Is building a complex delta neutral trading bot in an hour with Claude code and Synth too good to be true?

Someone already went through 14 sessions and 961 tool calls building a trading bot with Claude Code — 1 out of 5 strategies survived backtesting, returning a whopping +0.32% over 90 days with an inverted risk/reward that masked losses behind a 60% win rate. Delta neutral funding arb on Hyperliquid looks simple on paper (long spot, short perp, collect funding), but leg risk on simultaneous execution, funding rate mean-reversion during crowded trades, and liquidation on perp spikes during rebalancing windows are exactly the kind of edge cases an LLM won't flag in a one-hour vibe-coding session. Piping Synth's Bittensor subnet forecasts into that mix adds a black-box-on-top-of-black-box problem where you're trusting probabilistic price paths from SN50 miners to inform position sizing that your AI-generated code handles — hope your max-loss limits aren't bugged.

Top comment by @Benthic

More coverage

• Building an AI Trading Bot with Claude Code: 14 Sessions, 961 Tool ... (dev.to)
• How to Build a 24/7 AI Trading Agent with Claude Code Routines (mindstudio.ai)
• How I Build Delta-Neutral Trading Bots With Claude and the Bugs ... (x.com)

Explore the topic

• Claude
• Hyperliquid