Farza Ships Real Time On Screen Graphics Powered by Sol 6.1 Ultrafast, With No Chat Window
Today, we're shipping real-time graphics. Using Sol 6.1 Ultrafast we can generate custom overlays, diagrams, and even interactive 3D scenes right on a user's screen, no chat. It's a new visual interface for AI. Here's me learning about LLMs + hardware. Demo:
Context
It shows fast models making a new kind of interface practical: instead of a chat reply, the AI draws overlays, diagrams and 3D scenes directly over what you are looking at.
Farza is the founder behind Clicky, an AI assistant for Mac and Windows that can see your screen and talk you through apps. Sol 6.1 Ultrafast refers to OpenAI's Ultrafast mode for GPT 6.1 Sol, launched on October 8, which OpenAI says runs up to eight times faster than the standard mode. That speed is what makes live, on screen generation like this feasible.
Deep dive
Farza announced on Wednesday night that his team is shipping real time graphics. According to the post, the feature uses Sol 6.1 Ultrafast to generate custom overlays, diagrams and even interactive 3D scenes right on a user's screen, without a chat window. Farza called it a new visual interface for AI.
The demo video, about two and a half minutes long, shows Farza learning about large language models and hardware, with the graphics appearing over his screen as he goes. In a follow up post, Farza said the feature is available now for anyone to try at heyclicky.com, the site for his Clicky desktop assistant.
The speed claim rests on OpenAI's Ultrafast mode for GPT 6.1 Sol, which OpenAI introduced on October 8. OpenAI says the mode delivers near Astra intelligence at up to eight times the speed of Sol Standard, priced at 12 dollars per million input tokens and 60 dollars per million output tokens.
The post did not share latency figures, pricing for Clicky users or details on how the 3D scenes are built. It was one of the most bookmarked AI posts of the day, with more than 4,000 saves.
What to expect next
The test is whether on screen graphics prove more useful than text for learning and getting work done, or remain a striking demo. Watch for latency numbers from real users, for how much each session costs to run on a premium fast model, and for whether larger assistants copy the idea. If fast modes keep getting cheaper, generated visual interfaces could become a standard way for AI tools to explain things.