If anyone wants to become a millionaire, know that there’s currently a MASSIVE opportunity in digital twin applications in the age of AI. Really high demand, low supply, really expensive solutions, and long development processes. Everything can be reworked with AI. I just had
@dfintelligence
-
LLM coding test: Mistral Large 4 vs Kimi K3 vs GLM 5.3
By
–
Mon nouveau petit test: fait une animation en js/html d'un homme qui dessine un homme qui dessine un chat.
— Defend Intelligence (Anis Ayari) (@DFintelligence) 6 octobre 2026
Mistral Large 4 vs Kimi K3 vs GLM 5.3.
Allez j'arrête ici pour l'instant, je vais faire une descente en rafting à Hawaii, bsx a+. pic.twitter.com/HB3jN75SAZMy new little test: make a js/html animation of a man drawing a man drawing a cat. Mistral Large 4 vs Kimi K3 vs GLM 5.3. Alright, I'll stop here for now, I'm going rafting in Hawaii, bsx see you later.
-
Comparative Test Mistral Large 4 vs Kimi K3 vs GLM 5.3
By
–
Deuxieme test Mistral Large 4 vs Kimi K3 vs GLM 5.3.
— Defend Intelligence (Anis Ayari) (@DFintelligence) 6 octobre 2026
Au moins ca pédale VITE, les roues pas dans le bon sens mais c'est un détail hein. pic.twitter.com/kSzYXTOoE0Second test Mistral Large 4 vs Kimi K3 vs GLM 5.3. At least it pedals FAST, the wheels are on the wrong way but that's a detail, huh.
-
Questioning ‘from scratch’ training of a 1T model on 3,800 GPUs
By
–
How did you manage to train a 1T model « from scratch » on 3,800 GPUs? What do you mean « from scratch » ?
-
Doubt about training a 1T model from scratch on 3800 GPUs
By
–
I'm very, very eager to get more info on this. Because as it stands, I don't see how it's possible to train "from scratch" a model with 1,000 billion parameters (so 46 billion active) on 3,800 GPUs. So I'd really like more details, but obviously,
-
Critique of new model versus Chinese models and Mistral benchmarks
By
–
Ok, so it's worse than the Chinese models released several months ago, and Mistral, which as usual is still fudging the benchmarks and how it presents them. Distillation wasn't enough. They're going to get blown away in 2 weeks. That's good,
-
New Mistral model expected this week or next
By
–
Just so you know, according to my FDI information, there should be a new Mistral model this week or next week. They are preparing the release. Looking forward to it.
-
Agentic AI autonomy on smartphones vs desktop coding agents
By
–
Hear me out: when we will be able to have the same level of agentic autonomy on a smartphone as we have with Claude Code or Codex on our computers, it’s gonna be crazy. It will be a new era for mobile. Siri should have done that , but Apple suck at agentic AI for decades. I’m
-
Change the base model of an LLM in one line
By
–
With any serious LLM library, it takes one line to change the core model.
-
@dfintelligence — 2026-06-27
By
–
ATTENTION !!!! Ce serait une très grave erreur de se priver des modèles open source chinois ultra-performants en France. Des réponses orientées ou biaisées, oui, il y en a. Oui, il y en aura. Il y en en a aussi chez Mistral AI, GPT, Anthropic, etc. La probabilité qu'un agent