Build boldly. Keep it revertable.

Riu Salze is a Catalan-inspired identity that reinterprets the sound and meaning of my Korean name through a different linguistic lens.

Riu, Catalan for “river,” naturally echoes the sound of my surname, Ryu, while also reflecting the water imagery carried by my given name. Salze, meaning “willow,” mirrors the imagery associated with my Korean surname.

Rather than being a direct translation, Riu Salze is a linguistic reconstruction — preserving the sound, imagery, and identity of the original name without reproducing it literally — more or less.

Observe the structure.
Understand the structure.
Declare the structure.

Software engineer focused on system structure, data flow, and AI-assisted development architecture.

I treat programming languages and frameworks as implementation media. My primary interest lies in how systems are framed, observed, verified, and evolved.

I specialize in turning complex legacy systems into clear, observable, and reusable architectural flows.

I treat human language as a medium for design, collaboration, and execution. With AI — or whatever comes after it — I build across linguistic and geographical boundaries.

The work travels better than I do, though — if you want to find me in person, you’ll still have to come to Seoul.

🔭 Cosmic Horizon is where I document my engineering notes, development philosophy, and AI-assisted development models.

Not reading it is also an option. For now.

A viewer is only the surface. What matters is what survives after the session ends.

🛰️ A local desktop tool for reading Codex CLI JSONL sessions and exporting versionable worklog bundles.

A Codex session is also a trace of your reasoning — the assumptions you made, the directions you chose, the corrections you introduced, and the decisions that shaped what came next.

AI produces results within that path, but the path itself is yours.

Export Worklog is my way of preserving that path. It turns a Codex session into structured records grouped by [YOU] request boundaries, so the work can survive beyond the chat window as something reviewable, understandable, and versionable.

The feature itself is deliberately simple. You could probably build something like it too.

But are prompts and outputs really just disposable means to an end? I don't think so.

In AI pair programming, the code is only part of the work. Prompts, decisions, corrections, tool calls, failed approaches, and verification traces all help explain how that code came to exist. I believe that process deserves to be treated as an engineering record, not just temporary conversational exhaust.

That idea comes directly from Cosmic Horizon: observe the work, ground it in evidence, and leave enough context behind for the next person — including your future self — to understand what actually happened.

• Built with Rust, Tauri, Svelte, and TypeScript
• Reads local Codex CLI JSONL session logs
• Exports worklog bundles grouped by [YOU] request boundaries
• Preserves AI-assisted decisions and verification traces as reviewable records

You use Codex CLI, right? Still reading raw JSONL without a viewer? Bold choice. It’s open source, by the way. The price is unlikely to get better. Time is money, friend.

🧩 An earlier Kotlin/Swing implementation for reading local Codex CLI JSONL session logs.

Codex Chat Viewer was my first concrete step as a product engineer in this direction. Instead of just complaining that reviewing local Codex logs was inconvenient, I started building a tool to make the work easier to inspect, search, filter, and export.

It remains as the predecessor to Codex Session Observatory.

• Built with Kotlin/JVM and Swing
• Focused on local-first Codex session log viewing
• Turned raw JSONL logs into readable chat-style transcripts
• Started from a direct workflow friction I experienced while using Codex CLI
• Preserved as the earlier implementation before Codex Session Observatory

• Email: riusalze [at] gmail [dot] com
• LinkedIn: Riu Salze
• Cosmic Horizon: Engineering Notes — Wait... why are you here?

See? I did say “For now.”

Alright. Time to revert.