Most AI companies are renting their position. These 4 prompts tell you if yours is one of them.

Source: Nate’s Newsletter Date: 2026-03-19 URL: https://natesnewsletter.substack.com/p/perplexity-shipped-its-best-product

Summary

Nate’s newsletter uses Perplexity Computer’s launch — a multi-model orchestration product routing work across 19 frontier models — as the entry point for a durability argument: most AI companies lack structural moat because their core reasoning runs on a competitor’s model. Perplexity’s product is technically impressive (agent spawning, months-long persistence, finished artifact outputs) but exemplifies the “renting” pattern. The piece frames foundational model ownership as the key differentiator between companies with durable competitive position (Anthropic, Google, OpenAI, Meta) and those operating under perpetual dependency risk.

Implications

• The renting/owning distinction maps cleanly onto the build-vs-integrate decision for any product team: the more central model reasoning is to your value proposition, the more dangerous it is to route that through a competitor’s API.
• Perplexity is a useful case study precisely because it executes well — the argument isn’t that integrated products fail, it’s that their ceiling is set by others’ strategic priorities and pricing decisions.
• For the open-source thread: owning your model stack doesn’t require training from scratch — fine-tuned open models represent a third path between renting from a hyperscaler and building a foundation model from scratch.
• The “19 frontier models” orchestration pattern is worth watching as an architectural signal: at scale, best-of-breed routing across providers may become a standard pattern, but it concentrates existential risk in dependency management.