Journalist 🗞️ Historian 📚 Writer ✏️ European 🌍 Austrian 🎶 Styrian 🌳 Jedi ☄️ Eldar ✨ Schwoaza ⚽ Dad 👶
Wien, Österreich
Joined June 2010
- Tweets95.7K
- Following2K
- Followers4.4K
- Likes2K
Pinned Tweet
Hear ye, hear ye 🥁🙌
So happy and proud to announce that we won the Jean Monnet Prize for European Integration 2020 for our project “Home Is Where the Herz Is”
europeanconstitution.eu/2020…
Check it out ☺️❤️
metropole.at/herz
So proud of our team at @metropole_vie 🤩
Benjamin Wolf 🇺🇦 retweeted
Time to break the narrative
Europe is and has always been a catalyst of some of the best AI researchers in the world
(And yes know many, if not most, of these have ended up at US big tech or US frontier labs)
Looking at the impact Europeans have had on AI, we should be perceived as a AI powerhouse
Deepmind is a prime example
Europe has a deep history of research excellence, a high density of technical talent, and world-leading institutions -
European computer scientists have been central to many of the world's most important innovations, such as Linux and backpropagation
European researchers have played central roles in breakthrough innovations such as CRISPR, mRNA vaccines, and even the invention of the World Wide Web
Many, if not most, of today's leading researchers in AI and LLMs also originate from Europe
Time to break the narrative
And for Europe to break up with pessimism
Benjamin Wolf 🇺🇦 retweeted
easy to bash but seems harder to recognize mistral is actually delivering 💪🏻
Impressed by the new @MistralAI large-4 update! The model update is much better at coding and performs as the 2nd best OSS model on our coding benchmark
Benjamin Wolf 🇺🇦 retweeted
This week's @thecurrency column looks at trends in household incomes and equality in Ireland and Europe.
thecurrency.news/articles/24…
And it should be called an alliance. Clearer and more honest that associate membership, which would entail other things.
Benjamin Wolf 🇺🇦 retweeted
It's called an alliance. They are talking about an EU-Canada alliance. But it just freaks out member states for the EU to use the A-word so new faux terminology is needed to say the same thing.
Benjamin Wolf 🇺🇦 retweeted
y a encore quelques semaines, Mistral faisait essentiellement de l'inférence pour GLM 5.3
Mistral 4 est désormais plébiscité.
Congrats @MistralAI
Today, we release the weights of a new and improved Le Chonk, advancing the open frontier for agentic coding and cybersecurity.
💻 Built to build: Le Chonk ranks #1 on @cognition’s FrontierCode among open models, and sets a new benchmark for Pareto efficiency on @artificialanalysis.
🛡 Built to defend: State-of-the-art in cyberdefense, put through the paces by the world's leading cyber vendors, including @CrowdStrike, @cloudflare, @rubrik, @zscaler, and others.
Built for sovereignty: Make it yours. Download the weights on @huggingface , customize it with Mistral Forge, and deploy it wherever you need it.
Read more from our team and customers 👇:
Czechia needs to build more, plain and simple.
And Czechs need to spend more of their income.
Right, @CrisisStudent & @DanielKral1?
Where did productivity grow in Czechia and Poland since 2019?
A lot of attention has been paid to the stagnation of Czech hourly labour productivity since covid (+1% by H1 2026), often compared with Poland’s rapid growth (+17%). But the sectoral breakdown is very interesting.
Industries are grouped using the official OECD and Eurostat classifications: goods-producing industries by R&D spending relative to value added, services by the share of university graduates (knowledge-intensive above 33%). Whole economy data to H1 2026; sectors to 2025, Polish goods-producing industries to 2024 (latest available).
Each country is measured by its own numbers, so the same industry can land in different groups (e.g. electrical equipment is high & medium-high tech in Czechia but medium-tech in Poland).
Almost mirror images:
High & medium-high tech manufacturing: 🇨🇿 +26% vs 🇵🇱 −18%
Medium-tech manufacturing: 🇨🇿 +3% vs 🇵🇱 +19%
Low & medium-low tech industries: 🇨🇿 −8% vs 🇵🇱 +5%
Knowledge-intensive services: 🇨🇿 +5% vs 🇵🇱 +2%
Less knowledge-intensive services: 🇨🇿 −3% vs 🇵🇱 +20%
Whole economy: 🇨🇿 +1% vs 🇵🇱 +17%
🇨🇿 Czechia
High & medium-high tech manufacturing +26%, led by cars, electrical equipment and electronics, but only 8% of the economy.
Knowledge-intensive services +5%: ICT +21%, finance +31%, but real estate (−28%) dragged the group down. It only just made the cut (33.6% graduates); without it, it would be +20%.
The drag comes from domestic-demand sectors and energy-hit heavy industry (steel −52%, energy −30%). After the 2022–23 inflation shock, households cut spending and hospitality output fell 39% with the same hours, construction output −11% with hours +9%.
🇵🇱 Poland
Growth came from big, broad sectors: retail & wholesale +31%, metal products +29%, agriculture +40%.
High & medium-high tech manufacturing −18%, but only 2% of the economy and almost entirely cars, where the official numbers jump up and down suspiciously (−22% in 2023 despite rising production). A big rebound in 2025 can’t be ruled out.
Czech productivity growth follows sophistication almost perfectly, and its high & medium-high tech manufacturing and knowledge-intensive services grew in real terms about three times faster than the economy. But this core is only about 20% of value added, and workers are not moving into it from the low-productivity sectors.
Poland is the clear winner overall: growth in large, “ordinary” sectors lifts the whole economy. Part of the Czech weakness is cyclical, though: Czech households save 19% of their income, the second highest rate in the EU (Poland: 9%). More spending would let demand-hit sectors raise output without adding staff.
Benjamin Wolf 🇺🇦 retweeted
Where did productivity grow in Czechia and Poland since 2019?
A lot of attention has been paid to the stagnation of Czech hourly labour productivity since covid (+1% by H1 2026), often compared with Poland’s rapid growth (+17%). But the sectoral breakdown is very interesting.
Industries are grouped using the official OECD and Eurostat classifications: goods-producing industries by R&D spending relative to value added, services by the share of university graduates (knowledge-intensive above 33%). Whole economy data to H1 2026; sectors to 2025, Polish goods-producing industries to 2024 (latest available).
Each country is measured by its own numbers, so the same industry can land in different groups (e.g. electrical equipment is high & medium-high tech in Czechia but medium-tech in Poland).
Almost mirror images:
High & medium-high tech manufacturing: 🇨🇿 +26% vs 🇵🇱 −18%
Medium-tech manufacturing: 🇨🇿 +3% vs 🇵🇱 +19%
Low & medium-low tech industries: 🇨🇿 −8% vs 🇵🇱 +5%
Knowledge-intensive services: 🇨🇿 +5% vs 🇵🇱 +2%
Less knowledge-intensive services: 🇨🇿 −3% vs 🇵🇱 +20%
Whole economy: 🇨🇿 +1% vs 🇵🇱 +17%
🇨🇿 Czechia
High & medium-high tech manufacturing +26%, led by cars, electrical equipment and electronics, but only 8% of the economy.
Knowledge-intensive services +5%: ICT +21%, finance +31%, but real estate (−28%) dragged the group down. It only just made the cut (33.6% graduates); without it, it would be +20%.
The drag comes from domestic-demand sectors and energy-hit heavy industry (steel −52%, energy −30%). After the 2022–23 inflation shock, households cut spending and hospitality output fell 39% with the same hours, construction output −11% with hours +9%.
🇵🇱 Poland
Growth came from big, broad sectors: retail & wholesale +31%, metal products +29%, agriculture +40%.
High & medium-high tech manufacturing −18%, but only 2% of the economy and almost entirely cars, where the official numbers jump up and down suspiciously (−22% in 2023 despite rising production). A big rebound in 2025 can’t be ruled out.
Czech productivity growth follows sophistication almost perfectly, and its high & medium-high tech manufacturing and knowledge-intensive services grew in real terms about three times faster than the economy. But this core is only about 20% of value added, and workers are not moving into it from the low-productivity sectors.
Poland is the clear winner overall: growth in large, “ordinary” sectors lifts the whole economy. Part of the Czech weakness is cyclical, though: Czech households save 19% of their income, the second highest rate in the EU (Poland: 9%). More spending would let demand-hit sectors raise output without adding staff.
Benjamin Wolf 🇺🇦 retweeted
🇫🇷 Et si Mistral était en train de devenir beaucoup plus sérieux qu’on ne l’imagine ?
Hier, l’entreprise française a publié les poids de Mistral Large 4, surnommé « Le Chonk », après trois semaines supplémentaires d’entraînement. Et les progrès annoncés sont franchement intéressants !
On parle d’un modèle de 1 000 milliards de paramètres, dont 52 milliards actifs, avec une fenêtre de contexte d’un million de tokens.
Sur Terminal-Bench 4.0, ses performances sont passées de 28 % à 41 %. Sur l’indice de programmation agentique d’Artificial Analysis, Mistral annonce également des résultats comparables à GLM-5.3, tout en consommant 47 % de tokens en moins par tâche.
Mais il y a quelque chose que je trouve encore plus important : les poids du modèle sont désormais accessibles publiquement.
Cela signifie que des entreprises, des développeurs ou des institutions peuvent l’héberger sur leurs propres infrastructures, sans dépendre systématiquement d’une API américaine ou chinoise.
Et ce n’est pas simplement théorique. Orange Cyberdefense vient également d’annoncer un partenariat avec Mistral pour utiliser une variante spécialisée de Large 4 dans la détection de vulnérabilités et l’analyse des cyberattaques.
Je suis souvent très critique sur notre retard européen dans l’IA, mais je trouve qu’il faut aussi reconnaître les avancées lorsqu’elles sont là.
Mistral n’a pas encore démontré qu’il pouvait rivaliser avec les meilleurs modèles propriétaires sur tous les usages, mais il commence à construire quelque chose qui pourrait devenir extrêmement important pour notre indépendance technologique.
Et franchement, ça fait plaisir de voir une entreprise française se battre sur ce terrain !
Vous pensez que Mistral peut réellement devenir une alternative crédible aux géants américains et chinois ?
Benjamin Wolf 🇺🇦 retweeted
📈 Mistral AI monte en puissance !
🤖 Mistral AI a publié les poids de Mistral Large 4 (ou Chonk), un modèle massif de 1 000 milliards de paramètres et disponible sur Hugging Face.
💻 Cette intelligence artificielle s'impose comme un leader mondial du code agentique, rivalisant directement avec les meilleurs modèles chinois comme GLM-5.3 sur des bancs d'essai exigeants.
📊 Le modèle se distingue par sa grande sobriété, ne sollicitant que 52 milliards de paramètres par requête et consommant nettement moins de tokens que ses concurrents.
🛡️ Sa maîtrise de la cyberdéfense et sa capacité à détecter des failles de sécurité lui valent d'être adopté par des acteurs majeurs comme Microsoft, l'ENISA et Orange Cyberdefense.
🛰️ Des institutions scientifiques comme l'Agence spatiale européenne l'utilisent déjà pour exploiter des données complexes ou des images satellite.
Benjamin Wolf 🇺🇦 retweeted
The round after that is on its way. New generation of chips from Nvidia will mean about 2x to 3x as much training compute output for the same input electricity. I'm not sure we know how much bigger the new Mistral training set up will be, but they've already raised for it.
Benjamin Wolf 🇺🇦 retweeted
Mistral Large 4 is an excellent model. It's come through the later stages of training with the expected improvements in quality, intelligence, and efficiency. It will almost certainly produce an excellent range of smaller models. What's most interesting to me is how cheap it is.
Benjamin Wolf 🇺🇦 retweeted
This is a key point - the ECB's underlying thinking is steeped in supply pessimism. This is a defeatist framework that never gives a chance to genuinely positive reflationary pressures to manifest themselves and in fact risks entrenching various forms of hysteresis
Benjamin Wolf 🇺🇦 retweeted
3/ Importantly, using SPF expectations for GDP and HICP in the Orphanides-Wieland Taylor rule, which has a decent track record for the ECB, it's not all that obvious that the ECB should hike again in December. Unless you believe euro area potential growth is no higher than 1%.
Benjamin Wolf 🇺🇦 retweeted
Wages and salaries, adjusted for inflation, were down 0.6 percent in September from a year earlier.
Lots of talk about the "divide" between sentiment and data, but it doesn't seem hard to understand why people are unhappy when their pay is declining in real terms.
Benjamin Wolf 🇺🇦 retweeted
At this point, data and sentiment are aligned: the average worker is worse off today than a year ago.
And unlike the wealthy, most workers can't count on an AI-fueled stock market boom to take the sting out of a shrinking paycheck.
Benjamin Wolf 🇺🇦 retweeted
Euro area #GDP up by 0.4% in Q3 2026, +1.3% compared with Q3 2025: preliminary flash estimate from #Eurostat ec.europa.eu/eurostat/en/web…
ALT GDP growth rates over the previous quarter - % change, based on seasonally adjusted data
Benjamin Wolf 🇺🇦 retweeted
Fact: ALL the metals and minerals the energy transition needs between now and 2050 weigh less than the coal the world mines in ONE single year
Benjamin Wolf 🇺🇦 retweeted
A few thoughts on Lagarde resigning prematurely:
First, it's the wise thing to do with Le Pen (and other authoritarians) possibly looming. It means that the ECB Executive Board will be essentially stable until 2034/35. Why?
I will be leaving the @ECB in February 2027 because I want to have the freedom of speech, the freedom of expression, and the freedom to engage on matters that are not monetary policy matters.