Claude Code Hack For Trading

The speaker begins by explaining the need to set up their Polymarket bot on a server because they are currently in the United States, where they are not allowed to trade, though they ask the audience not to report this. They mention that streaming is difficult while traveling, so moving the bots to a server is necessary. Before setting up the bots, the speaker addresses an issue with "moondev.com arena." A user reported that the hourly data for the competition might not be live, claiming their model should have fired but didn't, resulting in a point deduction. The speaker asks for verification that all scripts are indeed firing on live data. The conversation then shifts to preparing the bots for deployment. The speaker asks for guidance on which files are needed to transfer a bot, specifically the WTA tennis bot, to a new server. They also mention canceling a subscription to "Kling" as it's no longer needed. Regarding the WTA tennis bot, it is described as fully self-contained, with no local helper imports, CSV/JSON reads, or database dependencies. The minimal deploy bundle only requires copying the bot script, and the environment variables exist in the repository, so they need to be installed manually. This pattern applies to other "no bots." An update on the arena issue confirms that it is firing on live data. Hourly ticks are active, with the last round firing at 15:05. Predictions data is live, not stubbed. The user's deduction of one point means their model did fire but predicted the wrong direction, not that it was skipped. However, a significant issue is flagged: the backend has restarted 1140 times, with an uptime of only 70 minutes on the current process, indicating high churn. This needs investigation to prevent future restarts, and the speaker inquires about processes consuming too much power. Further investigation into a node issue reveals that a local node only supports certain changes to the user positions API, which is otherwise fine. Polymarket prices are mirrored to a specific file, and the bug reporter claims "all crypto.json" is healthy. The speaker requests that these claims be verified. An API issue is also reported, with a user stating it's not working to their satisfaction. The speaker plans to test this internally and externally, asking for a recreation of the user's claims. While some fixes have been made to the server, the speaker wants to confirm everything is working. It is discovered that the backend restart issue is not a power resource problem, as the VPS has 11 GB of free RAM and the backend has 240 MB. The problem is that the watchdog log shows "miss one or two, miss two or two, restarting Moondav. Y Health misses." Backend boot takes 20-30 seconds due to heavy imports like BTC Polymarket web socket setup, BTC bars collector, and HL liquidation scanner. Once running, the "five every five minute fetch all" job does a 2.16 liquidation file, which can stall. The speaker asks for the top three heaviest processes to identify potential candidates for removal to "slim down" the system. The "fetch all data job" every five minutes is identified as one of the heaviest. The speaker questions where this data is outputted, considering it's also available via API for direct download, suggesting it might be removed from the website. They also question the necessity of the "IP API" for geo lookup, asking if it's for click tracking and suggesting its removal. Additionally, they propose deleting "scan hl liquidations" and deprecating "moondav.com/game," though they express fondness for the game. During testing, it was found that the API was running in a single worker system, which gets saturated. Rotating keys brings it up with eight workers, making things healthy. However, if System MD restarts it on its own, it will revert to a single worker, causing the issue to recur. A durable fix is requested. The speaker also briefly touches on email security, asking for advice on managing multiple email addresses for services to prevent spam.