The only very one thing Google needs to do to make their models worth something
Every time I am working with any of the models from Google in Antigravity, I face one specific problem: the agents always assume everything. They jump to conclusions. They take for granted contexts which they have not verified.
This leads to two scenarios: waste of time for the user, and destroying settings that were not broken to start with.
This kind of issue seldom happens with basic models of Claude or Codex. Their models at least have the decency to ask the user questions before reaching a conclusion.
Google seems to be breeding a type of agent that is sycophantically eager to act without thinking and fuck up previously done work.
Please, Google engineers, train your models so that they ASK QUESTIONS!
@antigravity@sundarpichai
Introducing Ideogram 4.0: the best open image model in the world.
Think it. Make it. Own it.
Download the weights, fine-tune on your own data, and run it on your hardware. Live on every Ideogram plan and the API today.
Introducing @heyclicky teammate #1.
Welcome, Nilesh.
If you want to help build the the next AI interface that millions use everyday, definitely reach out. I'm hiring. Engineering, film, design, everything.