Particle.news

OpenAI Codex Lead: Human Typing Speed Is the Bottleneck to AGI

He argues for agents that validate their own work to free humans from prompt writing.

Overview

  • On Lenny's Podcast, Alexander Embiricos described human typing or multitasking speed as the "current underappreciated limiting factor" to progress toward AGI.
  • He said current workflows slow because people must draft prompts and verify model outputs, keeping humans as the throughput constraint.
  • His proposed path forward is agent-level self-review so systems become "default useful" without constant human validation.
  • He predicted a "hockey stick" in productivity for early adopters starting next year, with gains later spreading to larger companies and feeding back into AI labs, a forecast rather than a confirmed plan.
  • He cautioned there is no one-size-fits-all route to fully automated workflows, and coverage by Business Insider and India Today presented his remarks as expert opinion with AGI still a theoretical goal.