I Asked Claude to Explain My Own Codebase to Me. Its Answer Was Better Than Mine.
I wrote most of it. Still couldn’t explain it as clearly. Make of that what you will.
Someone in a planning meeting asked me to walk the team through how our pricing engine actually worked — a service I’d built, largely alone, about eighteen months earlier. I said yes without thinking twice about it, because of course I could explain something I’d written. I started talking, got about two minutes in, and heard myself trail off into a vague, hand-wavy description of “and then it sort of adjusts based on a few factors,” which is not an explanation, it’s the sound of someone who used to understand something reaching for a version of it that no longer fully exists in their head. Someone asked a clarifying question I couldn’t cleanly answer. I said I’d follow up. That was uncomfortable in a specific, quiet way that stuck with me longer than the meeting itself.
That evening, mostly to figure out what I actually didn’t remember before writing the follow-up, I pointed Claude Code at the service and asked it to explain what it did. What came back was clearer, more complete, and more precisely organized than anything I’d said in that meeting — including details about how three specific pricing factors interacted that I’d apparently forgotten I’d even implemented, laid out in an order that made the whole system’s logic obvious in a way my own live explanation, from memory, under mild social pressure, had completely failed to achieve.