The core technology of Large Language Models is rapidly becoming a commodity, analogous to gasoline from different stations. This commoditization means that sustainable competitive advantage will not come from having a slightly better model, but from how companies use these models in conjunction with their unique, proprietary data.
A central argument is that the AI community has already met its long-standing definition of AGI, but has since 'moved the goalposts' toward 'superintelligence'. The current capabilities are more than sufficient to automate a vast number of enterprise tasks and generate immense economic value without needing further fundamental breakthroughs.
The conversation acknowledges the high failure rate of enterprise AI projects (95%) but reframes it as a positive sign of widespread experimentation. It contrasts the hype and bubble dynamics with concrete, successful deployments that are transforming core business processes in finance, drug discovery, and marketing.
The discussion highlights the rise of AI agents that can automate complex, multi-step processes, from generating equity research reports to managing marketing campaigns. The long-term vision, articulated by Glean, is for every employee to have a personalized AI companion that proactively manages tasks and understands their individual work context.
Keep pulling the thread on Ali Ghodsi, Arvind Jain.