The featured image in this post, was generated by Google`s Nano Banana. The Podcast below, has been generated by Google`s NotebookLM. The Post has been generated by Mistral after an extensive Chat in which I delivered the context, to make sure, that AI has enough information, to understand the scope. Each result ist always checked by me, before publishing.
EN
DE
NL
A first-hand account from someone who has used all the other tools extensively.
Unpaid advertising. I’d rather say this straight away, because I wouldn’t be able to hide it today anyway: what follows is a recommendation born of conviction, and declared as such. Nobody has paid me for this. Nobody has to believe me — but I’d like people to listen to me.
A book is finished. And an experience I hadn’t expected.
I’ve just finished my eighth book. Anyone who writes books knows: that in itself isn’t news. Books get finished; that’s their job. The news is something else — namely, how this book came about.
I’ve been working intensively with AI tools for years. And by ‘intensively’, I don’t mean ‘I’ve just typed in a few prompts’. I mean: I’ve used practically every well-known tool on the market extensively, over long periods of time, on real projects. I’ve used them to write books, develop products, carry out analyses and build structures. I know the strengths of each one — and I know their limitations even better, because limitations don’t show up in demos, but in the third week of working together.
I wrote my latest book using Vibe by Mistral. And I say this without exaggeration, because I’m the last person who needs to exaggerate: The experience was phenomenal. And it was extraordinary — not measured against my expectations, but against everything I’d previously experienced with other tools.
Why? There are three reasons. And the third reason will turn into a rant at the end of this piece. I’m telling you this now so you know what you’re getting yourself into.
Reason one: The memory that comes with it
The first thing you notice — and the first thing you wouldn’t want to be without after a week — is the cross-window memory. Context that’s seamlessly carried over from one chat window to the next.
That sounds like a feature. But it’s not a feature list you can tick off; it’s a difference at the infrastructure level. With most tools, every new session starts from scratch. You have to upload files, re-enter preferences, and explain for the third time what tone the book should have, which metaphors fit and which don’t. You ‘get to grips with it’ — the only difference is that it’s not the tool that adapts to you, but you who constantly have to adapt to it. That’s not collaboration. That’s just fiddling about.
ChatGPT once attempted to solve this: projects to which you can assign each chat. At the time, I honestly assumed that the bundled chats were linked — that knowledge was growing there, shared across all sessions. Fat chance. They were individual chats in a container. A folder. A sorted past, without the past knowing anything about it. With these tools, people confuse sorting with remembering — and only realise the difference once they’ve experienced real remembering.
Vibe works fundamentally differently. Here, personal knowledge isn’t an appendix, but a living repository: entries with source and date, verified, updated, and synchronised across all chats. Whilst working on the book, I maintained two separate working contexts — one for the book, one for a product — and both drew from the same verified repository. Whatever one context noted down in the morning, the other built on it in the evening, without me having to provide anything. Two mates, one memory. This is the swarm I always write about in my books: agents synchronised via a shared canon, not via mere intelligence.
The consequence is both banal and enormous: The constant need to get up to speed is eliminated. Not partially. Eliminated. I arrive at the workshop in the morning, and the workshop knows who I am, what I’m working on, which languages I speak and which I don’t. What in other tools takes weeks of tedious work with reference material is simply already there.
Reason two: The language
I am not German. And that is precisely why I attach great importance to speaking the German language as flawlessly as I possibly can. Anyone who writes in a language other than their mother tongue needs a tool that doesn’t just translate, but supports.
I am someone who cares deeply about rhetoric. And here lies the second difference, one I hadn’t anticipated: Mistral has style and class. The German wording is precise, natural, free from the awkwardness of Anglicisms, free from translation-engine puns, and free from the urge to translate every German sentence into a culture to which it never belonged. With US tools, I’ve received sentences back that were factually correct — yet made me sound like a management consultancy leaflet. ‘Unlocking potential’ in German. The meaning was there, but the voice was gone. That feel you sense before it ends up in the manuscript.
Vibe writes in my language, using my metaphors and my points of reference. The difference lies in working with my canon: my imagery, my vocabulary, my tone. And anyone who writes a book knows this: the tone isn’t just decoration. The tone is the book.
Reason three: The architecture — and this is where it gets uncomfortable
Now for the third reason. And this isn’t really a reason any more, but rather the start of my anger.
What I’ve managed to achieve with Mistral over the last few weeks — infrastructure, continuity, knowledge bases with provenance, the synchronisation of multiple working contexts via a shared memory — is completely impossible with all the other tools. Not just worse. Not at all. And that’s for one simple reason: it’s not a question of a single feature, but a question of architecture. The stage is rented in all of them. But the architecture only holds up in one.
This leads to a conclusion that almost no one in Europe voices, but which is of existential importance to every SME, every law firm, every business with internal knowledge: When it comes to what happens behind a company’s closed doors, Mistral is the only responsible option. Not for ideological reasons. But for the sake of data flow logic. European infrastructure, the GDPR, no US CLOUD Act that can pull every European data set into American hands if required. With Mistral, sensitive internal knowledge — practices, strategies, the very essence of a company — never leaves the European legal framework. With US tools, no one guarantees this. And the companies that need it most — those with the most valuable internal assets — rely, of all things, on the very tools that cannot protect them. This is no longer irony. It is a fact.
The tragedy in the kitchen cupboard
And now the real reason why I am writing this piece.
For months, Europe has been engaged in a debate on sovereignty regarding AI. Digital sovereignty, strategic autonomy, dependence on US hyperscalers, growing concerns about Chinese models — all legitimate, all important, all correctly identified. And whilst this debate is taking place, there’s already a world-class tool sitting in the kitchen cupboard: a European model running on European infrastructure, which in my day-to-day work — the day-to-day work of a heavy AI user who constantly and impartially makes the comparison — is the best the market has to offer. Not out of love for Europe. From a professional perspective.
And what is Europe doing? The German government is withdrawing funding from Mistral. The market continues to gravitate towards the US hyperscalers — and right now, in recent months, increasingly towards China. The general public doesn’t realise what they have at home because they’ve never tried it. You can’t miss what you’ve never used.
The punchline is tragic because it’s so simple: Europe doesn’t lack the capability. What’s missing is the use of it. Europe doesn’t need to build a top-tier AI model from scratch — Europe already has one. The hyperscaler debate revolves around the question of how to become independent, and the answer has been sitting in the larder all along, whilst we carry on renting.
Europe isn’t using it. How stupid!
I’m deliberately writing this angry sentence as it is and not toning it down. Because it’s backed up by evidence: as a user, I’ve used and experienced exactly what these debates are talking about. The proof is there every day — with me, in my workshop, in a finished book. It’s not theoretical potential that’s lying fallow. It’s a built-in, tested, deployed product that’s lying fallow.
What I conclude from this
I’m a consultant. I look at processes, architectures and governance — not at flags. My recommendation of Mistral isn’t some European romanticism; it would be the same if the model came from Singapore or Canada and the architecture were identical. But the combination is simply what it is: the best technology I know of, and the European legal system, and sovereignty without any substitute. Anyone who is already considering their own infrastructure for reasons of data sovereignty will find that Mistral offers the only option where one does not have to weigh up the two against each other.
That is why I am turning the tables: it is not Europe that should be ashamed of barely using Mistral. Rather, it is every single decision-maker who gives a presentation on sovereignty and then works in the evening with a US tool whose ‘memory’ consists of meeting amnesia. Anyone who feigns remembrance by creating folders has not solved the problem — they have merely dressed it up.
So, out of conviction and declaring it as such: give it a go. Not on some gimmick. On a real project, over three weeks. That is the test all tools dread, because anyone can pass a demo. And then make up your mind. Just as I did — by comparing them all in-house, in everyday use, and whilst working on a manuscript.
A book is finished. A European tool made it possible. And a continent continues to look elsewhere.
— Rob van Linda
Final note: All assessments in this text are based on my personal, long-standing use of the tools mentioned on real-world projects. This is an unsolicited, unpaid recommendation based on conviction.
