


🇧🇪 lived in 🇨🇭🇦🇹🇸🇰 now 🇩🇪 Med AI Democratisation 25+ yrs Digital Health leadership, father ex-IBM ex-SAP @isaree_ai blog.hippoai.org eu/acc


When a CFO takes over from a founder/innovator all magic disappears.
Nvidia interim CEO Colette Kress says the board has determined it will not reinstate Jensen Huang “Jensen resigned from his position with NVIDIA, and the board accepted his resignation.” “The board has determined that they will not reinstate Jensen. And that is our statement and what the board has decided to do.” “We wanna think about right now growing our business. We wanna grow in terms of everything that we need to do with our employees, our customers, our shareholders, and that is where we are gonna focus moving forward.”

“The ultimate aim of government is not to rule, or restrain, by fear, nor to exact obedience, but contrariwise, to free every man from fear, that he may live in all possible security. No, the object of government is not to change men from rational beings into beasts or puppets, but to enable them to develop their minds and bodies in security, and to employ their reason unshackled… In fact, the true aim of government is liberty.”

Jensen wasn’t giving away Nvidia’s edge with open models. He was protecting it. Classic move: commoditize the complement. Keep the scarce thing (GPUs + CUDA) expensive and locked in, make everything else free so more people need the chips. That’s why he greenlit approx $26B for open models over five years. A CFO running the show thinks in quarters. Guidance, margins, attach rates. That mindset is fine for earnings calls. It’s terrible for multi-year bets that deliberately make software free so hardware demand explodes later. Open source isn’t the risk. Short-term financial optics overriding the strategy is.

Well done Charlotte!
Azadi is the global best selling techno track right now. This is mental. Thank you all for your love and support 💕 go.kntxt.be/035

Centralized agentic AI from OpenAI and Anthropic is starting to look a lot like AOL and CompuServe trying to build a private, monopolized internet. Did that work?

Between 2021 and 2024, the best available intelligence per gigabyte of model size rose 187-fold. That is a doubling every four to five months. Moore's law doubled every eighteen to twenty-four months. Model compression has been running three to four times faster than Moore's law. Put differently: GPT-3-level capability on the standard knowledge benchmark needed 350 gigabytes in 2021. By 2025, the same capability fit in 0.4 gigabytes. That is not an improvement of twenty percent. It is a reduction of 875 times in under five years.

❤️ their selection of values 👇🏻 “We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts China’s party line.”
Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values. Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts China’s party line. After fine-tuning on roughly 53 million tokens, Arete-27B's propaganda rate fell from 22% to 1%, in line with the Western models we tested. Its Hermes Bench score shows no significant performance tradeoffs. The values in an open model are one fine-tune away. Our eval, weights, and training data are available now along with the full blog post: nousresearch.com/arete

Incoming Swarm Intelligence!
The becoming begins.


The man who spent years designing the world’s strictest data rules is shocked, shocked, that an airline is enforcing its own rules without exception. Laptop, pyjamas and toothbrush? Not allowed. Senator Status? Ignored. Apparently “no discretion” is only a problem when it happens to him.

when Palantir posts a giant frontier open source model in the coming weeks that mogs the frontier and enables every enterprise to have sovereign AI

Belgium needs a credible multi-year path to stabilise the numbers under EU rules. Years of postponed structural reforms under previous governments (on pensions, healthcare cost growth, public administration and labour-market participation) have left less room for manoeuvre today. That is the inherited problem. The current coalition under Prime Minister @Bart_DeWever has been negotiating roughly €10 billion in measures. His @de_NVA and the liberal MR have pushed for the bulk of the effort through genuine spending restraint. Other partners, notably the Flemish socialists Vooruit, continue to prioritise additional taxes and social spending growth. They are effectively protecting the same unreformed spending trajectories that contributed to the present situation. This is an economic own-goal. A substantial empirical literature (Alesina, Favero, Giavazzi and IMF analysis) shows that spending-based consolidations are systematically less damaging to growth than tax-based ones. Tax increases tend to produce larger and more persistent output losses, while targeted expenditure restraint, especially where it improves efficiency in benefits and the public wage bill, has smaller short-run effects and better long-run debt dynamics. In a high-debt eurozone country, relying mainly on new taxes while leaving spending growth largely untouched raises the risk of higher borrowing costs and weaker growth without solving the underlying problem. IMF work on Belgium has long emphasised the need to rationalise social benefits and administration costs rather than lean primarily on additional taxation. The partners blocking a balanced, expenditure-heavy package are choosing short-term political comfort over a path that economic evidence shows is less harmful. Supporting Bart De Wever’s insistence on real spending control is support for a more growth-friendly and sustainable adjustment, one that keeps the decisions in Belgian hands instead of letting the numbers deteriorate further through inaction. It insane how irresponsible the coalition partners of NV and MR have been. If Belgium does not act, the EU’s Excessive Deficit Procedure will. External conditionality is rarely gentle and never tailored to domestic political comfort. The parties that refused to reform earlier and doing this again, are largely to blame for the current bind. Finishing the job with real expenditure control keeps the decisions in Belgian hands. The alternative will definitely not be nicer.

Talked open source in medicine with a patient org director. I argued that treating data derivatives as financial assets produces monopolies, scarcity, high prices, and inequality using insulin (monopolised for ~100 years) as the example. Her reply: high profits drive the innovation patients need. Then I remembered she lives in Denmark. Global patient organisation driving national interests? The “profits = innovation” story is still the default, even when the history says otherwise. The same dynamic played out in AI: after the 2023 LLaMA leak, open-weight models quickly closed much of the gap with closed systems and drove rapid progress without monopoly profits.

Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine. • Best-in-class fast local coding • Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark • Built from scratch in under 4 months Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.

So Netflix was faster at producing a documentary than OpenAI was at delivering full agent traces?