Josh Woodward: Google Labs is Rapidly Building AI Products from 0-to-1
Watch on YouTube ↗ Summary based on the YouTube transcript and episode description.
Josh Woodward explains how Google Labs ships AI products in under 100 days and why 25% of Google’s code is now AI-written.
- 25% of all code written at Google is now AI-generated, up significantly in a short period.
- Google Labs targets 50–100 days from idea to users’ hands as a core cultural metric.
- Mariner (Google’s computer-use agent) went from concept to Chrome extension in 84 days; accuracy and speed remain unsolved.
- VO2 video model reduces the ‘cherry-pick rate’—how many generations needed to get one good clip—to roughly one.
- Generative video cost has dropped ~97x in one year for text models; Woodward expects video economics to turn viable in quarters, not years.
- Coding agents are underhyped despite hype: self-correcting, self-healing code models could dramatically change the progress curve.
- Long context / infinite context is a strategic lever Google is heavily investing in, especially for agentic chained workflows.
- Woodward argues taste and design will become scarce and high-value as AI slop proliferates at the application layer.
2025-03-18 · Watch on YouTube