OpenAI's Greg Brockman: Why Human Attention Is the New Bottleneck

Watch on YouTube ↗  |  April 30, 2026 at 17:27  |  28:27  |  Sequoia Capital
Speakers
Greg Brockman — Co-founder and President, OpenAI

Summary

Greg Brockman, co-founder and president of OpenAI, speaks with Sequoia partner Alfred Lin about OpenAI's compute strategy, scaling laws, AGI progress, agentic coding, security, and AI's role in science. He argues compute demand is effectively unlimited, scaling laws continue without a wall, and agentic coding tools are rapidly expanding into general computer work. He also highlights human attention as a new bottleneck and expects AI to accelerate scientific discovery.

  • OpenAI's business is buying, renting, and building compute to resell at a margin.
  • Greg says compute remains scarce and demand for intelligence is unlimited.
  • Scaling laws continue to improve model capability without a wall.
  • Agentic coding tools shifted from 20% to 80% of code writing in December.
  • Codex is expanding beyond software engineering to general computer work.
  • Human attention and evaluation are becoming key bottlenecks in AI-augmented work.
  • AI security and trusted access programs are a focus for responsible deployment.
  • Greg expects AI to drive a science renaissance, with early physics results.
Ideas
Greg Brockman Co-founder and President, OpenAI 1:05
Compute demand unlimited; supply remains scarce.
OpenAI's business is to buy, rent, and build compute and resell it at a positive margin. Since demand for intelligence and problem-solving is effectively unlimited, and they have never been able to keep up with demand despite aggressive compute ramp, compute remains structurally scarce. This supports continued long-term demand for AI compute infrastructure.
AI compute infrastructure HIGH
Greg Brockman Co-founder and President, OpenAI 2:39
Scaling laws continue; AI capability keeps rising.
Scaling laws remain a deep, empirical truth with no wall: pouring more compute into neural networks continues to make them more capable. OpenAI is investing in new architectures and algorithmic improvements, and Greg believes we are about 80% of the way to AGI. This supports long exposure to frontier AI model capability.
Frontier AI models HIGH
Greg Brockman Co-founder and President, OpenAI 7:34
Agentic coding tools becoming mainstream.
Agentic coding tools went from writing about 20% of code to 80% in just December, shifting from a side show to the main work. Codex is expanding from a software-engineering tool to a tool for anyone doing computer work, with new context and memory features like Chronicle. Greg expects this to spread across all computer work this year, supporting long agentic AI software.
AI-SECTOR HIGH
Up Next
1:14:02
MacroVoices #530 Daniel Lacalle: China and The Us Will Decide The Outcome of The Iran War
Macro Voices · Apr 30
22:35
Mastering AI Video Marketing w/ Magnific CEO Joaquín Cuenca Abela | AI Basics
This Week in Startups · Apr 30
32:27
This is AGI: Sequoia AI Ascent 2026 Keynote
Sequoia Capital · Apr 30
1:09:36
F1 Meets Finance in Miami | Open Interest 4/30/2026
Bloomberg Markets · Apr 30
2:11
Musk fires back at OpenAI attorney
CNBC · Apr 30

This Sequoia Capital video, published April 30, 2026, features Greg Brockman discussing AI compute infrastructure, Frontier AI models, AI-SECTOR. 3 trade ideas extracted by AI with direction and confidence scoring.

Speakers: Greg Brockman  · Tickers: AI compute infrastructure, Frontier AI models, AI-SECTOR