Stop paying frontier prices for work a cheaper AI would crush. Grab the model-picker prompt that routes the deck, the repo, and the call.

Source: Nate’s Newsletter Date: 2026-07-02 URL: https://natesnewsletter.substack.com/p/which-ai-model-to-use

Summary

Nate’s Newsletter argues that betting on a single “favorite” frontier model is risky since access gets throttled, repriced, or discontinued — “we rent the intelligence, we do not own it.” The piece packages a routing prompt that classifies a task first and assigns it to one of three tiers (daily driver, workhorse, frontier) rather than defaulting everything to the most expensive model, citing Coinbase and Cursor as companies already doing this.

Implications

Feeds the model-plurality / cost-routing thread: as the frontier ladder gets crowded (GPT-5.6, Opus/Sonnet line, GLM-5.2, DeepSeek V4, Kimi K2.6), the practical skill shifts from “which model is best” to “which task needs which tier” — the same dynamic that makes multi-model CLI support (Codex, Gemini CLI, OpenCode) valuable infrastructure rather than a convenience feature. Relevant to the tool-selection question this project’s reports feed into.