You leave the conversation and replay it on the walk home. Did I sound smart enough? Did I make the right joke? Was that pause too long? The other person, somehow, didn't make it into the replay. Only you did. The conversation was a small stage, and you were the only one on it β€” even though there were two of you in the room.

Maybe the problem isn't that you're not interesting. Maybe it's that you've been working on the wrong job. The job in a conversation isn't to be interesting. It's to be interested. And being interested, it turns out, is a structure β€” not a personality.

That made me think about the conversation layers experiment.

The question: What changes when I stop trying to be interesting and instead move through four layers of curiosity in each conversation?

The hypothesis is small. If I use one of four moves β€” an open question, a mirror, a reflective response, or an emotion label β€” in one real conversation a day for seven days, conversations will feel warmer, less performative, and more revealing for both people. Not because I've gotten more charismatic. Because curiosity, given a small staircase, finally has somewhere to climb.

The signal: five days of seven captured. Three conversations where the other person opened past surface facts. One pattern about which layer creates the most ease.

What you do for 7 days

The four layers, ordered from easiest to deepest:

  • Open question. Replace yes/no with how or what. "What was that like for you?"
  • Mirror. Repeat the last 1–3 words of their sentence with curious tone. "…too much pressure?"
  • Reflective response. Paraphrase what you think you heard. "So it sounds like the part that hurt was being unseen, not the workload itself."
  • Emotion label. Gently name the feeling under the words. "That sounds lonely."
  • For seven days:


    What this experiment grows

    It is not communication skills. It is the muscle of attention with another person. Most social tiredness comes from running two conversations at once β€” the real one, and the loop in your head about how you're doing in it. The four layers reroute attention to the actual conversation. The loop quiets, almost as a side effect.

    Curiously, after seven days, the layers themselves matter less than what they trained: the small willingness to stay one beat longer with what the other person said, before reaching for what you wanted to say next.

    Where it pairs

    This experiment walks with the witness deficit pain β€” the loneliness of being seen for output but never known in process. The deficit isn't only about being witnessed; it's also about practicing the witness. The four layers grow that muscle in both directions. The Witness Bot workflow becomes the natural daily companion β€” same attention, turned toward your own work after it ships.

    One week. Four small layers. One conversation a day. The other person, finally in the replay too.