Clearing chat history helps AI agents find better solutions

Microsoft researchers found that autonomous AI agents become trapped in "idea basins" — local solution regions they cannot escape — as their chat context grows, and their "Fork-and-Flush" method, which clears the chat history while keeping the workspace, achieved up to 66 percent better results on the same compute budget. The finding addresses how accumulating context degrades agent problem-solving.