Left my laptop running at 9am. Came back at 3pm to 5 shipped features. I spent less than 30 minutes total interacting with it. 234 million tokens later, some honest takes on @FactoryAI's Droid: --- THE GOOD --- The autonomous feeling is genuinely addictive You can truly be away. No more wearing an earpod waiting for the "click" sound when something is complete. It handled product complexity The task I gave it has less technical difficulty but more about navigating the product complexity, and it did it really well. Clean UI, seamless mid-session model switching I swapped models mid-run from Opus to GPT5.4 and it handled the transition without breaking context. Small piece meal commits Unlike working with Claude Code, Droid breaks tasks into smaller commits. History is easier to track. --- THE REAL --- Undershot task count by 50% (predicted 20, took 33) AI scope creeps as well I guess. Slower than me in a Claude Code flow state Sometimes, when looking at it, it feels slower than what I + CC can do. --- THOUGHTS --- ...