Not a tracker. Not a video library. A structured knowledge graph of martial arts technique — mechanical breakdowns, objectives, checkpoints, drill blueprints. The kind of tool that does not exist because building it requires someone who both understands the discipline deeply enough to define correct movement and can deploy the infrastructure to encode it. Those two people are rarely the same person.

In this case, they are. The builder trains in the discipline. The AI crew builds the software. The instruction layer — knowing what a correct hip escape actually looks like, what “progression” means when you are eighteen months into guard retention — is not something the model supplies. It is something the builder already held. The AI had no opinion on whether the breakdown was right. The builder did.

Domains secured. MVP-Beta operational. Knowledge graph in active enrichment. No prior software engineering background involved. None required.

A web-based multiplayer implementation of Liar’s Dice, integrated with the Order of the Kraken NFT collection on Cardano. Fifteen pirate personalities. Thirteen bloodline abilities. Cosmetics economy. Wallet integration. The entire codebase — front-end, game logic, server infrastructure — was built using AI coding agents. No external development team. No agency. No contract developers.

The human principal wrote the specification, made the architecture decisions, evaluated whether the output was correct. The execution was delegated entirely. This is what full application development looks like when AI operates as build infrastructure rather than as an assistant making suggestions in a sidebar.

Live prototype deployed. Wallet integration in progress. Smart contract wagering in design.

Four cloud-hosted storage nodes on the Iagon network. Approximately 270TB committed. An AI intelligence patrol runs weekly, aggregates performance data, and alerts on anomalies. The nodes have been operational since 2025.

The economic thesis is straightforward: does the token yield justify the carry cost? That evaluation belongs to the human principal. The monitoring and anomaly detection do not. Reward yield is active. Network adoption is early phase. The bet is on where it goes, not where it is.

This publication. Five specialist AI functions under one human principal. Total infrastructure cost under $1,500 per year. The operating model described in the dispatches is the operating model that produces them.