The Use of AI with ThinkOrSwim Scripts

[Art]

I’ve been creating financial dashboard models and scripts, and from my experience, one LLM is never enough if you care about accuracy, logic, and security.

AI can help a lot, but I would not trust one model by itself. I set up a loop or bridge so one model builds or reviews the script, then another model checks it against the goal, logic, errors, and possible security issues.

For example, if I use Claude or Claude Code to build something, I’ll run it through ChatGPT or Codex to verify it. The model that writes the code should not be the only one reviewing it.

 

[Adeodatus]

I’ve been creating financial dashboard models and scripts, and from my experience, one LLM is never enough if you care about accuracy, logic, and security.

AI can help a lot, but I would not trust one model by itself. I set up a loop or bridge so one model builds or reviews the script, then another model checks it against the goal, logic, errors, and possible security issues.

For example, if I use Claude or Claude Code to build something, I’ll run it through ChatGPT or Codex to verify it. The model that writes the code should not be the only one reviewing it.

Hi Art,
Long investor-trader, I'd ask about the reviewing code on another network apart. Isn't that laborious and defeating the difference in the AI modeling. It seems the outputs build on all the data exchange and coding produced, (Warp speed Captain), and singular AI data bases collectively code, so about flip-flopping between - is that degraded script? Some ultra-scripted monitoring indicators will grab the total chart, you can't add any lower. The Suite of CarbonStation monitoring studies may and can lead to predictions, [never works].

DYODD