Applied AI

Regen Explorer

Deterministic chain reads first, AI surfaces second

Role
Design and engineering
Year
2026
Status
Built and verified, deployment with the client
13
credit classes read live
60
projects
80
credit batches
2
tiers, one gated

The problem

Chain explorers with a chat box on top let the model answer questions the chain could have answered exactly. Anything a query can settle should never reach a language model.

What we built

Decisions That Mattered

1.

Split the tiers

Deterministic ledger reads run with no key and no model. AI surfaces sit behind a soft email gate. Which tier answers a question is a routing decision made before any inference happens.

2.

Find the endpoints that actually work

The obvious REST base returns 404 for the credit modules and only resolves metadata. Working against the live chain meant mapping which endpoints are singular, which are plural, which return 501, and encoding the answers with a fallback host.

3.

Degrade without breaking

Every AI surface has a documented fallback, so the application builds and runs with no keys present. A missing credential produces a reduced page, never an error page.

What this demonstrates

Knowing when not to use a model, and building the deterministic path first.

Next.js 16Anthropic MCP connectorUpstashZodCosmos REST

Get in touch

Start a Project

Tell us what you are trying to build and what is in the way. We will tell you early if we are not the right fit.

starrybodies@gmail.com