Every company, should learn from cursor, create a harness w strong closed models, then do RL on that data on top of a Chinese model, and try to beat the original model. Soon, maybe composer will overturn a few from tier coding agents, if thereβs enough data + good algorithms.
You can build you own composer w your own data then.
What happens to open Ai and Anthropic then?
@utk7arsh yo good one! DM;ing you.
By the way- we're working on tactile sensors, for fleet deployment to capture human embodiment sensory data - https://t.co/5dWKd6v4UK
These objects are only tool-shaped to the wielders who cannot actually have their own thinking tools to hold, wield, and shape the tool-shaped object and actually make it a tool. Don't listen to them.
@signulll Would you like record yourself hoping to train a robot policy on your physical action data, pair that w your llm agents, train a model on your digital + physical data. Get to personal AGI.
https://t.co/5dWKd6v4UK