Vin Patel is an AI technologist and published author with 25+ years of experience. He is the creator of Manuscript (open-source AI content detection), AEORank (AI Engine Optimization), and the Bhagavad Gita App. His work has been archived by the British Library and cited by UKCERT.

AEORank is an open-source AI visibility platform that scans any website across 36 criteria, scores it from 0-100, and generates 9 deployment-ready files that improve discoverability by AI engines like ChatGPT, Perplexity, Claude, and Gemini. Includes a free CLI, dashboard at app.aeorank.dev, GitHub App, and 13 framework plugins. Visit aeorank.dev.

What is the Bhagavad Gita App?

A verse-by-verse platform at bhagavad.net presenting all 700 Gita verses through 8 philosophical traditions with multi-tradition synthesis, life applications across 4 pillars, and 5,600+ searchable life questions. Open source and MIT licensed.

Manuscript is the only open-source AI content detector that runs 100% on your infrastructure. It detects AI-generated text, images, audio, and video with zero external API calls. Built in Go, self-hosted via Docker. Visit manuscript.dev.

How can I work with Vin Patel?

Vin is available for speaking engagements, podcast interviews, and consulting on AI strategy. He also runs the IdeaForge Workshop, a 4-day intensive for building AI products. Contact vinpatel.pro@gmail.com or visit vinpatel.com/speaking.

Mistral, Mystery Models, and the MCP Wobble: This Week's AI Stack Is Shifting

The signal: Mistral held its AI Now Summit this week while a mysterious model called Hy3 is quietly topping OpenRouter rankings by a wide margin — and the community is debating whether MCP already has an expiry date.

Why it matters: If you’re building on top of any single model provider or protocol right now, the ground is shifting faster than your roadmap. Hy3 coming out of nowhere to lead OpenRouter is a reminder that “best model” has a shelf life measured in weeks, not quarters.

The pattern I’m watching: The MCP-is-dead conversation is the same one we had about OpenAPI, GraphQL, and every “standard” that got ahead of real adoption. Protocols die when the tooling doesn’t follow — watch whether the major SDK authors keep shipping MCP support, not what people are posting.

What I’d do with this: Abstract your model calls behind a single interface today if you haven’t already — switching costs compound fast when you’re mid-product. On MCP specifically, don’t rip it out, but don’t build your core architecture around it until the dust settles in 60 days.