When the Mirror Has No Blind Spots
How to find the patterns in your recurring meetings that you’re too close to notice
Someone I coach has been juggling a lot. Interview prep for competitive roles, a startup she’s building with a co-founder, coursework, and leadership commitments. We’ve had three sessions together, and the work has been good. She’s executing, absorbing feedback, and making visible progress.
After our last session, I tried something I’ve been experimenting with. I fed all three sets of meeting notes to Claude and asked it to find patterns I was too close to see. The output was long, and most of it confirmed what I already knew.
Then I hit this line.
“You are both using the same observation to reach different conclusions, and you have not noticed you are reaching different conclusions.”
I read it twice.
She had been saying (albeit in different ways across all three sessions) that now is the time to focus on building a startup, since she has fewer constraints and personal obligations at this point in her life.
I had been saying: “Since you have fewer constraints and personal obligations at this point in your life, you should keep your options wide open and build a diverse skill base before committing to one path. Now is the time to explore professionally.”
Same facts. Opposite takeaways. Three sessions in, and neither of us had named that gap. I was steering her toward breadth. She was building toward depth. We were both nodding at each other’s reasoning and walking away with completely different conclusions about what it meant.
That is the kind of thing that lives quietly in a relationship for months. You don’t notice it because you’re inside it.
The practice
For any recurring meeting, create a shared document that stays open between sessions. As topics come up during the week, add them. Invite the other party to do the same. When you meet, that doc is your agenda. Run an AI note taker and drop the summary into the same doc afterward.
Over time, you accumulate something most relationships don’t have: a record of what keeps surfacing, how topics evolved, and what never got resolved.
Then feed it to Claude with a prompt that asks for the interpretation you can’t give yourself. You don’t need a summary; you already know what’s in it. Here is the exact prompt I used:
The prompt
I’m sharing a record of meeting notes and agenda topics from recurring conversations with [person/group]. Read through the full document and help me see patterns I’m probably too close to notice myself. Skip the summary. I already know what’s in it.
Specifically, I want to know:
What topics keep recurring? For each one, what does the recurrence tell me? Is it an unresolved issue, a relationship dynamic, a structural problem, or something I keep revisiting without fully closing?
Where am I the bottleneck? Look for topics where progress seems to stall on my end, where I’m asking for the same things repeatedly, or where I seem to be creating work for others.
What am I not bringing to these meetings that the pattern suggests I should be? What’s conspicuously absent?
How am I showing up in these conversations — as a problem-solver, an escalator, a decision-maker, a blocker, something else? What does the language and topic mix tell you?
If you had to name one thing I’m avoiding based on what’s in this doc, what would it be?
What patterns do you see in what the other party brings? What does their topic mix tell me about what they’re worried about and what they need from me?
Where are we both circling the same issue from different angles without naming it directly?
If you were watching this relationship as an outside observer, what would your impressions be? What am I missing?
Tell me what this looks like from outside the room.
What to expect from the output
It will be long. Some of it will be wrong, because Claude is working from notes, not from being in the room.
Your job is to filter. The observation you wouldn’t have surfaced on your own is usually buried somewhere in the middle, and that alone is worth the twenty minutes it takes to set this up. Look for surprises, where the observation does not match your mental model of what you were trying to accomplish or how you intended to be perceived. Dig into those, and verify that it’s a genuine conclusion and not a misinterpretation of the notes.
When you are satisfied with how it works, ask Claude to create a skill for you and save it. When you paste meeting notes in the input, the skill will trigger automatically.
While questions 1 through 5 improve self-awareness, questions 6 and 7 in the prompt above are where the most uncomfortable and useful observations tend to live. Most people don’t think to ask about patterns in the other party’s communications at all. They feed in their notes and ask Claude to reflect their own behavior back at them, which is useful, but it’s only half the picture.
The failure modes
Notes have a structural problem that transcripts partly solve, but neither eliminates all of these:
Frequency isn’t significance
If a topic appears in all three sessions, Claude will flag it as a pattern. But some topics recur because they’re genuinely unresolved. Others recur because they’re standing agenda items, relationship rituals, or topics one party always raises regardless of context. Claude can’t tell the difference without more context than notes typically provide.
Salience bias in what gets written down
You write down what struck you. What didn’t strike you doesn’t make it into the notes, which means Claude is analyzing a filtered sample and has no way to know what was filtered. The silence in the document isn’t neutral. It’s shaped by what you thought mattered. Claude can’t see what you left out.
The prompt shapes the output
The prompt asks Claude to find where you’re avoiding something, where you’re the bottleneck, what’s conspicuously absent. Those are leading questions. Claude will find answers to them because you asked for answers, not because the answers are necessarily real. A different prompt framing would surface different patterns from the same notes.
The notes are already interpreted
Most people don’t write raw notes. They write their version of what happened. “She seemed resistant” is already a conclusion. “He kept circling back to resourcing” might mean he’s worried about resourcing, or it might mean that’s the frame you applied to three different comments he made. When you feed interpreted notes to Claude, you’re not getting a second perspective. You’re getting your own perspective reflected back with more structure.
Using the raw transcript eliminates salience bias and improves frequency signal, but increases output volume and doesn't resolve the interpretation problem or the prompt-shapes-output problem.
The output that feels most insightful is the output most worth interrogating, because insight-feel and accuracy are not the same signal.
Where to start
Pick one recurring meeting where you’ve felt like something is slightly off but you can’t name it. Could be a direct report, a peer, a stakeholder, a mentor. Create the shared doc today. Add the first topic. Invite the other party.
You don’t need three sessions’ worth of notes to run the prompt. Even one session with a few agenda items will start to show you something. The pattern gets sharper over time.
The gap between what you think is happening and what is happening has been there the whole time. Now you can see it.