How This Startup Incubator Builds One Company Every Two Years

Quick Overview

The startup incubator, Every, follows a rigorous, iterative process exemplified by their P&C (Property & Casualty) venture, which involved continuous hypothesis testing, generating MBO-style objectives, and explicitly separating AI-assisted work from human-led validation to ensure high-quality, fact-based outcomes, rather than relying solely on AI-generated content.

Key Points: The incubator, Every, aims to build one company every two years, contrasting their approach with the rapid, often less rigorous, AI development seen post-ChatGPT. Their P&C company initiative involved a structured workflow (Steps 1-6) focusing on evidence-based hypothesis validation, using only internal Notion sources for context. The process mandates generating MBO-style objectives (What to validate, How to do it, Why it matters) for the next 1-2 weeks. They explicitly distinguish between AI-native companies (pushing the category ball forward) and AI-durable companies (using AI at the core) as two potential paths. The team actively uses tools like Augment Code for engineering tasks and has a rigorous process to avoid relying solely on unchecked AI output, often manually reviewing AI-generated content. The goal is to maintain high operational efficiency while ensuring quality, even when experimenting with new AI-driven product ideas. The speaker notes that their prior company, Moxy (a B2B vertical for aesthetic medical services), was successfully built using this structured, hypothesis-driven approach.

Context: The discussion centers on the operational philosophy of the startup incubator Every, specifically detailing the rigorous process they use to build companies, using their recent Property & Casualty (P&C) venture as a case study. The speakers contrast their structured, evidence-based approach with the rapid, sometimes superficial, deployment of AI tools seen elsewhere, emphasizing the importance of structured validation and clear goal-setting (MBOs) for success, even when leveraging AI assistance.