Claude Clue #2: Stop Commanding, Start Checkpointing

Most of us learned to interact with computers through search engines—type a command, get a result. So we bring that same energy to Claude: "Give me this," "Do that," "Write me a..." Then we're frustrated when it misses something obvious. But here's the thing: the model only knows what's in the context window right now. It can't read your mind about where this fits in the bigger picture. The fix is simple—drop guideposts throughout your conversation. These are little moments where you and Claude sync up before moving forward.

This can start before any code gets written. Instead of "add feature X," try "I want to add feature X—review the current code and make a plan for how that would work." Now you've got something to react to, give feedback on, and refine together. You're building shared context before the real work begins. Then as you go, keep those checkpoints coming: "Anything unclear before we tackle the next piece?" or "I don't love that structure, let's try it this way." It doesn't need to be formal. You're just making sure neither of you drifts off course—and suddenly those "now I have to fix everything" moments get a lot rarer.

Claude did the heavy lifting. I did the thinking.