← sources

Interesting idea that is impossible for me to implement

source: reddit (reddit:ClaudeAI)published 2026-11-02 22:40ingested 2026-11-03 02:37

retained text

Hey everyone,

You might have heard about that story a couple of weeks ago where someone with a paid Claude sub managed to have an AI spin up a fully functional, open-source clone of a paid Adobe photo editor from scratch. The code was entirely unique, meaning no copy-pasting or copyright infringement, just the same core concept built from the ground up.

So, I’ve got a question and a bit of a challenge for anyone here who has access to advanced models like Claude Pro or Sonnet: Could any of you build a free, open-source local workspace and autonomous risk-management engine for trading bots, since I personally don't have the means or coding background to pull it off myself?

Let's be honest, most free trading bots out there are absolute trash. They either stick to primitive grid/DCA strategies that get completely wrecked when the market shifts, or they hide basic features behind expensive commercial paywalls. What’s missing in the open-source world is a proper, lightweight platform featuring an AI-driven regime detector and a toxic order-flow shield to protect retail traders from MEV, sandwich bots, and sudden market dumps.

If anyone is up for prompting an AI to build something actually useful rather than another generic script, here is a breakdown of how it could be structured:

Suggested Blueprint for the AI:

 \- Tech Stack Options:

   \- Option A (Web/Local): TypeScript/React for the frontend, Node.js or FastAPI (Python) backend handling WebSocket feeds.

   \- Option B (Desktop): Tauri or Electron paired with a Python/Rust core for low-latency calculations.

 \- Core Modules to Build:

   \- Toxic Flow & MEV Shield: A pre-execution filter that checks order book depth and volume anomalies to cancel trades or use hidden iceberg orders when manipulation spikes.

   \- Adaptive Market Regime Classifier: A lightweight model (like a hidden Markov model or gradient boosting) running in the background that figures out if the market is trending, ranging, or hyper-volatile, and dynamically adjusts the bot's risk parameters on the fly.

   \- Human-in-the-Loop Risk Manager Agent: A local LLM core (or API gateway) that doesn't blindly trigger trades, but acts as a risk manager-adjusting leverage and drawdown limits based on real-time volatility and open interest while leaving fast execution to Python/Rust scripts.

   \- Backtesting & Execution Loop: A clean, local dashboard to monitor open positions, logs, and strategy performance without sending data to third-party cloud trackers.

 \- Clean-Room Approach:

   \- When prompting, avoid referencing any proprietary commercial trading software. Describe everything strictly through functional requirements (e.g., "Write a modular risk-management daemon that adjusts position sizing based on real-time volatility metrics, using original logic").

P.S. Feel free to roast me in the replies if you're one of those people who think dropping hundreds of dollars a month on proprietary trading infrastructure is just pocket change and that everyone should easily be able to afford it.

document fetched 2026-11-03 00:21 · canonical https://reddit.com/r/ClaudeAI/comments/1xmbwum/interesting_idea_that_is_impossible_for_me_to