Alexander Embiricos, head of product development for OpenAI’s Codex coding agent, stated on Lenny’s Podcast that human typing speed limits progress toward artificial general intelligence because it constrains prompting and output validation.
Embiricos identified the current underappreciated limiting factor to AGI as human typing speed or human multi-tasking speed on writing prompts. Codex functions as OpenAI’s coding agent, designed to assist with programming tasks. His comments appeared on the podcast episode aired on Sunday.
Artificial general intelligence refers to a theoretical form of AI capable of reasoning as well as or better than humans across diverse tasks. Major AI companies compete to develop AGI first.
Embiricos explained that an AI agent can monitor a user’s entire workflow, yet the process remains constrained if the agent does not validate its own outputs. He stated, “You can have an agent watch all the work you’re doing, but if you don’t have the agent also validating its work, then you’re still bottlenecked on, like, can you go review all that code?” This review step depends on the user’s typing speed.
To overcome this barrier, Embiricos advocated unburdening humans from the tasks of writing prompts and validating AI-generated work, given the slowness of human input. He emphasized rebuilding systems to position the agent as the default useful tool in workflows.
Such changes would enable hockey-stick growth, a pattern where productivity remains flat initially before spiking sharply, resembling the shape of a hockey stick. Embiricos said, “If we can rebuild systems to let the agent be default useful, we’ll start unlocking hockey sticks.”
Embiricos noted no single path exists to fully automated workflows. Instead, each use case demands a tailored approach to integrate AI agents effectively.
He projected that early adopters will experience these productivity spikes beginning next year. Over subsequent years, larger companies will achieve similar gains as they adopt AI agents more comprehensively.
AGI will emerge during the interval between early adopters’ initial productivity increases and the point when tech giants fully automate processes using AI agents. Embiricos stated, “Somewhere in between the time early adopters start to see gains in productivity and when tech giants manage to fully automate processes with AI agents is when we’ll see AGI.”
These productivity surges will channel resources back into AI research labs. Embiricos concluded, “That hockey-sticking will be flowing back into the AI labs, and that’s when we’ll basically be at the AGI.”
