Question 1

What is Connect Tycoon?

Accepted Answer

An open-source reinforcement-learning simulator from wonseokjung in which a model operates a company, making one decision at a time and learning from hundreds of simulated runs. It was built for a CONNECT AI LAB x AI CITY BUILDERS class and is Apache-2.0.

Question 2

How is Connect Tycoon connected to Jev or System One?

Accepted Answer

The repository description lists JEV, PPO and human-in-the-loop, and the README describes the core model the Jev way: given a situation, a question and options, it answers with a choice and probability or confidence rather than generating text, consulting a human when unsure.

Question 3

How do I run Connect Tycoon?

Accepted Answer

The digest shows the project ships as a simulator program in the repo plus Colab notebooks, including one that uploads your own RL model in a minute. The digest truncates the install section, so the exact pip or clone commands were not captured and should be read from the README.

Question 4

Do I need a GPU or an API key to try it?

Accepted Answer

Nothing in the captured text mentions a GPU, an API key or a paid service; the game runs locally in Python with Pygame, and the Colab notebooks are the hosted path. The README's requirements section was cut off in the digest, so this is not fully confirmed.

Question 5

What does the model actually learn?

Accepted Answer

Per the README, decisions about running the company return as money, customer satisfaction and incidents; repeated simulation runs let the model learn which choices lead to better outcomes, and uncertain moments are routed to a human instead of guessed.

Question 6

Who built Connect Tycoon and what licence does it use?

Accepted Answer

It is a project from wonseokjung, created 2026-10-11 and pushed 2026-10-17, with 7 stars and 15 forks. The repository is Apache-2.0 and the README credits the CONNECT AI LAB channel and AI CITY BUILDERS.

Question 7

What are the limits of this project?

Accepted Answer

It is a small game-style simulator, not a production decision service: the digest shows no benchmark numbers, no accuracy data and only a partial walkthrough, so the README's own claims about learning behaviour have not been independently checked.

Connect Tycoon

Highlights

Questions about this entry

From the Jev guides

From the Jev AI model community