# You Might Not Need Frontier

> A source-backed AI model selection guide for 98 curated prompt examples, with three current model choices, a conservative class cushion, explicit exceptions, and verification advice.

The guide was reviewed 2026-07-25. It is curated, not a frequency ranking, benchmark, or model-equivalence claim.

## Start here

- [Interactive guide](https://youmightnotneedfrontier.com/): Search the 98 examples with lexical-first, browser-local semantic matching.
- [Complete machine-readable guide](https://youmightnotneedfrontier.com/llms-full.txt): Every prompt, route, recommendation, frontier exception, verifier, and evidence link.
- [Markdown page mirror](https://youmightnotneedfrontier.com/index.md): A clean Markdown representation of the full one-page guide.
- [Prompt-specific router](https://router.samcarlton.com/): Review a prompt that does not closely match the curated examples.

## Guide facts

- 98 curated prompt examples
- 4 route types
- 17 current model records
- 8 reviewed sources
- Reviewed 2026-07-25
- Author: [Sam Carlton](https://samcarlton.com/)

## Method and provenance

- [Method](https://youmightnotneedfrontier.com/#method): Start low, verify, and escalate based on task demand, prompt readiness, risk, and verifier strength.
- [Evidence and sources](https://youmightnotneedfrontier.com/#sources): Dated source ledger and guide limitations.
- [When frontier earns its keep](https://youmightnotneedfrontier.com/#frontier): Six signals for retaining frontier capability.
- [Token efficiency versus model routing](https://youmightnotneedfrontier.com/#token-efficiency): Why lower model class does not automatically mean fewer end-to-end tokens.
- [Model source and pinning policy](https://youmightnotneedfrontier.com/#model-policy): Current model records, headroom rule, and evaluation boundary.