{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/d33fdb90fa48472ea4b32b966958cb1b\" frameborder=\"0\" width=\"1662\" height=\"1246\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":1246,"width":1662,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":1246,"thumbnail_width":1662,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/d33fdb90fa48472ea4b32b966958cb1b-f737dac3bb3e68de.gif","duration":184.351,"title":"Recursive Self Improving Minecraft Agent","description":"This Loom presents a recursive self-improving Minecraft agent project named Finding Nemo. The speakers, Arthur and Luis, explain that the goal is to build a bot that can adapt when its prior skills or world assumptions fail, contrasting efficient intelligence with crystallized intelligence. In a simple Minecraft world with three trees and three coal ores, the agent follows a torch-making skill chain requiring mining wood, crafting items, mining coal, and finally crafting the torch, while roadblocks like coal not dropping are introduced in some runs. They describe using traces that compare predicted rules to observed outcomes and note that the framework works as a proof of concept for adapting in changing environments."}