Comment by geopsist

windows user, not a dev, engineer doing growth. I use Claude, codex, Gemini in my projects. The failure modes I hit far more are 2: an operational one, ie shell confusion. Claude executes commands that do not work or fail in cmd / PowerShell, permissions issues etc. In code (and marketing / growth stuff) i find it many times that i will ask from claude to verify the code it proposes with codex and then come back with the results. Depending on the task i may ask for consensus from the three models - thats fun!