Thanks to @GeminiApp 3, I'm again using google Tasks. Google did with gemini what Siri couldn't achieved. I now have a working agent using the apps I use correctly.
@he_the_mystery@godofprompt I had to take control several times, it struggle connecting the components. I was amazed it started to search for the proper logic of the components. I just test a bit, I think if I provide it with the docs it will act better
@perplexity_ai Comet interaction with google sheets will be incredible. I discarded creating a personal income/expenses app through CC. It'll be easier that this agent handles it.
Great minds like great challenges. Only when a “problem” is known/visible, labs will work on to solve it. There won’t be a “our model doesn’t perform well on x” and just leave that benchmark alone. They will solve X and recursively continue onto the next Y… Until AI to solve all
OpenAI’s GDPval made me think - Make visible the fundamental problem and it will be solved. Which goes along with formulating better questions (the “prompt engineering”).