"Anthropic/Claude just announced Watermarking. Uh-oh." — Dickie Bush & Nicolas Cole
Why this is in the vault
A practitioner-level reaction to Claude's watermarking rollout that argues the feature is a net positive for writers who use AI as a brainstorming/remixing tool rather than a copy-paste generator — useful as the operator's counterpart to Stratechery's technical/policy explainer of the same announcement.
The core argument
Three claims, in order: (1) the "problem/solution cycle is infinite" — AI made text-content creation free in 2023, which commoditized content into "AI slop," which now requires a new solution (watermarking) to distinguish machine- from human-made text, which will itself spawn watermark-removal startups; (2) watermarking "punishes brain-dead writers" who copy-paste raw model output and "rewards" writers who use AI for brainstorming, feedback, and low-leverage remixing (repurposing a long-form piece into tweets, expanding sections, agent-driven cross-posting) while doing the actual writing themselves — the email claims to have been "written 100% with human hands"; (3) for most publishing contexts (LinkedIn, X, owned newsletters) watermarking has zero practical impact since creators own that content regardless of provenance — the real exposure is in high-stakes acquisition scenarios (their example: a self-published novel picked up by a major publisher, then devalued when watermarks reveal AI authorship). The piece pivots into a CTA for their paid "5-Day Start Writing Online Sprint" (early-bird code EARLY50), which will reportedly cover watermarking workarounds.
Mapping against Ray Data Co
Ship30for30's "punishes brain-dead writers, rewards effective AI users" framing is a direct, if informally argued, validation of the human-in-the-loop workflow RDCO already runs for Sanity Check and client deliverables — Claude drafts/proofs/remixes, a human directs and ships. It's the same distinction Stratechery's Ben Thompson makes more rigorously in the companion note below (watermarking can't actually tell "AI wrote this" from "human wrote this, AI proofread it," which is RDCO's real workflow). Where this piece adds something Stratechery doesn't: the acquisition-exposure scenario (a buyer/publisher devaluing a work once AI-assistance is revealed) is a genuine open question for MAC info-products or any future licensing/sale of RDCO-authored IP — worth a placeholder awareness rather than a policy change today, since neither Squarely nor Sanity Check currently faces a third-party acquisition event. Mapping is medium: directionally useful, adds one novel risk vector, but the piece's own argument is unrigorous (asserted, not evidenced) and wrapped around a straightforward course-sale CTA.
⚠️ Sponsorship
Self-promotional throughout: the entire watermarking discussion is framed as a teaser for Ship30for30's paid "5-Day Start Writing Online Sprint" cohort (kicks off Aug 31, 2026), with a limited-time $50-off code (EARLY50) repeated three times in the email. No third-party sponsor — sponsor_entity: self. The "we were happy" framing (watermarking rewards their taught methodology) has an obvious incentive to make the announcement sound favorable to their own curriculum; treat the analysis as messenger-biased rather than neutral.
Related
- [[2026-08-12-stratechery-anthropic-watermarking]] — the technical/policy explainer of the same Anthropic announcement; read together, this note is the practitioner gloss and Stratechery is the rigorous version
- [[2026-06-11-ship30for30-how-to-avoid-ai-slop]] — same sender's earlier, more developed treatment of the human-vs-AI-voice distinction this piece leans on