cat | cllm translate to french > docs_french.md what's next?

@MaziyarPanahi Every citizen is required to contribute at least 10 training tokens to each new frontier model

Asked ChatGPT to compare and contrast the coconut (reasoning in latent space with reasoning tokens & passing hidden states) to the new recurrent depth paper. Coolest was GPT saying a hybrid approach; I had been thinking before I even got to the bottom

@tyhouch Of course, maybe I'm wrong and Claude would be more serious about it that way. Models behave oddly :D the "or kittens die" thing is ~unneeded now both in my experience and re: latent space pod and that prompting guy podcast.

God, I can't wait until the pro sub includes getting a R/T voice agent to call people to ask questions for me.

This is good advice. I am hiring people right now. If you are expecting to do more engineering and less research, brush up on distributed systems, get grounded in infosec, graph theory. Economics and causal modeling. And statistics++++ because data data data.

@MaziyarPanahi 💯 And even in "real" investments at real scale there can be a lot of distortions. (covenants, liquidation preferences, warrants, etc) that make valuations deceptive.

@tyhouch Another trick is to introduce the contradictory opinion as belonging to someone else, “someone told me” or “ someone said”, and then take a skeptical tone and ask it to analyze this other person‘s theory. Also works well if you want two different models to debate. 😂

@MaziyarPanahi What if OpenAI was ahead of the game, and they don't think (completely) in words at all (re: COCONUT paper)?

If god did not exist, it would be necessary to invent him. --Voltaire, Chief AI Scientist of the Enlightenment

@GregKamradt Totally. And I agree human language would get in the way of thinking efficiently as we see thinking happen now (eg, generating tokens).

@GregKamradt Will it? I agree that first principles are interesting but what about interaction with the world?

You used to buy visual paradigm and spend a good amount of time to do this. Now? 10s mid-brainstorming. #ageofwonders

@KevinTamura What if your app just did everything? Because between the agents in the app and the agents building the app, it was immediately adaptive?

@chris_j_paxton Super interested in the details. I thought this was better for OpenAI than for figure. (Maybe @adcock_brett did too hah)