Start research by exhausting frontier models first, only then claim a durable insight

FinanceYF5 · x · 2026-09-18

Continuing the thread, the author argues research must change its starting point: every project should first try to solve the problem with existing frontier models, analyzing their performance and failure modes in detail.

Only once current models demonstrably fail should new methods be proposed — and they must offer a core insight that survives the next generation of model updates.

Related event: VIGA thread: exhaust LLMs first; paper cycles lag behind models(2 posts)→

Original post →

More from Research

Research channel →