High ยท Killed
Startup death signals in the AI cycle
A practical checklist for spotting AI startups likely to be absorbed, repriced, or killed by platform bundling before the market admits it.
The point of the checklist
The question is not whether a product is clever. The question is whether its value survives when model quality improves, model cost falls, and the main platforms bundle the workflow.
The fastest failure mode
A startup is most exposed when it sells horizontal productivity gains without measurable customer-side output. Those claims are easy for platforms to copy and hard for customers to verify.
How to use this page
Use the signals as an editorial screen for future anti-ai.app events. When a new AI update ships, ask which exposed startup patterns match these signals and move them onto the casualty map.
Signals to watch
- No proprietary distribution: The company rents attention and has no owned channel, system of record, or embedded workflow.
- No vertical workflow ownership: The product can be replaced by a better general assistant because it does not own a specialized operating context.
- Token-cost pricing: The product's pricing story weakens as foundation model API prices fall.
- Free-tier collision: The core job appears in ChatGPT, Claude, Gemini, Copilot, or a cloud agent at no extra vendor contract.
- Services hidden as software: Revenue depends on implementation labor while the company asks for software multiples.
- No measured output: The pitch is productivity, but the buyer cannot tie usage to cycle time, revenue, cost, or risk reduction.
Representative companies
These are examples of the category pressure or survivor pattern, not a claim that the companies have failed.
- Zapier (zapier.com)
- Replit (replit.com)
- Airtable (airtable.com)