Episode 1 of 6, What AI Actually Does: what is an AI agent, really? I ran Qwen3-8B on my MacBook on ten small tasks, 100...

Episode 1 of 6, What AI Actually Does: what is an AI agent, really? I ran Qwen3-8B on my MacBook on ten small tasks, 100 runs as a plain chat and 100 in a loop of Python (34 lines): 41/100 vs 76/100. The lift is on tasks that need a file (0/40 vs 23/40, with 28 made-up answers from the chat), none on the control tasks, and about 5.5x the tokens. https://youtu.be/5E1eCCjWZc8What would you hand an agent first?#AI #AIAgents #Qwen3 #llamacpp

Read Original

Related