Ex-OpenAI researcher reveals the culture of betting on long-term, improbable projects like GPT-live

jachiam0 · x · 2026-08-24

Former OpenAI researcher @kundan2510 reflects on 14 months at the company, highlighting leadership's full support for seemingly impossible projects like full-duplex models (gpt-live). The rule is simple: if it's theoretically possible and magical, they pursue it despite technical hurdles. As understanding grows, the problem becomes solvable, reflecting a culture of long-term, high-upside research bets.

Related event: Ex-OpenAI Researcher Reveals Leadership's Full Backing for Full-Duplex Speech Models(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →