In partnership with
News of the day
1. Mistral launches Forge, a platform for enterprises to build custom AI models trained on their own data, addressing the common failure point of generic models lacking business understanding. → Read more
2. OpenAI launches GPT-5.4 mini and nano, drastically cutting costs for image description and AI analysis, making large-scale data processing affordable. → Read more
3. Pentagon plans secure environments for AI firms to train models on classified data, enhancing military AI but posing new security risks. → Read more
4. World ID's new Agent Kit uses iris-scan tech to link human identity to AI agents, preventing Sybil attacks and ensuring responsible AI use online. → Read more
Our take
Hi Dotikers!
Yesterday, we were tracking the rise of OpenClaw and how Nvidia was embracing the open-source agent ecosystem rather than watching from the sidelines. Today, Nvidia is still in the picture, but this time as a backdrop, and it's Mistral taking center stage.
The French startup announced Mistral Forge at Nvidia's GTC, the annual enterprise AI summit. The concept: enabling organizations to build their own models from scratch, trained on their internal data, not on the web's archives. This isn't a cosmetic upgrade. Most market players offer fine-tuning or RAG, two approaches that rely on an existing model adapted at the margins. Mistral claims to do something different: training a model end-to-end, using the same recipes the company employs for its own flagship models. It's the difference between repainting a car and building the engine.
The positioning is consistent with Mistral's trajectory since its inception. While OpenAI and Anthropic saturated the consumer market with B2C products, Mistral cultivated relationships with enterprises and governments. The result: Arthur Mensch announces a billion-dollar annual recurring revenue milestone for this year. That's no small figure for a European company navigating a market largely dominated by American players.
Forge's first partners illustrate the target well: the European Space Agency, ASML, Ericsson, and Singaporean government agencies. Sectors where data sovereignty isn't a sales pitch but a genuine operational constraint. Training a model on thirty years of internal documentation from a semiconductor manufacturer or a telecom operator is a use case that makes RAG clearly insufficient.
Skeptics are right, however, to raise the question of cost. Building a model from scratch remains out of reach for the vast majority of companies. Forge targets a narrow segment, those with the budget, the data, and the technical teams. For everyone else, fine-tuning remains more pragmatic. Mistral knows this and is hiring field engineers to support its clients, following a model inherited from Palantir: sell a product, then send people to install it. Consultant or software vendor? The answer is probably both, and that's precisely what makes Mistral's position hard to replicate.
Alex.
Here’s how I use Attio to run my day.
Attio is the AI CRM with conversational AI built directly into your workspace. Every morning, Ask Attio handles my prep:
Surfaces insights from calls and conversations across my entire CRM
Update records and create tasks without manual entry
Answers questions about deals, accounts, and customer signals that used to take hours to find
All in seconds. No searching, no switching tabs, no manual updates.
Ready to scale faster?
Meme of the day
