The trading section of Hey Jarvis covers grid bots on crypto, an RSI(2) stock bot, a backtest script, and a dashboard that shows what they are doing. I run all of it on paper money. After it was published I kept going, and the biggest thing I added was an options bot. I want to be upfront about it, because it is the one addition I would not hand to everyone who builds a Jarvis.
What it does
It started as a SPY iron condor, which sells a call spread and a put spread at the same time and collects a small credit if the price stays between them. After a backtest across 2007 to 2026, I switched it to a bull put spread, the simpler one-sided version. It sells a put below the market, buys a cheaper one further down for protection, and keeps the credit if SPY stays above the short strike until it closes.
The rules are deliberately narrow. It looks at the market once a week, Monday at 10:00 Central, and nowhere else. It only enters when SPY is above its 200-day average. Every proposal comes to my phone on Telegram, and if I do not approve it within 15 minutes it is denied. It is built so it cannot place a real-money order: the broker connection is created in paper mode, and there is no code path that creates it any other way.
Here is how it looks on the Options tile in OVERWATCH, the dashboard that grew out of the one in the book. The tile still says condor because that is how the bot started; the open position is the bull put.
Why it may not be for you
Look at the middle card. The most this position can make is $172. The most it can lose is $1,828. That is the deal you take with a credit spread: you win small and often, and every so often you lose much more than you won. The backtest behind it, which priced the options with models because real historical quotes are hard to get, showed drawdowns of close to 20 percent in some periods. It can hold up to six positions at once, and a sharp drop in SPY hits all of them together. There is also no stop loss: in testing, stops cost more than they saved, so the protective put is the only limit. Those are reasonable choices in a test. They feel very different in your own account.
So I would skip this addition if any of these sound like you:
- You have not traded options before. Learn how a spread wins and loses with a book and a calculator first, not with a bot.
- You want set-and-forget. This needs a tap from you every week and a look at it when the market moves.
- You would be tempted to flip it to real money after a good month. A good month proves very little here.
- A $1,800 loss on one position would hurt. Then the position is too big for you, whatever the backtest says.
I added it because options risk is something I wanted to understand by building it, slowly, on paper, with the same RCA & trial and error process I sometimes use in quality work. That is a reason that fits me. It is not a recommendation, and nothing here is financial advice.
What paper trading taught me
The most useful lesson so far had nothing to do with strategy. In the first weeks, a four-leg order filled 91 milliseconds after I sent it, below the minimum price I had set, with three of the four legs printed outside the quoted market. A real exchange would not have filled it that way. The paper simulator was simply not honoring the order price, which meant every fill number in my log was measuring the simulator, not the bot. I still catch myself wanting to read paper results as proof, and that episode is why I don't.
It also showed me a missing check. The bot was not testing how old its price quotes were, so it could have judged a trade on stale numbers. It now refuses any quote older than five minutes.
The bot also refuses more than it trades. Before every Monday proposal it checks whether short-term volatility is above longer-term volatility, a sign the market is stressed, and blocks the week if it is. When a week is blocked, it tells me why on Telegram. I also kept the daily chain check from the condor version running as a second opinion, and it says no often.
That panel is the condor check, not the bull put's own gate, and it disqualified the day because volatility was below its minimum. I keep it because it is a cheap, independent look at the same market. A bot that rarely trades is easy to underestimate, but for this kind of strategy, the days it says no are part of the design.
If you want to build it anyway
Start with what the book already gives you: the trading dashboard, the backtest script, and the habit of running everything on paper first. The book does not cover options, and I would not try this before the stock and crypto bots are boring and reliable on your machine. If you do build it, give it the same three guards mine has: paper only by construction, a human approval on every trade, and a hard limit on how much any one position can lose.
Mine also has a kill switch. Sending /halt from my phone stops every new proposal and order, and the halt survives a restart; it only clears from the console at my desk, after I have looked at what happened. The bot halts itself if it ever finds SPY shares in the account, which would mean a put was assigned. If its broker readings stop making sense, it blocks new entries until they do. It also checks in with Jarvis every five minutes, so if it goes quiet, I hear about it. One honest limit: a halt freezes the bot, it does not close positions. Anything already open sits at its defined maximum loss until I close it myself.
Hey Jarvis: Your Machine. Your AI. Your Rules. walks through the whole build, from the wake word to the trading dashboard this grew out of. And if you are here for the problem solving side, the method I use to chase down issues like that paper-fill one starts with the free worksheet at https://solved360.com/pages/free.