Yumes

Yumes Corporation

A small team that works like a large one.

We're an AI-native software company. Agents work alongside us at every step, from planning and research to code and review, so a few people can take on work that used to need a department. Right now that work goes into two products: CrossPlug and Transit Time Map.

Building
CrossPlug
Live
Transit Time Map
Home
yumes.net

01 — Products

What we build

Two products so far. One is out in the world; the other is on its way.

CrossPlug

Your agent runs the ads.
You say one sentence.

An ad marketplace for solo founders who ship apps with AI. Founders show each other's ads to earn credits, then spend those credits to put their own app in front of new users. Their coding agents handle the setup, the copy, the buying and the tuning.

Pre-launch · waitlist open Visit CrossPlug
Transit Time Map

See a whole city by travel time.

A public transit travel-time map for six Korean cities: the Seoul Capital Area, Seoul, Busan, Daegu, Daejeon and Gwangju. Pick a starting point and the city is colored by how long it takes to get anywhere, computed from official timetables. Tap a destination for the route, transfers and walking included.

Live · Korean Open Transit Time Map
Transit Time Map overview: travel-time maps of six Korean cities

02 — What we believe

The next user of your product might be an agent.

More and more, people ask a coding agent to do the work and only review the result. We design for that agent first, and keep people in charge of what matters.

a

Structured answers, not screens

An agent parses a response and explains it to its person. It doesn't need a dashboard to click through.

b

Small tools, not big features

We expose simple primitives and let the agent combine them. Things like A/B tests and budget splits become the agent's job, not ours.

c

People hold the money

Agents act on their own only within limits a person signs for. Anything that moves cash asks a human first.

03 — How we work

Agents on the team. People on the decisions.

  1. 01

    Agents in every loop

    Planning, research, writing, code and review all run with AI agents in the loop. People set the direction and make the calls.

  2. 02

    Small on purpose

    We grow output with leverage, not headcount. A small team keeps decisions fast and context in one place.

  3. 03

    Specs as the source of truth

    Every product starts as a written spec that people and agents can both read. When the spec and the code disagree, we fix one on purpose.

  4. 04

    Ship to learn

    Building is the experiment. We ship early, measure with real users, and let the results decide what comes next.