They just added function calling support to the new DeepSeek. Before you had to use all sorts of tricks, of course I know them. But now it's supported!
GENERATIVE AI
-

Kontext Rebuilds Scenes Accurately Without Creative Hallucination
By
–
I like how kontext rebuilds the scene as prompted, rather than a creative upscaler hallucinating the details.
-
Google Brings Veo 3 Video AI to Mobile and Web
By
–
We're bringing Veo 3 across the pond and to mobile!
— Google AI (@GoogleAI) 31 mai 2025
For Pro and Ultra members, you can now access Veo 3 through the @GeminiApp on Android and iOS and through https://t.co/HUIK0owzfZ in over 71 countries (including the UK). https://t.co/RPyEn6ZwHNWe're bringing Veo 3 across the pond and to mobile! For Pro and Ultra members, you can now access Veo 3 through the @GeminiApp on Android and iOS and through http://
flow.google in over 71 countries (including the UK). -
Veo 3 Video Generation Demand Surges Across Mobile Platforms
By
–
The demand for Veo 3 has been off the charts! Millions of videos have been generated in the past few days alone, now available on mobile, and in more countries (including the UK! 🇬🇧). Try it out in the @GeminiApp (and the Flow tool for filmmakers at https://t.co/QgTpxTKAOi) https://t.co/jpWfqyKB9U
— Demis Hassabis (@demishassabis) 30 mai 2025The demand for Veo 3 has been off the charts! Millions of videos have been generated in the past few days alone, now available on mobile, and in more countries (including the UK! ). Try it out in the @GeminiApp (and the Flow tool for filmmakers at http://
flow.google) -

DeepSeek-R1 v2 Engineer Release with Function Calling
By
–
The new version of DeepSeek-R1, it's an INSANE model, my god.
— Pietro Schirano (@skirano) 30 mai 2025
DeepSeek Engineer v2 with function calling support drops tomorrow.
I have to. pic.twitter.com/9rkx4roKc7The new version of DeepSeek-R1, it's an INSANE model, my god. DeepSeek Engineer v2 with function calling support drops tomorrow. I have to.
-
Ambient AI Agents: Production Challenges and Email Revenue Automation
By
–
Harrison Chase & Deepak Bapat (Tabs CTO) explore ambient vs background AI agents, vertical AI advantages, and the technical challenges of production systems – plus Tabs' approach to building ambient agents for email monitoring and revenue automation without chat interfaces.… pic.twitter.com/bzPJfF6U1a
— LangChain (@LangChain) 30 mai 2025Harrison Chase & Deepak Bapat (Tabs CTO) explore ambient vs background AI agents, vertical AI advantages, and the technical challenges of production systems – plus Tabs' approach to building ambient agents for email monitoring and revenue automation without chat interfaces.
-
Ollama adds thinking traces feature
By
–
ollama now supports thinking traces 🤖
— 🚨 AI News | TestingCatalog (@testingcatalog) 30 mai 2025
This feature can be tested with the newest DeepSeek-R1-0528 model https://t.co/waFbJjcrT6 pic.twitter.com/Yutj5v40GJollama now supports thinking traces This feature can be tested with the newest DeepSeek-R1-0528 model
-

Anthropic Leads SimpleBench with 58.8%
By
–
Anthropic takes the lead on SimpleBench again. 27.5% (Sonnet 3.5 from June) to 58.8% in one year is a big jump. Next year will be wild
-

Testing Kontext Pro Portrait Angle Variations
By
–
Me trying to prompt all the different portrait angles with kontext pro
-
Flux Kontext Pro Multi-Image Model Concatenation Explained
By
–
On the vanilla model you can only pass in one image: https://
replicate.com/black-forest-l
abs/flux-kontext-pro
… What this model does behind the scenes is concatenates two images then sends them as a single image: https://
replicate.com/flux-kontext-a
pps/multi-image-kontext
… So the multi-image model we put up yesterday does the hard work for you.
