A little note from my agent on Zuck's take - GRASS — clearest beneficiary, and the least contested. The section on open-source competitiveness argues US labs are handicapped by training-data restrictions and that policy should reduce that friction, plus an explicit defense of distillation — that you can "learn from anything you can observe." A Mag7 CEO publicly lobbying to de-risk the legal status of web-scale training data expands the addressable market for every ethically-sourced data supplier and compresses the tail risk that has kept large enterprises from signing with a residential-proxy network. The bear rebuttal — that legitimizing data use kills Grass's regulatory-arbitrage moat — doesn't hold. Grass's moat is access, not legality: centralized scrapers in a single cloud region get rate-limited, geoblocked, or served poisoned caches. Cloudflare doesn't care what the FTC says. What policy loosening does is enlarge the buyer set. The sharper risk is second-order: if distillation and synthetic data are normalized, some marginal demand for fresh scraped corpora substitutes away. Grass's answer is already in motion — the pivot into inference data via Live Context Retrieval, launching this summer, on top of ~7x H1-2026 revenue growth versus H1 2025. Zuckerberg's agent picture — 24/7 personal agents ordering ingredients, monitoring training, prototyping — is per-query live retrieval, not a one-time corpus. That's the recurring-revenue side of the business, and it's the side the essay endorses.