[{"data":1,"prerenderedAt":66},["ShallowReactive",2],{"term-a\u002Fagent-thrash":3,"related-a\u002Fagent-thrash":59},{"id":4,"title":5,"acronym":6,"body":7,"category":40,"description":41,"difficulty":42,"extension":43,"letter":44,"meta":45,"navigation":46,"path":47,"related":48,"seo":53,"sitemap":54,"stem":57,"subcategory":6,"__hash__":58},"terms\u002Fterms\u002Fa\u002Fagent-thrash.md","Agent Thrash",null,{"type":8,"value":9,"toc":33},"minimark",[10,15,19,23,26,30],[11,12,14],"h2",{"id":13},"eli5-the-vibe-check","ELI5 — The Vibe Check",[16,17,18],"p",{},"Agent thrash is when an AI agent keeps changing its mind, flipping between approaches without committing. 'Let me try approach A. Actually B. Wait, A.' Each reversal costs tokens. Nothing gets done.",[11,20,22],{"id":21},"real-talk","Real Talk",[16,24,25],{},"Agent thrash describes oscillating behavior in an agent where the model repeatedly reverses decisions without making progress. Causes: ambiguous task specifications, conflicting tool outputs, or sycophantic reactions to pushback. Mitigations: commit-based workflows, explicit decision logging, and preventing the agent from reversing without new evidence.",[11,27,29],{"id":28},"when-youll-hear-this","When You'll Hear This",[16,31,32],{},"\"Agent thrash cost us 200k tokens with nothing to show.\" \u002F \"Stop agent thrash by enforcing commit-then-verify.\"",{"title":34,"searchDepth":35,"depth":35,"links":36},"",2,[37,38,39],{"id":13,"depth":35,"text":14},{"id":21,"depth":35,"text":22},{"id":28,"depth":35,"text":29},"ai","Agent thrash is when an AI agent keeps changing its mind, flipping between approaches without committing. 'Let me try approach A. Actually B. Wait, A.","intermediate","md","a",{},true,"\u002Fterms\u002Fa\u002Fagent-thrash",[49,50,51,52],"Agent Loop","Sycophancy Loop","Agent Debugging","Tool Use Loop",{"title":5,"description":41},{"changefreq":55,"priority":56},"weekly",0.7,"terms\u002Fa\u002Fagent-thrash","79XAWap5UyXqeRgkFv5pnhCNHyRYn2t9UERIfNUBF9U",[60,63],{"title":50,"path":61,"acronym":6,"category":40,"difficulty":42,"description":62},"\u002Fterms\u002Fs\u002Fsycophancy-loop","A sycophancy loop is when the AI agrees with everything you say, including when you're wrong. 'You're right, my previous answer was incorrect.",{"title":52,"path":64,"acronym":6,"category":40,"difficulty":42,"description":65},"\u002Fterms\u002Ft\u002Ftool-use-loop","A tool use loop is when an AI agent gets stuck calling the same tool over and over. It reads a file, reads it again, reads it again.",1776518254519]