Why I Built a Terminal-Native Note-Taking App
Why I Built a Terminal-Native Note-Taking App
I spend most of my day in a terminal. Compiling, grepping, SSHing into boxes, reading logs. The terminal is fast, composable, and completely under my control. So it always struck me as odd that the moment I wanted to write something down, I had to leave it — open a browser tab, wait for Electron to load, sign in to a cloud account, and hand my notes over to someone else's server.
I tried everything. Notion is beautiful but requires a login and syncs to the cloud. Obsidian is closer — local files, Markdown — but it's Electron, which means it's a web browser wearing a trench coat, consuming 200 MB of RAM to display plain text. Bear, Roam, Logseq — all excellent, all built for a different person than me.
The straw that broke it
One afternoon I was SSH'd into a production server diagnosing an incident. I wanted to capture what I was finding in real time — commands I ran, output that mattered, hypotheses I was testing. My note app was on my laptop. My terminal was on the server. Copying between them was friction I didn't need mid-incident.
I wanted to type syahi and have notes appear right there, in the same terminal, with no round-trip to a browser.
What I actually wanted
- Start in under 100ms
- Works over SSH
- No internet connection required
- Notes stored locally as SQLite — one file I can back up, version, copy
- Full-text search that works instantly, not after a 30-second index
- A real editor with find, replace, undo, and selection — not a textarea
- Templates for the things I set up repeatedly (sprint planning, incident runbooks, meeting notes)
- Completely offline AI when I want it — not a call to OpenAI
None of that required a browser. All of it could be done in C.
Building it
Syahi started as a weekend experiment: a raw-termios REPL over a SQLite backend. A week later it had a full-screen editor. A month later it had templates, FTS5 search, and recording. Six months later it had a local AI inference engine — two quantised models embedded directly in the binary, no Python, no API key, no internet.
The binary is about 2 MB. It starts instantly. It runs on Linux, macOS, Windows, Android, and iOS from the same C11 codebase.
Who it's for
Syahi is for developers who live in the terminal and don't want to leave it just to write something down. It's for researchers who want semantic search over their notes without sending them to a cloud. It's for teams who need seat-managed licensing without giving everyone another SaaS login.
If that's you — download it and run syahi in your next terminal session.