They moved from prompting AI to leading an AI team that built a trading
A working AI-powered trading workstation with simulated live data, portfolio tracking and an AI Copilot, built by a coordinated team of AI agents.
This student set out to build something ambitious: a full trading workstation with live simulated markets and an AI Copilot that understands plain language. They did it by directing a team of specialised AI agents, reviewing every step themselves. Their story shows what becomes possible when you stop just prompting AI and start engineering with it.
FinAlly
Before
Before this project, their relationship with AI was about prompting: asking for code and seeing what came back. A full trading application, with a frontend, backend, database, live data and an AI assistant, is a lot of moving parts for one person to hold. The challenge was turning fast AI output into something that actually worked end to end.
“I learned how to move from prompting AI to engineering with AI.”
In their own words
The turning point
In the program they adopted a vibe-coding and multi-agent engineering workflow. Instead of one assistant, they set up a coordinated team of frontend, backend, database, LLM, DevOps and integration-testing agents. Their own role shifted to guiding, reviewing outputs, approving actions and testing by hand, with Git branches and checkpoints protecting each working version.
“I guided the process, reviewed outputs, approved actions, tested the application manually, and used Git branches and checkpoints to protect working versions.”
In their own words
After
They now have FinAlly, a working workstation that combines simulated live market data, portfolio management, manual buy and sell execution, and P&L tracking. Its AI Copilot can analyse a portfolio, recommend diversification, add symbols to a watchlist and carry out approved multi-step trading actions. They built it with tools like Claude Code Agent Teams, Next.js, FastAPI, Python and Docker.
What it lets people do
FinAlly lets people practise trading in a simulated live market, track their portfolio and profit and loss, and place trades by hand. They can also simply ask the Copilot in plain language to review their holdings or take a series of actions, which only run once approved.
What they learned
Their biggest lesson was the move from prompting AI to engineering with AI: context engineering, splitting work into specialised agent roles, testing continuously and using Git as a safety system. Above all, they learned that AI can generate large amounts of code quickly, but a human in the loop is what makes it trustworthy.
“I also learned that AI can generate large amounts of code quickly, but human verification remains essential.”
In their own words
FinAlly was built during AI Mastery.
Read the full submission
Written 9 August 2026 Submitted by the storyteller
Every story is written by a Mindvalley student or by someone our teachers have worked with. How we collect and check stories
More from AI Mastery
