Lineage artwork for Eval — Lineage visual · lineage texture platform eval

Eval

Eval

Nothing here is accepted on vibes. Eval is the studio's judgment layer (scoring, receipting, and remembering), so quality is measured instead of assumed.

Judged before accepted

The Auteurs: Powers #3, The Glitch

A glitch plants fake facts and false "all good" checks. The Night Shift catches it with canon audits, a reviewer from a second model family, the claim ledger and an optional attached classifier.

Transcript

Somewhere in your project… a lie is hiding. The Glitch. He plants a fact that never happened… a bug that swears it's fine… a chapter that forgets chapter one. Alone, your AI won't catch him. It wrote the mistake. So I carry the whole belt. Canon audits on every chapter. Green passes. Red gets flagged. A second AI, from a different family, reviews the work… and reports what it finds. Claims without proof get blocked before they go live. And a smart, fast classifier you attach, like Jev, can check every scene… one fact at a time. The build stays true. The manuscript stays true. Nice try, Glitch. Auteur Intelligence. Be your own auteur.

Captions (.vtt)

Drafts, builds, shots, and answers are scored, often by a different model family than the one that made them, before they touch the project. The checker is never the builder.

Receipts, always

Every attempt leaves a receipt: what ran, what it cost, what was judged, what happened. Failures are kept, classified, and useful: you always know how a seat performed.

“Score tonight’s three drafts against my last accepted chapter and tell me which one earns a read.”

How eval looks at a picture

A render request never returns one take on faith. Candidates come back and are scored against a rubric (does the face match the character the project knows, does the composition serve the shot, did the image actually follow the prompt, is anything artifacted), and the winner is accepted with its scores on the record. The rejects keep their scores too, which is how the next request starts smarter.

How eval powers the workshop

Eval is the workshop's compass. Every Workshop run ends in a judged pass: the contender attempts sit side by side with the judge's pick and its reasoning, you accept the winner or turn it down, and both outcomes land in memory. Scored attempts and remembered verdicts are what the next session starts from.

The ledger remembers models, too

Eval does not only judge the work; it keeps book on the workers. Every model you run builds a track record from its receipts: what it delivers, what it fumbles, how long a finished task takes and how many tokens it spends. Failures are kept and classified, not hidden. That record is the model's dossier, and it is why trusting an inexpensive open-source model here can be a measured decision, not a leap of faith: it has a scorecard, and you choose with it in front of you.

It learns your taste

Eval outcomes feed memory, and memory shapes the next run. The system gets measurably better at your preferences over time, in code, prose, or picture.

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.