The pipeline.

Eleven stages, one page each. Same shape every time.

Every stage page tells you the same six things. What goes in. What comes out. Which prompt version is current. Which models are recommended. How it fails. What the evals look like. If a page deviates from that shape, that is a documentation bug. Tell us.

The map

• 01discover()machine
• 02web_audit()machine
• 03social_research()machine
• 04qualify()machine
• 05extract_signals()machine
• 06draft_brief()machine
• 07enrich_contact()human gate
• 08draft_email()machine
• 09human_review()human gate
• 10send()human gate
• 11track()machine

Reading order

If you are an engineer evaluating Paitho, read extract_signals() first. That is the stage where the product earns its name. If you are an operator, read qualify() first. That is where most of your leads die.

Conventions on these pages

Function names are written like code: qualify(). Versions are semver: v1.4.2. Prompts are stored in the prompt library under the same name and version. The page titled qualify() is the canonical reference for whichever version is currently in production. Older versions are archived but readable.