Vue FesJapan 2026
The main visual of Vue Fes Japan 2026. It features a silhouette of a V, representing Vue, and a circle, representing the Japanese flag, side by side. The colors are deep matcha-like shades derived from the two colors, and the circle features a Japanese gold marbling pattern.
OtemachiPlace Hall & Conference

Speaker

MedPeer Track
15:05 - 15:35
Keisuke Tsuji

Deterministic Frontend Architecture Is the Way to Go

"Should we put this in the xxx directory?" "Should we extract this into utils?" "Should we turn this into a composable?" "Can we write this with computed instead of watch?"

These are the kinds of design discussions that come up again and again in frontend development.

As generative AI keeps accelerating development speed, is it really realistic to keep making these judgment calls one by one?

In a new product I'm involved with as a frontend engineer, built on Nuxt, we've decided as deterministically as possible where files and functions should go, how things should be named, and how APIs should be used, and backed all of it with linter guardrails.

As a result, we've reached a point where most PRs can be reviewed in just a few minutes, and we've built a setup where people can develop productively even without deep frontend expertise, giving us a development environment that's both robust and fast.

In this session, using the architecture of a Nuxt project currently in development as an example, I'll introduce the concrete decision rules we've established in the field to build an architecture that leaves no room for hesitation.

Keisuke Tsuji

Title
Software Engineer

Keisuke Tsuji

Title
Software Engineer

Deterministic Frontend Architecture Is the Way to Go

"Should we put this in the xxx directory?" "Should we extract this into utils?" "Should we turn this into a composable?" "Can we write this with computed instead of watch?"

These are the kinds of design discussions that come up again and again in frontend development.

As generative AI keeps accelerating development speed, is it really realistic to keep making these judgment calls one by one?

In a new product I'm involved with as a frontend engineer, built on Nuxt, we've decided as deterministically as possible where files and functions should go, how things should be named, and how APIs should be used, and backed all of it with linter guardrails.

As a result, we've reached a point where most PRs can be reviewed in just a few minutes, and we've built a setup where people can develop productively even without deep frontend expertise, giving us a development environment that's both robust and fast.

In this session, using the architecture of a Nuxt project currently in development as an example, I'll introduce the concrete decision rules we've established in the field to build an architecture that leaves no room for hesitation.