A Four-Quadrant Framework for Deciding Whether Your AI Startup Is Defensible

joecole · x · 2026-09-03

Max Kolysh lays out a four-quadrant framework for AI startup defensibility: bounded short-horizon tasks (e.g. "order me DoorDash") will succumb to rapidly improving open-source models; unbounded long-horizon tasks (frontier coding) demand huge capital and a knife fight with the labs. The two viable quadrants for startups: bounded long-horizon tasks (specialized vertical workflows tied to proprietary systems and customer data) and unbounded short-horizon tasks (design and content creation), where taste, context and rapid iteration compound. Framed against an article noting 2028 AI capex may exceed France's budget.

Original post →

More from AGI Musings

AGI Musings channel →