Crosshatch

Crosshatch is best read as a design document rather than a vendor. The homepage itself frames the site as a record — "This doc records all our Docs site copy Blogs should they be useful to anyone attempting a new (AI-agnostic!) identity layer for personalized AI" — and states plainly, "We shut down July 2025." The material left behind is substantive. The 11/17/2023 manifesto argues that the surveilling Data Moats which could power hyper-personalization "just aren't that big. Even the big ones aren't that big. They're fragmented and sparse," and that the experiences worth personalizing — books read, trips taken, trails run — are scattered across services but are still the user's own records. The proposed answer was tap-to-consent sharing: an AI-agnostic identity layer where a single consent tap lets an app draw context from another service, and where the user retains control ("Might delete later. If it were safe."). The 12/11/2023 essay explores the consumer-facing version of this through the Jeff persona, arguing that Spotify Wrapped is compelling because it celebrates real moments, and asking why those moments couldn't be celebrated "from everywhere on everything." Strengths as a reference: the consent model is stated concretely, the critique of the data-moat assumption is specific, and the AI-agnostic framing is a real architectural stance rather than a slogan. Weaknesses: it is archived, so none of it is maintained; there is no live API, no SDK, and no working infrastructure; and the business did not survive, which is itself the most important data point for anyone planning to depend on it. Where it fits: as required reading for a team designing consent, portability, or context-sharing into a personalization feature, and as a cautionary example when writing an internal proposal. Where it doesn't: anywhere in a production architecture, a vendor-evaluation shortlist, or an enterprise procurement process. Treat every page as a 2023–2025 historical artifact.