Skip to content

Case study 05 · AI-native editorial system

Giving AI a system to think and build with.

ATLAS layers editorial intent on top of SYX so AI-assisted long-form experiences can vary creatively without abandoning technical governance.

Role
System designer
Mental model
ATLAS
Execution
SYX
Focus
Editorial identity · motion · accessibility

Explore repository ↗

Live product evidence

Same system. Two radically different editorial registers.

ATELIER and OBSIDIANA are rendered directly from the published GitHub Pages experiences. The previews are intentionally non-interactive here; each project can be opened separately to explore the full motion and scroll behaviour.

ATELIEROpen live ↗
ATELIERLive preview unavailable — open project
Cinematic profilesHuman · editorial · motion-led
OBSIDIANAOpen live ↗
OBSIDIANALive preview unavailable — open project
Forensic dossiersData · investigation · controlled tension
ATLASEditorial intent
ATELIEROBSIDIANA
SYX · governed executiontokens · accessibility · motion patterns · architecture

Problem

AI can generate pages before it understands the system.

Generic prompting can produce code, but it does not automatically understand editorial intent, identity rules, motion language or the architectural constraints of a design system.

Separation of concerns

ATLAS decides what the experience should feel like. SYX decides how it is built.

ATLAS acts as the editorial mental model. SYX acts as the technical executor. That split keeps creative direction distinct from implementation rules.

Two registers

A system should create variation without losing governance.

ATELIER is designed for cinematic human profiles; OBSIDIANA for forensic economic dossiers. They deliberately differ in typography, palette and motion while sharing the underlying system and accessibility expectations.

Knowledge, not prompts

Reusable motion patterns become part of the system context.

The project documents reusable motion patterns with narrative intent, parameters, GSAP controls, prompts and reduced-motion fallbacks. That knowledge can be loaded by specialized SYX modes instead of rediscovered in every task.

Takeaway

The differentiator is not generating more. It is governing generation better.

This project shows how I am evolving from using AI as an assistant to designing the context, knowledge and rules that allow AI-assisted work to remain coherent.

Open ATELIER ↗Open OBSIDIANA ↗Explore repository ↗