The CUA model was primarily trained on web tasks, but early users have also found it surprisingly good at desktop applications. It’s still early days, though—we have plenty more to improve!
MULTIMODAL AI
-
AI Workflow: Claude Website to Creative Output
By
–
Now this is sick
— Linus ✦ Ekenstam (@LinusEkenstam) 11 mars 2025
Claude 3.7 –> Website
Screen Recording –> Grab first frame
Magnific Sturctured Ref –> into crayons
Runway Restyle –> Final output
Pure magic, AI was used in all steps,
I just moved files around and instruct. pic.twitter.com/DJtGFZ4EKONow this is sick Claude 3.7 –> Website
Screen Recording –> Grab first frame
Magnific Sturctured Ref –> into crayons Runway Restyle –> Final output Pure magic, AI was used in all steps, I just moved files around and instruct. -
Impact of faster AI text-to-speech for creators and businesses
By
–
That's huge! Faster TTS means more possibilities for creators and businesses.
-
ChatGPT compares AI-generated images
By
–
One is clearly made by DALL-E 3 The other one is 4o? ChatGPT is getting image response comparison: “You are giving feedback on a new version of ChatGPT”. Never saw this working for image responses before.
-
Sony AI Prototype Enables Real-Time NPC Conversations in Games
By
–
Sony revealed a new prototype of an AI-powered video game character for Playstation’s Horizon Forbidden West
— Rowan Cheung (@rowancheung) 11 mars 2025
We're heading towards a future where NPCs will be capable of real-time conversations with players—thanks to the power of AIpic.twitter.com/5a8IoIXG4OSony revealed a new prototype of an AI-powered video game character for Playstation’s Horizon Forbidden West We're heading towards a future where NPCs will be capable of real-time conversations with players—thanks to the power of AI
-
Pika AI Virtual Clothing Try-On Without Shopping Stress
By
–
Try on every option for every dress code…without needing to stress, shop, or show up 2.5 hours late.
— Pika (@pika_labs) 10 mars 2025
Give it a whirl at Pika dot art, or on the iOS app#Pikaswaps
Credits @glamseasons pic.twitter.com/8kIBZxUS3ATry on every option for every dress code…without needing to stress, shop, or show up 2.5 hours late. Give it a whirl at Pika dot art, or on the iOS app #Pikaswaps Credits @glamseasons
-
Copilot Launches Voice, Reasoning, and Mac App Updates
By
–
Busy last few weeks + more coming…
-free unlimited voice + reasoning
-Think Deeper runs on o3-mini high now
-voice speaks 40+ languages
-Mac app in a dock near you
-gaming coming to Copilot Labs
Tip of the iceberg. Can't wait to show you what else the team is working on -
FLUX Dominates Image Generation Market with 40 Percent Share
By
–
In image generation, Black Forest Labs’ FLUX family of models burst onto the scene in mid-2024 and has maintained its dominant position as the clear frontrunner since, capturing close to 40% of image gen messages, but with Google’s Imagen3 also rising fast. (4/6)
-
OpenAI Prepares Omnimodel Launch with Full Multimodal Support
By
–
Omnimodel Launch in ChatGPT Soon? References to internal codenames for video and audio handling have been found, suggesting that OpenAI is preparing to roll out full multimodal support in ChatGPT.
This aligns with the Omnimodel capabilities that were showcased only as demos -
Google AI Studio may add Gemini embedding model
By
–
Google AI Studio may get support for the recently announced "gemini-embedding-exp-03-07" model. This model, as well as "imagen-3.0-generate-002", were added to the hidden model's list. * It is still possible that these options will never become visible to end users