My OpenClaw token burn rate estimate: - Fixing various bugs, debugging, verification, etc., burned 70% - Daily random check-ins (hello, still alive? how’s it going, etc.), burned 20% - Other stuff, burned the remaining 10% For high-creativity tasks, I’ve already renewed my subscription to a good IDE and upgraded with Claude Code/Codex/Gemini… I honestly can’t believe the output from a so-called autonomous AI Agent after burning through a massive amount of my tokens. I just don’t trust the quality of its results… That said, if you treat it as trying something new, then it’s not a big deal.