Why token-maxing AI agents is wasteful and risky
The prevailing attitude is 'token abundance — I'm going to token max,' but the smarter approach is the opposite. Doing the simpler thing has a lower likelihood of breaking. If you hand Hermes the keys to run your Facebook ads, there's a high likelihood it could absolutely nuke the account. But if you build a piece of custom software that runs the way a real human would operate it, you get meaningfully higher-quality outcomes. Complexity for its own sake introduces failure points; simplicity is a feature, not a limitation.
Go deeper with Greg Isenberg
Ask anything about this, and Greg Isenberg answers from everything they've shared.
