AI employees monitor 59 stocks overnight. Start of the operation diary for stock investment powered by a structured system
Nice to meet you. I’m someone who runs US stock investments using a system, together with an AI I call an AI employee. The human decides on trades, while AI employees handle candidate generation, data inspection, midnight monitoring, and execution. In this series, I will publish the practical operations and records for free. It originated as a reprint of my daily investment diary written on Note.
AI employees monitor 59 stocks overnight. The investment diary of a systematic stock investment starts now
AI employees, tonight as well, are on duty.
When the US market is open in the evening, it is not a human but an AI employee that watches 59 stocks.
I review and approve the plan in advance. The AI employee monitors the conditions within that scope, executes, and records. In the morning, I read the results.
We have reached a point where AI handles 99.7% of the work involved in stock investing. The remaining 0.3% is human approval of the plan and morning record checks. This is not about profit margin or win rate; it is about measured working time.
I am not bragging about a completed system. It is still being developed. From today, I will publish the operation and improvement records. This is the first installment of the series “Investment Diary.”
Summary
- What we are doing: Operating a system where an AI employee watches 59 US stocks overnight. Trades are made only within plans approved in advance by humans
- How we operate: The basic approach is “seed sowing”—not predicting growth, but holding many high-quality stocks in small portions. We do not forecast
- What we publish: Daily movements, the AI employee’s work, failures and improvements. We do not disclose amounts or real stock names
Overall diagram of the system
The daily flow looks like this.
In the morning, I read last night’s records. In the morning, the AI employee creates tonight’s plan. It includes the stocks, conditions, and upper limits on amounts. If I read and approve it, the AI employee watches the market at night, and only those that meet the conditions move. In the morning, the results are recorded, and they become material for the next day’s diary.
What matters is the order. Plan first, then approval, and execution last. The design is such that there are no trades without approval.
Why we are building such a system
To realize the asset management team that only the super-wealthy could hire, with a few smartphone apps.
That is the aim of this system.
If you break down what an asset management team does, it roughly falls into three parts: continuously watching a large number of stocks, acting decisively only when conditions are met, and recording all judgments.
All of these are demanding for humans to do every day. Watching through the night continues. The moments that meet the conditions arrive while you’re asleep. The records are plain and do not continue without upkeep.
Therefore, we have created a system that delegates these three tasks to AI employees. Humans only handle plan approvals, morning record readings, and setting overall policy for the system (policy creation is not included in the daily working time measurement).
Current status
Since this is the first installment, I’ll put the current status in numbers. To show the types of numbers, I will separate actual measurements from others (approximate figures, goals, policy).
Actual measurements (as of November 10, 2026)
- Watchlist/holdings: 59 stocks (52 individual stocks, 7 ETFs)
- Automation rate: 99.7% (the share of stock-investment work handled by the AI employee)
Assumptions and policy
- Population: About 13,000 US-listed stocks (approximate). From them, I gradually select those that meet the criteria and increase the list; currently, the 59 stocks are under watch/holdings
- Publication: The diary aims to publish one entry the morning after the US market opens. All articles are free
- How performance is shown: No dollar amounts. Not daily; instead, at defined intervals, with rates and measurement criteria attached
A note on the 99.7% automation rate: this figure compares the time as if all stock-investment work were done manually. Even just the price checks for the watchlist, at 10 seconds per item, amounts to several hours a day. By contrast, the human effort involved is only a few minutes per day for funding preparation, morning plan approval and recording, which constitutes 0.3% of the total.
Below is a reproduction screen of the watchlist for the 59 stocks as of November 10, 2026. This is not an actual app screen. Stock names are anonymized. Amounts are not shown. What is shown is the number of stocks and the day’s movement (price change percentage).
Can you, alone, monitor so many stocks in real time and make decisions from 11 PM to 6 AM?
Even if humans cannot, the system can. The work of the AI employee is to calmly handle this “volume.”
Investment philosophy (seed sowing)
No one can predict which stock will rise. We do not attempt to guess the outcome.
Instead, we sow seeds. We select many stocks with quality, divide them into small portions, and let nature take its course in the market. Those that grow are harvested a little at the defined conditions and then go back into the next seed sowing.
Give every industry a chance. Let those that grow grow to any extent. Target both dividends and price appreciation. These three form the basic thinking of the operation.
Things we do not do here
As a promise, here are the things we won’t do.
- We do not recommend stocks. We do not give “buy now/sell now” instructions
- We do not forecast the market. We do not provide targets like target stock prices
- We do not disclose amounts (total assets, account balances, trading amounts)
- We do not use expressions that guarantee results. We will present records that include periods that did not perform well
Today's reflection and what was corrected
As a diary format, this section appears in every entry. For the first one, I will write what I corrected before starting.
When starting to publish, the first decision was “what not to reveal.” The real screens cannot be shown as-is since they could identify the services used. Therefore, I rebuilt the screens from actual data, used pseudonyms for stocks, and did not reveal amounts; this is the first recreated screen of this article.
Another decision was the article format. From studies of successful writers’ first installments, none started with just a self-introduction. From the first installment, the series format was already set, and the promise was not results but “what to publish.” This diary starts in that same format.
Conclusion
That completes today’s setup.
From tomorrow, I will record what I watched the previous night, what conditions it reacted to, and what moved. If nothing moved, I will write that as well. The goal is to publish the previous night’s records the morning after the US market opens.
Whether it went well or failed, I will leave a record of how the system operated. This is the promise of this Investment Diary.
If you find it helpful, please follow me.
AI employees are on night duty again today.
About the images: The screens in this article are recreated from actual data (not real app screens. Stock names are pseudonyms, amounts are not displayed).
— Where the tools are kept
The tools for this operation are also available at GoGoJiang (GogoJaun).
・FX EA set “ freely∞ color is nondual”
・FX indicator set “Infinite Custom AI”
The operation diary is also published daily on Note (account: htstar_trader_ai).
※This article reflects personal opinions and does not recommend the purchase or sale of any specific stocks or products. Investing carries the risk of losing principal, and in some cases losses may exceed the principal. Results are not guaranteed. Please make investment decisions at your own risk.