Lineage artwork for Workshop — Lineage visual · lineage texture platform workshop

Workshop

Workshop

The room for problems that resist a single pass. Workshop runs structured attempts, eval scores them, memory keeps what worked, and the problem gets genuinely easier each session.

Workshop anything

A scene that will not land. A character voice that drifts. A campaign architecture. What moved in your field this month. If it is hard and it matters, it belongs in the workshop.

On demand, on the models you already have

The Auteurs: Powers #1, Workshop

Which AI should write your opening scene? With Workshop, coming to Auteur, two or three models you already use compete on your task. You see the estimate first and set the cap, a judge model explains its pick, and the final call is yours.

Transcript

Which AI writes the best opening scene? Claude? GPT? The model you run at home? Most people just… flip a coin. With Workshop, you pick two or three AI models you already use… and they compete on your task. It runs on the subscriptions and keys you already have. You see what it will cost first… and you set the cap. It runs when you say… go. A judge model reads every draft… and tells you which is strongest, and why. Side by side. No guessing. But the final call is yours. Accept the winner… and it goes straight into your project. Stop guessing which AI to use. Let them compete… and you pick the winner. Workshop. Coming to Auteur. Be your own auteur.

Captions (.vtt)

You start a Workshop when a task needs more than one pass. Two or three contender models take it on, chosen from the AI subscriptions and keys you already have: in practice your Claude and ChatGPT or Codex subscriptions, plus any API keys with known prices. Before anything runs, you see an estimate of the cost or messages it will use and set a cap for the run.

“Workshop this scene. Claude against Codex, a judge to pick, and cap it at twenty messages.”

Side by side, your call

The attempts come back side by side, with a judge model's pick and the reasoning behind it. You accept the winner and it goes into your project; nothing enters the project until you do. Workshop runs when you ask and never changes your models on its own.

Powered by eval and memory

Workshop is where the other two pillars compound: eval measures each attempt, memory carries the learning forward, and the third session starts where the second left off, smarter.

Powered by the Algernon Engine, in every product

Memory

The project remembers what the chat window forgets, approved canon kept separate from suggestions, on every surface, under any model. It travels with the project, not the app.

Go deeper → ⊚

Eval

Work is judged, scored, and receipted instead of accepted on vibes, evidence before acceptance.

Go deeper → ⌘

Workshop

Hard problems, on demand: two or three of your own models attempt it within a cap you set, a judge explains its pick, and you decide what enters the project.