Claude Just Changed the Stock Market Forever

ELI5/TLDR

A former JP Morgan guy shows how to use Claude’s desktop app to automatically trade stocks through Alpaca’s API — no coding required, just talking to it. He walks through three strategies: a trailing stop bot that protects your downside, a copy-trading bot that mirrors US politicians’ stock trades (which, yes, consistently beat the market), and the wheel strategy, where you sell options and collect premiums like an insurance company. Everything runs on paper money, which is fake money in a real market.

The Full Story

The Gap Between Wall Street and Everyone Else

Samin Yasar spent the early years of his career at JP Morgan, where he learned that the distance between institutional traders and regular people comes down to three things: data, execution, and intelligence. Wall Street knows when a billionaire makes a massive bet. They know when a senator buys shares before an announcement. They have systems placing trades around the clock. And they have teams turning all that information into decisions.

That access used to cost hundreds of thousands of dollars a year. The pitch here is that Claude, combined with a few free tools, closes that gap.

Setting Up: Claude Meets a Brokerage

The setup is straightforward. You need the Claude desktop app (pro or max plan) and an account on Alpaca, a brokerage that offers free API access. A brokerage, for the uninitiated, is just the middleman between you and the stock exchange — it used to be a person you called on the phone, now it’s an app.

On Alpaca, you create a paper trading account (fake money, real market prices), generate API keys, and paste the endpoint, key, and secret into Claude. Then you test it:

“Can you please buy one share of Apple? I want to see it inside my account.”

It works. Claude places the order through Alpaca’s API. You can now trade stocks by talking.

Strategy 1: The Trailing Stop Bot

The first bot is a trailing stop — a strategy where your sell floor rises with the stock price but never falls back down. Think of it like a ratchet.

You buy a stock at $100 and set a floor at $95. If it drops there, Claude sells. If it climbs to $110, Claude drags the floor up to $105. You can never lose more than a fixed percentage, and your gains get locked in as the price rises. If the stock drops immediately, you take your small loss and move on.

“The worst thing you can really do is hand your AI a pile of money and be like, ‘Go figure it out.’”

The key insight: you provide the rules — your risk tolerance, your read on the market — and Claude executes them with speed and discipline you couldn’t manage yourself. It checks every 5 minutes during market hours, adjusts stop losses, and handles re-entry. He also adds “ladder buys,” where Claude buys more shares at progressively lower prices if a stock drops 15%, 20%, 30%, averaging down instead of just bleeding.

Strategy 2: Copy Trading Politicians

This is where things get interesting. Members of Congress are required by law to disclose their stock trades. The data shows many of them consistently beat the market, which is not particularly surprising given they sit on regulatory committees and get briefed on policy changes before the public hears about them.

A free service called Capitol Trades scrapes and organizes this data. Claude connects to it through an MCP — which is essentially a plug that connects Claude to an external data source, the way a power outlet connects an appliance to the grid.

Claude picks a politician to copy. In this case, it chose Michael McCaul, who was the most active and top-performing trader at the time. The trades are slightly stale because of disclosure delays, but Congressional trades tend to be longer-term positions — they’re not day trading. The backtested claim: copying McCaul’s trades over the past year would have returned 34.8% on a $50k account, compared to 15% for just holding the S&P 500. A 2.2x outperformance.

Strategy 3: The Wheel (Becoming the Insurance Company)

The final strategy requires understanding options, which Yasar explains through an analogy that actually works.

An option is insurance on a stock. You pay your car insurance company $100 a month. If you crash, they cover it. If nothing happens, they keep your money. Stock options work identically: you pay a premium for the right to buy or sell a stock at a fixed price by a certain date. If the stock moves in your favor, you exercise the contract. If it doesn’t, the other side keeps your premium.

There are two types. A call option is the right to buy at a locked price — like putting a deposit on an apartment to lock in the rent. A put option is the right to sell at a locked price — like insurance against a price drop.

Here’s the twist: you can be the insurance company. When you sell options, people pay you premiums. Most contracts expire without anything happening. You keep the money. Insurance companies make billions this way.

The wheel strategy chains this together in two stages:

Stage 1: Sell puts. Tesla is at $250, but you’d buy it at $230. You sell a put at the $230 strike and collect $500 in premiums. If Tesla stays above $230, the contract expires, you keep the $500, and do it again next week. If Tesla drops below $230, you have to buy — but your effective cost is $225 because of the premium you already collected.

Stage 2: Sell covered calls. Now you own the shares. You sell a call at $260. Someone pays you another $500 for the right to buy your shares at that price. If Tesla stays below $260, contract expires, you keep the shares and the $500. If it goes above $260, your shares get sold at $260 — and you’ve made $35 per share in stock gains plus $10 in total premiums. That’s $4,500 on 100 shares.

Then you go back to Stage 1. The wheel keeps spinning. You collect income regardless of whether the stock goes up, down, or sideways.

“The wheel strategy sounds great until you try to run it for yourself.”

The management is what kills people. Picking strike prices, tracking expirations, rolling contracts — most people give up after a few weeks. Claude handles all of it, checking positions every 15 minutes, closing contracts early when they hit 50% profit, and generating daily summaries at market close.

Claude’s Take

The technical walkthrough is solid. The explanations of trailing stops, options, and the wheel strategy are genuinely well done — the car insurance analogy for options is one of the better ones out there, and the apartment deposit analogy for calls is a nice touch. If you’ve never understood options, this video will get you there.

That said, a few things deserve the raised eyebrow.

The backtested politician-copying returns (34.8% vs 15% for the S&P) are presented as though they’re repeatable. Backtesting is the financial equivalent of predicting yesterday’s weather — it always looks great. Congressional disclosure delays, changing market conditions, and the fact that everyone is now watching these trades (diluting the edge) all complicate the real-world picture. The strategy is not wrong, it’s just shinier in hindsight than it will be going forward.

The video also glosses over a meaningful detail: Claude’s scheduled tasks only run when your computer is on. This is mentioned in passing but deserves more weight. A trading bot that sleeps when your laptop does is not exactly the institutional-grade execution being promised.

The biggest omission is risk. The wheel strategy is presented as collecting income “no matter which direction the stock moves,” which is technically true but misleading. If the stock you sold puts on craters 40%, you’re buying it at a price well above market value. The premium you collected barely cushions that. The wheel works beautifully in sideways and mildly trending markets. In a crash, it hands you a bag.

Paper trading is emphasized, which is good. But the emotional distance between “I lost $5,000 of fake money” and “I lost $5,000 of real money” is roughly the distance between Earth and Jupiter. Every strategy works in paper trading because you don’t flinch.

The core idea — using Claude to automate trading mechanics so you can focus on strategy and rules — is genuinely useful. The danger is mistaking automation for alpha. Claude will execute your rules perfectly. Whether your rules are good is a separate question entirely.