I travel a lot, and every time I did, my trade execution nosedived. Missed trades, missed invalidations, and no real idea what my portfolio was doing. In the video, I break down the always-on setup behind my AI agent, connecting Bybit and Alpaca, the memory folder, FMP and Quiver, and the Telegram bot that ties it all together. Then I run it live from the beach, which is the fun part.

Most people still use Claude like a chatbot, so when you ask it about the market it's just scraping the internet. It never sees the actual filings. In the video, I connect Claude Code to Quiver, run the prompts I use on insider, congressional and lobbying data, and screen the whole market for stocks where several executives are buying at once. Then I build a Telegram bot that pings my phone when something happens, which is how I actually use it day to day.

Claude can take five seconds to make a trading call. Jev makes one in around 400 milliseconds, so I've split the job between them. In the video, I show you how the strategy engine makes every call, how I find strategies worth running, and the news engine I built that reads headlines as they land. Then I connect it to Alpaca and Bybit, plus the manual approvals I still use on bigger trades.

Warren Buffett's investing framework can now run as an AI agent. So I set one up in Claude and gave it $100,000 to manage. In the video, I set up the free AI Hedge Fund repo, ask Buffett about Apple, Nvidia and a full watchlist, and backtest his picks against the S&P 500. Then I let him rebalance my Alpaca paper account, which is where it gets real.

Building your own trading bot from scratch is a lot of work, and you don't always have to. Professional developers have already built them and put them on GitHub for free. In the video, I go through the three best repos I found for actually executing trades, set each one up with Claude Code and run them live, and say when you're still better off building it yourself.