7. Amazon doubles down on AI devices A refreshed Echo lineup, Fire TV with Vega OS, Kindle Scribe Colorsoft, and “Alexa+”—a proactive AI assistant with memory and multitask workflows.
MULTIMODAL AI
-
Black Forest Labs Kontext AI Models Edit and Generate Images
By
–
Black Forest Labs’ Kontext AI models can edit pics as well as generate them #AI #AIio #AIInnovation #ML #DataScience #Futureofwork @timnitgebru @oriolvinyalsml @ceobillionaire @soumithchintala @waitin4agi_ @sallyeaves @bernardmarr
-
Diverse prompts enhance model reasoning
By
–
this is also backed by work in Dipper (prompt ensembles) showing that feeding multiple diverse prompts in parallel improves reasoning and reduces blind spots. and the more you mix prompt styles – contrast, random pivoting, persona shifts – the more the model “sees” new
-
Hiring Researchers Obsessed with Pixel Generation
By
–
We are hiring researchers who are obsessed with generating pixels 🤩 https://t.co/797Dd6i9Zm
— Fei-Fei Li (@drfeifei) 3 octobre 2025We are hiring researchers who are obsessed with generating pixels
-
Google releases NanoBanana model for production via Gemini API
By
–
Here’s a quick look at what we shipped this week — @NanoBanana is now generally available and ready for production use for developers via the Gemini API on @GoogleAIStudio and @GoogleCloud Vertex AI for enterprise customers. — We also released 10 new aspect ratios and
-
AI Memory Evolution: Beyond Personalization to Co-Memory Systems
By
–
Talking about AI memory as just good personalization totally misses the bigger picture. It's actually becoming co-memory, remembering the world with you, saving your random ideas, even proactively reminding or resurfacing things. It's so much more than tailored answers
-
Alibaba reveals Qwen3 and Wan 2.5
By
–
BREAKING: Alibaba just unveiled Qwen3 and Wan 2.5 at Apsara 2025. We tested them with real builds and the results are wild. Here’s what Qwen3-Max, Qwen3-VL, Qwen3-Omni, and Wan 2.5 can actually do in the hands of a creator. @AlibabaGroup
-
Hume’s Octave 2: Best Instant Translation Voice Model
By
–
Hume's Octave 2 is the best instant translation voice model I've seen – demo is out of this world: https://
demo.hume.ai/evi-4 -
OpenAI Sora 2 Impresses With New Capabilities
By
–
Ça recommence à bouger très vite dans le domaine de l’IA (honnêtement, ça n’a jamais vraiment arrêté, mais certains disaient que ça “ralentissait”).
— VISION IA (@vision_ia) 2 octobre 2025
Et on dois vraiment reconnaître le mérite de OpenAI,
car Sora 2 est impressionnant sur de nombreux points :
– un son de haute… pic.twitter.com/I1O5xZUyG2Ça recommence à bouger très vite dans le domaine de l’IA (honnêtement, ça n’a jamais vraiment arrêté, mais certains disaient que ça “ralentissait”). Et on dois vraiment reconnaître le mérite de OpenAI,
car Sora 2 est impressionnant sur de nombreux points : – un son de haute