Abstract. A growing class of AI deployments is a fleet: many scheduled agent runs, each starting with
an empty context, that cooperate over days through shared files and may span several models and
vendors. Every run pays to rebuild its working state before doing any work.
Abstract. A growing class of AI deployments is a fleet: many scheduled agent runs, each starting with
an empty context, that cooperate over days through shared files and may span several models and
vendors. Every run pays to rebuild its working state before doing any work.
@amyoder I replaced every single researcher I had with AI. Cleaned out a 20 person lab. It’s a little quiet and I lost all personal relationships but the profit is great
@macrocephalopod@GollumGekko1 The markets you’re talking about sometimes have 5000$ total liquidity for the duration of trading. It’s really too thin for the blue whales.
@Kalshi is a fantastic and futuristic company. Contracts trade fundamentally different from traditional bets. I only hear complaints from non users. Thank you for the immense enjoyment I get from the app. It’s my favorite social media.
HOW TO SAVE TOKENS:
When you use AI for mundane tasks, use older model tokens.
When you use AI for complex tasks use frontier model tokens.
When you dream, hope, imagine,create, use your brain tokens. The brain however is token limited. So be efficient.