informal007

Google is putting much higher priority on frontier models over Flash models, even though they have TPUs.

| Quarter | Revenue (USD billions) | QoQ Growth |

| Q1 2026 | 4.731 | +131.9% |

| Q2 2026 | 11.555 | +144.1% |

| Q3 2026 | 16.800 | +45.4% |

The growth rate slowed in the latest quarter, which I think is why they decided to go public now.

It won't be free for Apple to use Meta's model in the future.

It’s pretty slow for me. I’m not sure if it’s due to I trigger an agentic task by Gemini website entrance.

Google have strict limits on new model this time and are pretty conservative. As we know they spent lots of time on internal testing since release.

All the signs mean Gemini 4 argon is a pretty large model than previous.

Even on Hacker News, comments that briing unrelated but valid points often get more attention

I'm thinking if this is a more fair comparison among other models, like ChatGPT, Claude and Deepseek

Much like short-form video, technology that minimizes deep thinking seems bound to gain popularity

A dump question, can we use Internet Archive for that?

Interesting! Thanks for sharing

One difference for human to understand the video is that we only care the changes on a picture compare to LLM

it shows the possibility that SVG replace PNG/JPG even video.

This help me understand meaning of Pi as open source ai coding agent

Tried today, I like the record track feature