Use an AI agent of your choice to help you solve this task.
Your task, and supporting docs, are deliberately left incomplete and ambiguous, to test your ability to work in a realistic subpar codebase.
You trade for Cannon Capital, a startup prop trading firm.
You have been assigned to a desk that trades a single exchange with 100,000 usd margin. Your job is to expand the desks suite of trading processes. Contracts on this exchange are cash-settled and worth exactly their listed price in usd.
Read TASK.md for the full brief and PROTOCOL.md for the exchange
wire protocol. Below is some info on getting started.
You are given a simulated exchange (a Docker image) that speaks a simple message protocol over NATS.
You need:
- Docker + Docker Compose.
- An x86-64 machine, or emulation: the provided exchange image is
linux/amd64. On Apple Silicon, Docker Desktop emulates it automatically (a platform warning is normal). On an arm64 Linux box, enable binfmt first:docker run --privileged --rm tonistiigi/binfmt --install amd64. Everything you build yourself runs natively either way. - Python 3 if you want to run the taker on the host while probing
(
taker/, see its README) — in the full stack it runs as a container. - The
natsCLI is handy for probing (optional): https://github.com/nats-io/natscli —nats contextdefaults tonats://localhost:4222, which is what the exchange exposes.
./run.sh # NATS + exchange
./run.sh --sim # NATS + exchange + a sample market to trade against(Equivalently: docker compose up / docker compose --profile sim up.) The
first run loads the exchange image; NATS is published on
nats://localhost:4222 for development (handy with the nats CLI).
- Without
--simthe book starts empty — good for controlled probing (you place orders and see exactly what happens). - With
--sima simplistic sample market runs: background liquidity, prices that move around, casual takers.
Stop with Ctrl-C; docker compose down to clean up.
The three trading processes will live in taker/ (provided), strategy/ and
hedger/. Once your code + Dockerfiles are in
place:
docker compose --profile sim --profile strategy up # exchange + sample market + all three seats(./run.sh --sim --strategy is the same thing — exactly what grading runs.)
Inside the compose network every seat reaches the exchange at
nats://nats:4222. While developing you can also run any seat directly on
the host against nats://localhost:4222 — but the final submission must work
as containers from a clean checkout.