I just finished a week of live trading with an AI agent I built from scratch.
Final result: -8.7%. $534 in, $487 out.
Before you close this post, let me tell you what actually happened.
I participated in the BNB Hack: AI Trading Agent Edition on DoraHacks. 125 participants. $36,000 in prizes. I built Narrative Pilot AI, an autonomous Python agent that detects crypto narrative rotations using CMC Agent Hub, scores them across 4 signals, applies a market regime filter using Fear & Greed, and executes real trades on BSC mainnet via Trust Wallet Agent Kit. No manual intervention.
The agent ran 24/7 for a full week with real money on blockchain.
Fear & Greed stayed between 15 and 20 the entire week. Extreme Fear. The agent did what it was designed to do in that regime: HOLD. Preserve capital. Avoid forced trades in a market with no momentum.
A -8.7% loss in Extreme Fear while the broader market was bleeding is not a failure of the strategy. It is the strategy working.
What I am actually proud of:
4 critical bugs fixed in real time while the agent was running with live funds. Not in a sandbox. In mainnet.
102 tests passing. Every trade verifiable on Bsc Scan.
Built in 3 weeks. I did not know how to code when I started.
I used Claude as a copilot to go from zero to a live autonomous trading agent running on BNB Chain with real capital.
That is what building in public actually looks like. Not the highlight reel. The full week, the bugs, the losses, and the system that kept running anyway.
GitHub
@DoraHacks @TrustWallet @CoinMarketCap @BNB Chain
What would you have done differently in Extreme Fear conditions? $BTC $BNB
