OpenAI Codex has pulled ahead of Claude Code in daily agentic workflows, according to Riley Brown and Ras Mic in this 78-minute breakdown of every major AI agent release from Q1 2026. The video covers Opus 4.5, the SpaceX acquisition deal with Cursor, and the emergence of OpenClaw as a framework for orchestrating agents on what they call the agentic personal computer.
The most technically dense sections are worth reading in full: the memory layer discussion at 33:12, the computer use speed benchmarks at 36:23, and the Codex vs Claude UX comparison at 08:42. Their core argument is that integrations outperform prompting skills, and that keeping agents narrow beats building broad ones, which cuts against most current hype around autonomous super-apps. Google gets singled out as structurally slow to respond.
The final hour shifts to predictions: agents acting as customers in agent commerce, knowledge work disruption as a near-term event rather than a forecast, and why first-mover advantage still exists right now for builders. The framing is not aspirational. These two are building in the stack they are describing.
[WATCH ON YOUTUBE →]