Skeptic: Closed Models Got Faster After DFlash Paper, Hinting at Borrowed Secrets
xeophon · x · 2026-08-13
A user points out that while people think closed model labs have secret sauce, all closed models mysteriously got faster after the DFlash paper release, suggesting they may have borrowed from open research.
More from AGI Musings
- The Economics of Open vs. Closed AI: Value Shifts Downstream — mikeflache · 2026-08-13
- Scholar Claims: Those Romanticizing Traditional Education Will Be Economically Replaced by AI — RexDouglass · 2026-08-13
- Philosopher Argues for Decoupling Survival from Competition in the AI Era — marigo · 2026-08-13
- Tech Managers Who Don't Understand AI Are the Biggest Corporate Risk — davidhoang · 2026-08-13
- MIT Tech Review: Diffusion Models Among Four Approaches Defining Next-Gen LLMs — volokuleshov · 2026-08-13
- AI bubble? Anthropic revenue soars from $5B to $47B, CoreWeave backlog $104B, says ChrisGPT — ChrisGPT · 2026-08-13