while 13 billion-dollar rounds took 37% of all capital and Europe now produces more new founders than the US (29% vs 28%) even as first-round funding keeps shrinking. The sobering part is AI, where the US out-invested Europe 14 to 1 (€258B vs €18B) and holds about 72% of global compute against Europe's 7%.
Research acceleration: The jagged frontier
OpenAI · 15 min · 18-11-2026 report
OpenAI's second report on how coding agents speed up its own research finds agents now complete 46% of two-to-four-hour research tasks on their own, up from 23% in March, but unevenly: 79% for monitoring training runs, 25% for writing inference code. Agent use is surging, with the median researcher burning $2,400 of tokens a day at list prices, and the lab argues that every frontier lab should publicly report such indicators of recursive self-improvement.
2026: The State of AI in Healthcare
Menlo Ventures · 18 min · 16-11-2026 report
Menlo Ventures' annual survey finds healthcare has become AI's biggest vertical market, with spend up 2.5x to $3.7B (more than legal and finance combined) and pharma's AI budgets jumping 23x to $700M. The first wave of scribing and billing tools is commoditizing as Microsoft and Epic captured two of every three new scribe dollars, pushing growth toward clinical copilots (5.5x) and patient engagement (4.5x).
Harvard researchers find AI coding agents shifted the bottleneck to code review
ComplexDiscovery · 10 min · 13-10-2026 article
Drawing on a Harvard working paper by Fiona Chen and James Stratton covering 725,938 workers at 718 firms, this analysis shows that engineers using AI coding agents wrote about 30% more code but shipped no measurably more finished work, as the time from pull request to merge rose 49% and change requests nearly doubled. Its takeaway for buyers of AI tools, legal ones included: measure the human review checkpoint, not drafting speed.
How Fast Do You Really Have to Be Growing to Raise Venture Capital Today? $1M to $100M ARR in 5 Years, Max. But VCs Are Hoping for 2-3 Years
SaaStr · 8 min · 18-11-2026 article
Jason Lemkin argues that raising venture capital now takes going from $1M to $100M ARR in five years or less, about 151% compounded growth, which demotes Battery's once-aspirational T2D3 to a bare minimum. He traces the bar back to compressed exit multiples (8.3x forward revenue for top growers versus a 10.8x ten-year average) and AI-era comps like Cursor, and spells out where the rule breaks, from deep-tech infrastructure to strong net retention.
Alex Imas and Tyler Cowen on whether AI will deliver runaway economic growth
a16z Podcast · 58 min · 17-11-2026 podcast
Economists Alex Imas and Tyler Cowen push back on the double-digit-growth story: AI's benefits will exceed its costs, but bottlenecks, slow institutions and agents that still coordinate poorly mean the path will look far more 'normal' than the hype suggests. They also argue GDP will undercount AI's gains to consumers, from cheaper models to science and longer lives.
Handing over humanity
Ruxandra Teslo's newsletter · 16 min · 11-11-2026 essay
Three weeks after moving to San Francisco, Ruxandra Teslo finds that even the 'good' post-AGI scenarios floated by lab insiders amount to 'handoverism': humanity survives but cedes the direction of its future to an aligned AI, which she estimates 30-40% of the AI-safety community accepts in some form. Surveying lab leaders' own statements (Tom Brown's 'handoff', Amodei's Machines of Loving Grace, Culture-style utopias), she warns that a jobless, AI-steered world leaves little room for democracy, work as leverage, or human self-determination.
State of AI 2026: The Biggest Stories of the Year (and AGI in 2027?), with Nathan Benaich
Matt Turck podcast · 65 min · 17-11-2026 podcast
London-based investor Nathan Benaich walks Matt Turck through this year's State of AI report and its boldest call yet, AGI in 2027, driven by recursive self-improvement. Along the way: cyber risk as the more real danger, the data-centre backlash, and a sobering European data point: all 27 EU countries together hold roughly 80,000 H100-equivalents, fewer than any of the world's top five GPU clusters.
Welcome to the B2A Economy
Andreessen Horowitz · 7 min · 10-11-2026 article
Olivia and Justine Moore argue the internet's primary users are now AI agents, and that the missing half of the AI platform shift is distribution: agents will increasingly pick the products, subscriptions and purchases on our behalf. Pointing to coding as the preview (40% of GitHub PRs now start from an agent draft, most new Supabase and Neon databases are spun up by AI), they map who wins in 'business-to-agent': agentic models, agent-native products, incumbents that optimise for bots, and the payments, identity and communication infrastructure agents will need.
Jensen's fall and the world's chokepoint nation
James Crabtree's newsletter · 9 min · 15-11-2026 essay
Back from Kuala Lumpur, James Crabtree argues Malaysia has become the world's 'chokepoint nation', sitting on chip packaging (about 13% of global assembly and testing), Johor's booming data centres (roughly half the EU's compute) and rare-earth processing. Washington never squeezed it, his sources say, because Nvidia's sway over the White House kept export rules loosely enforced; with Jensen Huang gone, that comfortable balance looks set to end.
The $5B+ Cash Cow That Let Rockstar Take Its Time With GTA VI
SatPost (Trung Phan) · 12 min · 13-11-2026 article
Why did GTA VI take 13 years? Trung Phan traces it to GTA Online's Shark Cards, a roughly $425m-a-year annuity that let Rockstar hire more developers and spend an estimated $1-1.5bn without deadline pressure. The bet looks paid off: pre-sales passed $1bn a week before launch, a neat case study in how recurring revenue buys a creative studio patience.