Frontier labs build lock-in using proprietary memory and reasoning layers
A small X exchange from Nikesh Arora and Brad Gerstner points to a familiar enterprise tradeoff: better performance now, less model flexibility later.
1812175649.7K
Story Brief
On X, Nikesh Arora backed a prediction that frontier AI labs will build durable lock-in through proprietary memory, reasoning and tool layers, writing, "Yes. What else will fund the compute?" In a reply on X, Brad Gerstner added that enterprise buyers have shifted more AI spending toward closed-source frontier models over the past six months because, he wrote, decision-makers see a single model's advantages as worth the downside of being "single threaded."
“Percentage share of wallet / dollars is going up at closed source / frontier models over last 6 months. Decision makers are saying with their choices that the advantages accruing to a single model outweigh the downside of being single threaded for now.”
Brad Gerstner@altcap