@andrewchen Memory and the context derived from it would make consumer AI apps highly sticky. Also, apps will be explicitly designed to make sharing context effortless for users.
The key question in the closed vs. open-weight debate is who funds frontier research and training, and how? Open-weight providers have not unlocked a research breakthrough that makes training dramatically more efficient in compute or data. #agi#openai#claude#kimi
@bindureddy Fable spins up hundreds of subagents and probably has a feedback loop, which could also be a reason why it does well. How would you replicate that?
Today’s coding agents and LLMs are brilliant.
The missing piece is not intelligence. It is judgment and the battle tested learnings that prevent disasters.
@Appyg99 Unless someone thinks through how to solve the cold start, it won’t be worth building.
Scale AI may be a chance. They have 1 million users with verified skills on the platform.