What's cooking?

This project grew out of a very practical problem: AI makes it easy to create a lot of useful work, and surprisingly easy to lose track of what that work has become.

There's already real competition for the shallow version of this: tools that scan a ChatGPT or Claude export and surface forgotten threads and abandoned ideas. That part of the problem already has answers, and we're not interested in rebuilding one of those.

What those tools don't do is treat AI-assisted output as an actual body of work. Not a report you run once on an export and read, but a living record that stays with the work as it moves from conversation to draft to shipped product: the guide that started as an idea, the framework buried inside a long chat, the draft that became a product, the unfinished project worth returning to. Cross-tool and ongoing, tied to the actual artifacts rather than reconstructed from a one-time chat dump.

We're also exploring how much of that upkeep AI itself can carry, so the record stays current without becoming one more thing only you remember to do.

So the bigger question is not just where did I save that? It is: What have we built, where did it come from, what belongs together, what is still unfinished, and what could we do with it next?

This system began as a way to use AI like a capable creative crew without handing over authorship, voice, continuity, or control. The tools have moved fast since we started, so we're taking another look at the system with an even stronger focus on human direction, judgment, creative decisions, and authorship.

Part of how it does that is a simple three-tier way of naming how much of a given piece came from the writer versus the AI, and what evidence shows the writer stayed in control. Tier 1 — Human Authored / AI Supported: the writer creates the prose; AI helps with research, brainstorming, organization, spelling, and grammar. Tier 2 — AI Assisted: the writer writes or substantially rewrites the work, while AI contributes suggestions, alternatives, critique, or limited passages. Tier 3 — AI Directed: the human owns the concept, structure, characters, decisions, and final approval, while AI generates significant portions of the prose under that direction. This is a working tool for the writer's own clarity, not a public certification mark: the question underneath it is what did AI do, what did the writer do, and what evidence shows the writer remained in control?

The final stage of the system is the Writer's Publication Readiness System: a five-stage final review, covering version control, continuity, prose, cold-reader review, and final handoff, that moves a manuscript from draft-complete to reader-ready. It replaces scattered rereading and last-minute fixes with a visible, repeatable process. AI can help identify gaps, inconsistencies, and potential problems, but human judgment remains the final gate. The two systems are being reconsidered together as one pipeline, and the second half is built to answer one question: Is this work truly ready for its next reader?

This educational project began with working musicians and has grown into a larger conversation about creativity in the age of AI: authorship, provenance, identity, rights, human value, career resilience, audience trust, and the qualities technology can imitate without actually living them.

There are already learning materials, student activities, educator resources, and pieces of a program that could live in classrooms, workshops, and other learning environments. Now it needs more real creative voices in the room.

This project sits alongside the tiering idea in the Writer's AI Director System, but asks the question from the creator's side rather than the reviewer's: once AI can draft, remix, and imitate at scale, what's left that's actually yours? The question isn't whether AI touched the work. It's whether something uncopyable, something that could only have come from you, made it in.

The Uncopyable Source gives creators a Source Card built from five inputs no AI can generate on its own: memory, observation, experience, contraband (the messy, half-legal influences everyone actually draws from), and taste. The guide walks through naming your own sources, feeding them into AI-assisted work on purpose, and ending up with something that still sounds like you.

This concept started with a simple observation. AI-created music artists are building real audiences and real catalogs, but most have no path to live performance, which is where most of an artist's income actually comes from.

Major labels have already proven the core mechanism works. Companies behind virtual K-pop groups cast real performers to embody the persona on stage, and those groups are already touring. That's not a dead end for this concept, it's validation that the mechanics hold up. What those productions have in common is a full label's budget and infrastructure behind them.

Synth Stage Studio is narrowing to the gap that leaves open: independent and mid-tier AI music artists who have the songs and the fans but not a production company. Same core idea, professional performers cast to embody an artist's persona live, a structure for booking and managing those performances, and a licensing model, just built to be adoptable by a small agency or an independent artist's own team rather than requiring major-label resources.

We do not know yet whether that becomes a licensing framework, a boutique agency, a toolkit artists set up themselves, or something else. That is the question we are bringing here now.