diary
- to maintain mistakes and learn
i want this project to be an automated stock tracker
the goal is: ai agent will check the market frequently, and it will have portfolio of human, as it moniters through out the day, it
suggests to buy or sell stocks to human via telegram/whatsapp. the goal is to maximize profit and make sure human doesn’t loose
money.
i have some ideas
-
the agent will be given current human portfolio, what stock she owns
-
there is a main agent, who is a smart model (gpt 5.6 sol max), and he manages agents for other subagents. for example, one subagent
may be 5.5 medium or 5.3 codex spark tracks the news and keep checking for government contracts that can raise the stock or may
reduce the stock
- and there is another subagent who is taking care that stock doesn’t fall too loss, like may be for stop loss. but he will only
report to big smart agent, he wil reason about whether or not to make a buy or not decision
- and may be there is one agent just looking for current stocks in portfolio, that is going up. and suggesting to sell
each of the agent can be of different smartness, like for news one could use a fast ,smart agent as it checks summaries
- and there might be one agent who looks quantitatively at the data , but i don’t know how good can this be? its difficult no, uness
u have renaissance like powerful compute.
also, we should make sure that the human does not loose money, its important to play safely while also making profits
there are some details, which I didn’t think like
- how frequently the check should be, every 15 mins? every 30 mins?