Design System
A token system used across an organization's product surfaces.
Compresses the time required to build a new screen, and continues to be maintained by the in-house team without our involvement.
We build
structures that
operate.
A studio is not measured by what it makes,
but by what continues to operate after it leaves the room.
.02 / Systems
Four kinds of systems we build
Each is delivered as an operating system, not a deliverable. It is meant to be used by people who were not in the room when it was made.
-
01
Design systems
that scale across teams.Tokens, components, editorial rules, and the operating documentation around them. Tuned so that an in-house team can extend, revise, and audit without our involvement.
-
02
AI workflows
that act, not assist.Agents, prompt libraries, evaluation harnesses, and human- review loops — assembled so that a measurable share of operational work runs end-to-end without supervision.
-
03
Product architectures
that hold their shape.The information model, the interface logic, and the decision rules. Built so the product can be re-skinned, re-priced or re-organized without rewriting its core.
-
04
Knowledge structures
that remember.Internal documentation, taxonomies, and retrieval systems. Treated as organizational memory, not content marketing — the kind of asset that compounds when people leave.
.03 / Method
How a system is built
Four movements. Diagnosis comes before design, and operation comes before launch.
-
M.01
Diagnose the operating problem. Most projects begin one layer too high. We start by locating the actual constraint — the team, the workflow, or the decision that the system will need to support every week, not just at launch.
-
M.02
Design the structure, not the surface. We build the model first: tokens, taxonomies, decision rules, role boundaries. The visible interface is the last layer — and it is the easiest to change because the structure underneath is sound.
-
M.03
Compose with AI as a co-operator. AI is not a feature we add. It is a working member of the system — assigned tasks, given guardrails, evaluated like any other operator. We design the points where it acts and the points where it must hand off to a person.
-
M.04
Hand over a system that runs. Delivery is not a file or a deck. It is a system that an in-house team can operate, extend and audit. Our last act is to make ourselves unnecessary.
.04 / Outputs
What our systems produce
We measure outputs in operating capacity — not in pixels shipped or hours billed.
AI Workflow
A triage system that absorbs a team's regular inbound volume.
Operates end-to-end with a single human reviewer at the routing layer.
Product Architecture
An information model that survives full re-skins.
Same core, multiple brands, no data migration.
Knowledge Structure
A retrieval system that replaces a team's onboarding documents.
New hires reach independent operation noticeably earlier, and the system stays current because the editorial cost of keeping it that way is low.
.05 / Position
How Sinh Vũ works differently
The difference is not aesthetic. It is structural — a different stance on what a studio is for.
.01 / Stance
Surface-first: design the look, then figure out the system.
System-first: design the structure, and the surface follows.
.02 / AI
AI as a feature added to the product after launch.
AI as a co-operator inside the system from day one.
.03 / Time
A one-time delivery, then a maintenance retainer.
A system designed to run and evolve in your hands.
.06 / Invitation