Course 01

Build your own arbitrage bot with an AI agent

Courses where you build something real. You direct an AI agent, learn one idea per day, check the result and finish with a working bot. Paper mode only: public data, no real money.

16 lessonsabout 25 hoursfrom zero

The route

  1. Warm-uplessons 1-4
  2. Workshoplessons 5-8
  3. Whole botlessons 9-12
  4. Releaselessons 13-16

What is this academy

Here you learn by building your own bot. The agent writes the code. You decide what to build, learn the idea behind each step, check the result against numbers you can verify, and save a working version.

01

Practice over lectures

Every lesson ends with a change in your bot and a number you can check by hand.

02

One idea per lesson

One method, one article, one video. No bundles of three algorithms in one piece.

03

Your own pace

No registration, no schedule. Skip what you know.

Course 01

Build your own arbitrage bot

Build a working bot in paper mode with an AI agent, day by day. Each lesson teaches one idea from zero, then you give the agent a task, check the result and save a version. 16 lessons, 4 worlds.

Who it is for

A high-school student, a beginner developer, a trader. No background needed and no programming: the agent writes the code. Public exchange data only, read-only, no API keys, no real money.

What you can do at the end

Your own two-stage arbitrage bot in paper mode with a dashboard and a config file. Stage 1 finds candidates by best bid and ask, stage 2 walks the order book to see how much money can pass. You can explain every part and its limits.

Primersoptional

+1

Logarithms in 10 minutes

Why rates multiply and logs add. Needed before lesson 6.

Article
+2

Graphs from zero

Nodes, edges, paths and cycles. Needed before lesson 5.

Article
+3

Mean and standard deviation

The two numbers behind every z-score. Needed before lesson 15.

Article
+4

Vectors and matrices

Lists and tables of numbers, and numpy broadcasting. Needed before lesson 13.

Article
+5

Probability and the bell curve

Probability as a countable share, and why a z-score of 2 is rare. Needed before lessons 11 and 15.

Article
+6

Reading Rust in ten minutes

Variables, loops, vectors and functions in Rust. Needed before lesson 13.

Article

Warm-updays 1-4

01

First scanner in an hour

The agent writes a scanner for two exchanges, you run it and read the gap.

Article
02

Kinds of arbitrage and your bot's type

Cross-exchange, cyclic, statistical, futures-spot. You choose the type of your bot.

Article
03

Stages and the bot's design

Screen, simulate the fill, execute. The agent drafts the design, you review it.

Article
04

Compare best prices

Spread, fees, units. Stage 1 of your bot works on 5 venues.

Article

Workshop: finding candidates fasterdays 5-8

05

Rates as a graph

Currencies are nodes, rates are edges. Your bot builds a rate graph from live quotes.

Article
06

Bellman-Ford from zero

A negative cycle is a profitable chain. Your bot finds one.

Article
07

SPFA from zero

The lazy queue version. Compare its speed on the same data.

Article
08

RICH from zero

What it improves and when it pays off. Benchmark on a bigger graph.

Article

A whole botdays 9-12

09

Reading and walking an order book

Emulate a buy and a sell through depth, leg by leg.

Article
10

Volume, fees and slippage

How much money can pass, and where profit disappears.

Article
11

Reality check

Latency, phantom quotes, gas. A replay tester with recorded data.

Article
12

Paper trading, config and risk limits

A loop that scans, simulates, trades on paper and stops at limits.

Article

Releasedays 13-16

13

Speed: vectors, matrices, Rust

Rust for the hot path of your bot.

Article
14

Futures-spot and funding as a second strategy

Basis, funding rate, carry.

Article
15

Your own extension: pairs trading

Two coins on a leash: spread, z-score, a backtest and what fees do to it.

Article
16

Release: show your bot

Paper mode with a dashboard, a README, a public repository.

Article

Questions

Do I need to know trading?

No. Every term is explained before it is used.

Do I need to code?

No. The agent writes the code. You read the output and decide what to change.

What do I need?

A computer and an AI agent subscription (Claude Pro or ChatGPT Plus, about $20 a month). The bot uses public data and never real money.

Which languages?

English, Russian, Chinese (simplified and traditional): articles and videos.