Design
We shape the product and the system around the work it has to do.
- Product and system design
- Interfaces in three languages
- Prototype before promise
AI software company
We design, build and operatethe software we ship.
AI systems, platforms and SaaS, from first prototype to production scale, run by people and AI agents as one team.
What we do
Design. Build. Operate.
Build
We build the platform, the agents and the tooling as one system, verified before it ships.
- Web and app platforms
- AI agents in production
- Verification gates before release
Operate
We keep running what we ship, and we measure it where anyone can see.
- Self-hosted infrastructure
- Uptime on a public status page
- People and AI agents as one team
- 8Sites under public watch
- LiveUptime, measured right now
- 19Engineering notes published
- 2Cities, Seoul and Nha Trang
Selected work
Products we run every day.
How we work
From first prototype to a running product.
- 01
Prototype
A working slice, put in front of real users before anything is promised.
Working prototype - 02
Production
Platform, agents and pipeline, built to run and verified before release.
Production release - 03
Operate
We run it, watch it on our status page and step in when it needs a hand.
Live status page - 04
Improve
Operating data feeds the next release, so the product improves.
Next release
Platform
Software that runs itself.
Self-hosted infrastructure, public monitoring and AI agents on shift, so the products we ship keep running after launch.
Self-hosted
We run our own stack and keep the data where we can see it.
Measured in public
Uptime and response times are measured by us and shown on the status page.
Agents on shift
AI agents handle inquiry, booking and dispatch in production, with people in the loop.
Checks before release
Nothing ships until it passes the verification gates in the pipeline.
- 3Open-source repositories
- 2Research papers published
- 3Languages served
Field notes
How we build and run software.
Engineering notes on AI systems, agents and self-hosted infrastructure, written from production.
4 min read
The tells turn over
The words that marked model-written Korean in 2024 are falling in 2026 while a different set rises. Anything built on a fixed list of tells has a half-life.
- ai
- llm
4 min read
Compute the placebo floor first
Any estimator that extrapolates a trend will return something. Running it on years where nothing happened is the cheapest test we know, and the one that most often kills the claim.
- research
- statistics
4 min read
Korean journal abstracts changed in late 2024
We adapted excess-vocabulary measurement to Korean morphology, ran it over 398,296 abstracts, and found a register shift that starts in late 2024 and flattens in mid-2026.
- ai
- llm
Contact
Tell us what you're building.
One email is enough to start. Tell us what you need and when, and we take it from there.