Training data access is not a long-term moat—it's abundant and freely available (internet scraping, public domain books), and advances in training on small datasets will make proprietary data less valuable over time

causalpending

Speaker

Mark Anderson

Evidence Quote

“Is training data hard to come by? You know, no, it's it's actually, you know, pretty straightforward. You can scrape the internet.”

Source

Moment of Zen w/ Marc Andreessen on SBF/FTX, Twitter, and Open AI/Chat GPT— Moment of Zen
Created: 8/10/2026, 10:07:04 PM

Your notes

Loading notes...