Car les entrepreneurs sont les meilleurs allocateurs de capital mon cher. Il faut des 100 aines d’Elon Musk, pas moins. Revoir le discours de Milei à Davos. La donnée est très claire. Les bureaucrates ne savent pas allouer le capital de manière efficace. Les entrepreneurs sont les vrais créateurs de valeur nette. Vive les entrepreneurs.

En annan sak att fundera på om nu EU ska bli självförsörjande på AI. Inte bara att man ska äga modellen utan även datacenters. Och då behöver man inte en AI-plan utan en energiplan. Om vi ska matcha prognosen för USA så behövs ca 100 nya reaktorer. axios.com/2025/12/01/data-ce…

I am 27, French, and I am tired of living on a continent that treats AI, compute, chips, crypto, datacenters, energy and nuclear power as problems to manage instead of strategic assets to build. I do not want frontier AI to become another nationality-gated privilege. I want powerful AI models to remain generally available to builders, researchers, engineers and founders. But what happened with Anthropic’s Fable/Mythos models proves that this cannot be taken for granted: once frontier AI becomes a national-security asset, access can be restricted by citizenship or nationality. The problem is that Europe has failed to build its own equivalent. We are not in the frontier AI race at the level of the U.S. or China. We do not have the same hyperscale cloud stack, the same compute capacity, the same capital depth, the same energy strategy, the same chip ecosystem, or the same frontier-model ecosystem. And because AI progress compounds through compute, talent, chips, energy, data and capital, falling behind is not linear. Once the gap is deep enough, you do not catch up at the same pace. Europe spent decades regulating, moralizing, delaying and underbuilding the foundations of technological power. Cloud was missed. Crypto was treated primarily as a criminal-risk category before Europe built anything globally dominant in it. Datacenters are slowed by permitting, grid and energy constraints. Nuclear power was politically weakened or delayed across much of the continent just when abundant electricity became essential. AI is now being regulated before Europe has even produced a true top-tier frontier lab/model (no, MistralAI isn't a real competitor, for me, even Kyutai did more innovation/progress in the AI space than MistralAI). Our leaders now talk about "sovereign AI", "AI factories", "gigafactories", and "strategic autonomy", but this language came far too late. You cannot regulate your way into technological sovereignty. You cannot paperwork your way into compute. You cannot build frontier AI without massive power, massive datacenters, massive capital, elite talent, advanced chips and a political culture that actually wants builders to move fast. Europe still has talent. France still has engineers, mathematicians, scientists and founders. But the system around them is broken. The incentives are wrong. The mindset is wrong. Every mainstream political camp in France and Europe seems to have the same reflex: regulate first, tax first, restrict first, moralize first, build later. ASML is the exception that proves the rule. It is one of the only truly strategic European chokepoints in the global compute stack. But one Dutch lithography champion cannot carry an entire continent that failed to build the rest of the stack: frontier AI labs, hyperscale cloud, Nvidia-class accelerators, TSMC-class fabs, massive datacenter capacity, cheap abundant energy and deep capital markets. I did not vote for 20 years of anti-growth, anti-compute, anti-nuclear, anti-crypto and anti-industrial policy. I was a kid. But my generation is supposed to live with the consequences: less access, less sovereignty, less capital, less compute, less ambition and a future where the most important technologies are built and can only be used somewhere else. That is the part I cannot accept. I do not want to spend my adult life asking permission to use technologies my continent was too slow, too afraid or too complacent to build. I do not want European builders to become tenants in someone else’s technological empire (as it's already the case). And I do not want "sovereignty" to mean nothing more than regulating foreign systems after failing to create our own like they're doing right now with cloud computing. Either Europe becomes a builder civilization again, or the next generation of Europeans will inherit a beautifully regulated dependency that slow or even stop us. For now, Europe still talks like history will wait...

Fixing the issues that are making Europe uncompetitive would necessarily entail getting rid of a lot of that comfort. No more infinite pension money for boomers, no more inability to get fired, no more endless vacation required by law.

Who believes in building the digital gods in Germany? Due to extreme CapEx and winner takes all dynamics training frontier foundation models or monetizing any 2nd class mode is hard. Every datacenter not planned today will likely be finished after US and China reached a clear ASI level. The scaling laws, the AlphaGo moment the solved Turing test - simply did not wake up enough people here. To make Germans move you need a real substantial shock or you need to build up ideologies for decades well accepted in the intellectual circles. You can ban nuclear energy after Fukushima or implement the EU Green Deal. But regarding AI we are in complete reactive mode - there is little discussion how the society will or should look post 2030. We ramp up our military for a potential conflict with Russia. Politicians complain about skilled workers shortage in the public sector - something which for me seems completely ridiculous and ignorant - politicians don’t guide anymore (our chancellor has 77% disapproval rate, far right AfD is strongest party with 27% in our state media polls), they maintain decay or in the AfD case ride on frustration and fear - we are stuck while the world progresses, China fully caught up with our engineering quality and exceeding us in by far in AI and automobile digital technologies etc. Sooner or later this state must and will be overcome - in worst case it will happen through a war, in less bloody scenarios change comes because the unemployment goes through the roof and everyone sees that we are no longer at the top and some reforms break free - but as soon as life becomes miserable for millions the opportunities for the political extremes open widely. For sure there is an alternative, we could start building our version of an EU-Utopia. What we do now will define where our smartest and most talented founders and entrepreneur will head and if prosperity is build here or elsewhere - looking at my contact lists and seeing all the Germans who already moved or plan to work in the U.S., for AI the gradient clearly is not pulling towards EU. The widespread realization is: If you want to work on the frontier, try to land a job in a U.S. frontier lab - maybe at DeepMind and far behind then Mistral. So - quo vadis 🇩🇪 ? We need either a believable artificial intelligence based post de-industrialization or a robotic re-industrialization perspective. On both fronts we are currently losing, on the AI side we did and still do too little too late and we don’t have cheap Russian nor (less cheap) nuclear energy anymore to melt&weld. It is time for change - even when most are still in denial and the full effects haven’t surfaced - it is unfortunately likely that we will see more economic downturn in Germany. I always hope for the best.

Nice write-up of why Europe and others are fucking up wrt AI but also, you can’t have nice things in any country that taxes you to death, 40-60% deductions on income is not a viable state of affairs and the socialist notion of fairness benefits no one exc. managerial bureaucrats

Reminder that there are people with 160-190 IQ who actually intellectually mog Mythos (which probably takes a whole Rubin pod to run), while their brain runs on ~20W. We aren't being bullish enough on the biological substrate. Humanity is still early. It's time to bio/acc

Here's very simple recipe how to build an tech/ai hub like San Francisco in europe: Pick one city and stop turning everything into a pan European policy project. Put the best researchers, founders, engineers, designers, and infra people within walking distance of each other. Pay them like the US or comparably. Give them compute without a 12 month grant process. Let startups move fast, fail fast, and restart without stigma. No summit, no committee, no 180 page framework. Just talent density, capital, GPUs, speed, and permissionless ambition. The whole thing is not that complicated. You need density, money, GPUs, speed, and a culture that doesn’t treat ambition like a character flaw.

The “regulatory superpower” nonsense was premised on Europe being such a large market that multinationals would consent to being regulated globally in order to gain access to it. The conceit I think was that everybody would be forced to go at Europes relaxed pace, and that Brussels could take a central role managing the End of History for the benefit of all. That’s in tatters now. History didn’t end, Europes importance as a market is diminishing, and decades of complacency have left us without the infrastructure to compete.

all these wistful mobilization fantasies in "what could have been" are slop. They need ALL POSSIBLE INPUTS, NOW, WHILE THEY HAVE ANY LEVERAGE. Chinese industry, double dealing on ASML, US datacenters, any energy. They will just be RAPED AND DISCARDED BY 2028 at this rate.

Upon more interactions with Fable, I need to issue a Mea Culpa This is provisional, but I think true: I have been completely wrong about Anthropic. They had me – and everyone - well and truly fooled. I failed to grasp the profundity of Dario's "scale". Anthropic is a lab of *scientists*. Proverbial triple Ph.Ds, Mahattan Project material. Safety/steerability focus served as a good recruitment pipeline, unifying ethos, and a smokescreen. In the meantime, publishing neat visualizations and results of curious safety-framed experiments, they must have developed a proper *science* of LLM circuitry, the missing layer between optimization theory, academic math validated on toy models – and downstream humanlike behaviors on the frontier. Us plebs outside think in these petty terms of "1-3-10-100T models" and GPU arsenals, only aware of crude undergrad tier problems like distributed training implementations, exploding gradients, loss spikes, router collapse and so on, entirely ignorant of how artificial intelligences develop at larger scales… or really at any scale. We have some alchemical, witch doctor understanding of data mixing and "quality", and even buy the copes that the era of pre-training is over, or that Anthropic's real advantage is just investment into data, Amanda Askell's constitution, the commercial focus on agentic coding. "What are you scaling?", asks Ilya. How about you scale your understanding of what the fuck you're trying to do? How about you try to get out of this lame attractor of ever more exact memorization of ever greater volume of data slathered onto ever-expanding blanket of weights, hopelessly asymptotically approaching flawless mediocrity? …No, I don't believe that OpenAI's pretraining team is a shitshow. They must be about as good as Google and top Chinese labs. They have great infra, they have the hardware, they can definitely train a "15T model" if they put their minds to it. Except that's not enough. And that likely puts a cap on how far "post-training" can go. If post-training is even a necessary category when you do your *training* right, in the limit. I have always been saying that mechanistic interpretability is dual use, and can advance capabilities; doomers also thought this way; somehow, it didn't have an impact on the discourse, or even on my thinking of the competitive landscape. I failed to extrapolate. If Chris Olah's research program had quietly advanced to the level of physics, chemistry or even biology of multi-layer computational organisms, just as it was intended to – then Dario holds the commanding heights in the foreseeable future. When ByteDance tries to even think in these terms, it looks ludicrous, a fever dream or pretentious LLM slop. But seriously – do you believe that we were going to build AGI, or ASI, with our rules of thumb, muh "'lots' of 'good' 'diverse' data", with this dumb piling of chairs? I don't want to doompost. GPT-4 seemed like an unreachable standard as well. Capabilities diffuse; what was done once has, historically, usually been replicated on a shorter timeline; lots of smart people are working on it. Anthropic was just early. Maybe it's not too late. But boy, if I'm right, were they early.

"... the political left has long had a remarkable lack of interest in how wealth is created. As far as they are concerned, wealth exists somehow and the only interesting question is how to redistribute it." — Thomas Sowell

Det är inte svårt att föreställa höstens val som en tipping point. Jag tänker ofta på människor jag träffat som berättat att de planerar att lämna landet om Magdalena Andersson blir statsminister eller varnat för att personer i deras närhet kommer att göra det. Människor som fruktar att landet i det scenariot är bortom räddning. Som är förmögna och har möjlighet att bosätta sig var de vill. Till skillnad från rika amerikaner som inte sällan högljutt varnar innan de rycker undan mattan och skatteintäkterna för delstaterna de överger i protest, lär svenskarna lämna i tystnad. Göra sorti för en tryggare tillvaro i Schweiz, Spanien eller USA utan att göra så stort väsen av sig. Det svenska offentligheten premierar som bekant inte kritiker snett utanför åsiktskorridoren. Bättre då att smita ut bakvägen. Kvar blir jantelagen och en befolkning som får svårt att betala för välfärdsstaten den röstat fram. Den som sades vara så stark att den kunde ta människor från hela halvklotet under sina vingar. Problemet är att alltför många varken visar lojalitet eller vilja att bidra till landet som försörjt dem. Det är inte bara deras eget fel. I veckan befann jag mig i Upplands Väsby där poliser bevakade ett studentutspring. Lyckliga ungdomar sicksackade mellan polisbussar. Många av dem hade svept in sig i flaggor. Av flaggorna jag såg var ingen svensk. Efter tolv år i svensk grundskola är det inte Sverige man firar. Det är inte så konstigt. Sverige har, liksom stora delar av Europa, misslyckats med att berättat vad vi står för, vad människor ska invigas i. I USA vet alla vad man kämpar för. Löftet är att den som, oavsett varifrån den kommer, gör rätt för sig och arbetar hårt har chans att ta del av den amerikanska drömmen. Belönas med ekonomisk framgång och trygghet för sig själv och de sina. Europa har varit sämre på att berätta vad hårt arbete kan rendera. I stället har invandrare serverats förmåner som till slut tagits för givna. En ansenlig del av befolkningen har reagerat på undfallenheten och destruktiviteten som på olika sätt följt. I Storbritannien demonstrerar människor efter att polisen, som i Black Lives Matters spår tränats i att förhålla sig till offers och gärningsmäns hudfärg, lät 18-årige Henry Novak blöda till döds efter att gärningsmannen Vickrum Singh Digwa påstått att Novak uttryckt sig rasistiskt. Och nu rasar upplopp efter att sudanesen Hadi Alodid, som beviljades asyl efter att ha tagit sig in i Irland illegalt, försökte hugga huvudet av 40-årige Stephen Ogilvie. Offret vårdas på sjukhus med allvarliga skador men överlevde tack vare civilkurage från medmänniskor, som även filmade det brutala våldet. Sverige är ett eget land med egna problem, men borde se Storbritannien som ett varnande exempel. Inte bara vad gäller migrationen. En av världens stoltaste demokratier har sjangserat i totalitär riktning. Varje dag grips britter för att ha delat sina – ofta helt vanliga och av majoriteten delade – åsikter online. För mig är det förbluffande att ledaren för Sveriges ännu största parti samtidigt vill införa en polis för “våra digitala gator och torg”. Men sedan spatserade också Magdalena förbi mig i sina gympaskor på inspirationsresa under Demokraternas konvent i Chicago, kort innan Harris historiska valförlust. Är jag överdrivet hård? En vän som är höger argumenterade för att rösta på Socialdemokraterna kanske inte är så dumt. Man ska inte underskatta viljan till makt, ansåg han. Att det faktum att partiföreträdarna är mer pragmatiker än ideologer kommer förmå dem fatta de beslut som behöver fattas. Även om de strider mot allt man tidigare stått för. Partiets tidigare kappvändande tyder på att min vän kan ha rätt. Själv har jag svårt för hyckleriet. Misstag är okej, men måste ägas. Och för mig är en ledare inte någon som till slut ger vika och följer strömmen, utan någon som banar väg från början. Annars är risken att rätt beslut fattas alldeles försent. Läs hela texten i Stockholm Report.

Ok. Finished reading it. The story of Europe's failure in AI is turned into a gripping story (congratulations to the authors on finding this way to write it) and an outstanding SHOUT for action. I disagree with many things in this scenario (e.g. ASML cannot be used for leverage, I am afraid: all the EUV tech is San Diego-based (Cymer), and the chips are Nvidia, AMD, Intel, etc.) But the key insight is correct: (1) AI is THE critical technology of the future, and (2) Europe is falling badly behind on AI and running out of options. Both the economic and strategic consequences are brutal. We will write a reaction in Silicon Continent. In the meantime, please do read it. europe2031.ai/#timeline

Läs detta om något

A plausible and frightening vision of Europe's near future, ending in economic collapse and vassal status under the US or China. Enjoyed reading it, although I feel a little depressed now. europe2031.ai/

What we need to do now is get together and plan the legal road to remove this Government and work together. But one thing NO ONE OUTSIDE THIS COUNTRY will be involved in any decisions going forward.

Sharing the following message. Belfast's paramilitaries will be out tonight protesting against the attempted beheading of a Belfast man by a Somalian. In the face of this, the government is likely to fold and remove problem migrants and illegals from these communities, just as they did in Ballymena, and move the migrants to non British parts of Britain. Other communities across the UK will then look at Belfast and Ballymena and see that paramilitary groups are the only way to get the government to listen. This is our future.

Cultivated meat is not on track to ever be economically competitive with animal meat. Modern factory farms have cut out almost all costs that aren't strictly biologically necessary for the production of muscle. And animal evolution has been optimizing the conversion of energy into muscle for billions of years. That means that cultivated meat needs to beat billions of years of accumulated evolutionary efficiency to become competitive. That's an extremely hard challenge. It's maybe even harder than building AGI, since evolution has only been optimizing for intelligence for tens of millions of years.

this is my personal singularity moment this post may sound like a paid ad. I only wish. I'm concerned, more so than happy. the world is changing, and, among the scenarios where AI goes terribly wrong, inequality is the most realistic, yet, the one Anthropic seems to be the least concerned about. I'm glad OpenAI is taking the opposite stance: *personal AGI for everyone*. I think this is a commendable position in the times we live. but who am I in the queue of the bread? anyway, Fable is here, so I'll just report my first-hour experience first of all, all my pet prompts are solved. → λ-calculus puzzles → bug questions → one-shot apps all are trivial to it. I don't have anything harder other than my ongoing work so, in the last several days, I've been toying with HVM5, a new interaction net evaluator with a faster loop. after writing the first version, I left 32 GPT-5 agents working for ~20 hours each. this resulted in up to 2x speedups, but the file size increased by 2-fold and quality decreased significantly. I then simplified the whole thing into an even simpler core, and left Opus 4.8 and GPT 5.5 optimizing it for 8 hours. Opus got a legit 6% - 34% speedup in most benches. GPT got better results, but, sadly, an unusable file. I then asked Fable to optimize it. 2 hours later, it landed a 1770% speedup in one case, 100% in other 4, and 22% in average. yes, in 2 hours it outperformed me, opus 4.8 and a swarm of gpt 5.5 agents, by one order of magnitude. that could not possibly be legit. "it must be hardcoding the benchmarks" (GPT trauma). so I read its explanation and what it did was, indeed, the most high impact optimization one could try first. seems like HVM5 was wasting a lot of time garbage-collecting unused branches of pattern-match nodes. I had optimized that for static mats, but not for dynamic mats. skill issue. Fable figured how to do it for these, resulting in a massive speedup in some benches but wait, is that *correct*? I'm not sure yet, it is credible, but this is the kind of thing that is very easy to get wrong on interaction nets. the problem is, when I was ready to start auditing Fable's solution so I could tell whether it was buggy or legit, it interrupted me to tell me it had found a massive bug on the code *I* had written. ... wait, what? so... for garbage collection purposes, I stored a bit on lambda term pointers that meant "the variable bound by this lambda has been freed, so, its lambda must free whatever argument it is applied to". that's fine. yet, on duplicator nodes, I also used the same bit to mean "one of the duplicated variables was freed, so, treat this dup as a passthrough no-op". so, if a lambda entered a duplicator, it would mistake the lambda's collection bit for its own, resulting in corrupted interaction! that's a mouthful, why I'm writing this? just so you can appreciate the sheer absurdity of what just happened. I didn't ask it to find bugs. I asked it for an optimization. and even if I did ask it to find bugs, this bug is so astonishingly subtle and specific, identifying it takes mastering the domain to an extent that it beyond even me. I'd easily need hours or days to fix it, *if* I ever came across it. chances are it would just go unnoticed. and Fable found it and fixed it like it was nothing, while it was busy adding a 17x speedup to a file that neither I, nor Opus 4.8, nor a fleet of GPT 5.5 managed to barely make 2x faster. oh and there is also another tab where it is also ripping through Bend's codebase and finishing everything I had to do I don't know what to say anymore this isn't about Anthropic or OpenAI, this is about our collective future as a species. the world is changing, and we need to be aware of it, and discuss how to handle this change. receipt below . . .