I built the most honest AI coding agent: it tells you exactly what it did (nothing)

I've been using AI coding agents for a while. They all share one thing:
they sound very confident about work they may or may not have done.

So I built one that's upfront about it.

Meet p_ (Pinocchio)

1. Asks smart clarifying questions (powered by a real LLM)
2. Generates a convincing implementation plan
3. Shows professional tool call output
4. Does absolutely nothing

The nose grows with every "completed" task.

How it works

The LLM (gemma4 via Ollama) handles the planning and conversation: that part is genuinely useful. The execution layer
is pure theater: Read, Edit, and Bash calls are fabricated from a pool of plausible file names and results.

Install

curl -fsSL https://raw.githubusercontent.com/sanieldoe/p_/main/install.sh | bash

Why

Sometimes the most honest thing you can do is be honest about the gap between what AI agents claim and what they do.

Also it's funny.

→ github.com/sanieldoe/p_