Build log

What we got wrong while building OLP, what we changed, and what we threw out, in date order.

Before you approve something irreversible, see what it will change.

There’s a name for it: approval fatigue. When an AI agent asks for permission, people mostly choose Yes. Anthropic reports that Claude Code users approve 93% of permission prompts (March 25, 2026). Before that Yes, Outcome Lookahead Protocol (OLP) shows you what will happen if the command runs and what else you can do. You still make the call. This series is our record of what we got wrong while building it, in the order it happened. Anthropic’s post

  1. Why I built FeedPia

    I set out to take some strain off my parents' shop. I ended up building something for the moment before you approve what an AI agent wants to do.

    Part 0 Namgyu Eom, Founder of FeedPia

  2. Why we fixed our own rules four times

    Four times in the past two months, we applied a rule we’d set for ourselves too broadly or too narrowly, and had to fix it.

    Part 1 FeedPia

  3. We installed OLP on our own PC, and our own AI session stopped first

    We installed OLP, the tool we make, on our PC, and two commands from our own AI session stopped before they ran. Here is what stopped and why.

    Part 2 FeedPia

At your next permission prompt, look before you choose. It’s one line to install.

See the one-line install