AI Co-founder (in-browser prototype)

Chat

Scores six dimensions · answers in five sections

Memory 0/20

Enter to send · Shift + Enter for a new line

The original workflow in the n8n editor

What replaced what

Chat Trigger: This chat window
AI Agent + system prompt: Rule engine that follows the same prompt: six dimensions, clarifying questions, five-section answer
DeepSeek V3.1 model: Keyword classifier (12 startup categories) + signal extraction + 1–5 scoring with visible reasons
Simple Memory (20): A 20-message window, so follow-ups update the analysis instead of starting over

How the verdict is calculated

Each dimension starts at a category baseline and moves in half-points for signals found in your text: a named customer, a stated pain, a revenue model, differentiation words, regulated domains (medical, finance, children), hardware or logistics, network effects, geography and traction. The verdict is a weighted mean with feasibility counting most (35%): High ≥ 3.55, Medium ≥ 2.7, otherwise Low. Open "why this score" under any answer to see every step.