Gryps Research Engine

The Gryps Research Engine is the weekly research and publication pipeline behind blog.gryps.finance↗. It pulls institutional perp DEX market intelligence. Volume, execution quality, institutional movement, SEI ecosystem developments, the research landscape. Then drafts long-form analytical pieces, three-pass fact-checks every substantive claim, and stages the result in Sanity for editorial review.

The pipeline runs four orchestrated skills: an Intelligence Engine that runs Monday and gathers the week's structured intel; a Content Pipeline that takes a thesis or intel file and produces a full article through brief, draft, self-edit, and fact-check; a three-pass Fact-Checker that scores every claim and auto-stages to CMS when accuracy clears 85%; a Sanity Push that converts markdown to Portable Text and creates the unpublished draft. A weekly Distributor handles the social calendar on top.

The architecture is the noteworthy part. Most "AI content engines" are a single skill that drafts a blog post. The Gryps Engine is a four-skill orchestration where each skill has a job that respects the editorial register: the Intelligence Engine doesn't draft; the Content Pipeline doesn't fact-check; the Fact-Checker doesn't publish; the Sanity Push doesn't auto-publish. The handoffs preserve the discipline the institutional audience expects.

Honestly, the most surprising lesson is that the rate-limiting step isn't AI quality. It's editorial taste. The Engine is functionally ready to publish more than it does; the constraint is whether each piece earns its place in the institutional research register. That constraint is on me, not the agents.