I turned $50 into $500 with an AI trading bot. Then I stopped checking it for a few days.
When I came back, the balance was zero. 814 trades. $193 in losses. $115 in fees. All gone.
Here’s what went wrong, and how I built a bot that now fixes itself automatically.
What you’ll learn:
• How the original Battle Royale bot went from $50 to $500, then back to zero
• Why manually tweaking your own bot makes things worse (my win rate dropped from 19% to 12%)
• What Karpathy’s “auto research” concept is and why it changes everything
• How I built a self-improving trading bot using Claude Code and OpenClaw
• Why the system auto-rejects strategies that cheat by looking at future data
• How to set this up yourself in about 30 minutes
Timestamps:
00: 00 Intro
0: 26 Battle Royale recap
1: 12 What if I just kept it running?
3: 52 Why it failed
5: 36 Karpathy’s autoresearch
6: 40 How it works
10: 59 How to build this yourself
12: 16 The 3 files you need
13: 01 Installation
16: 22 The future & next steps
18: 39 Conclusion
Links:
• Karpathy’s AutoResearch repo: https://github.com/karpathy/miniAutoResearch
• OpenClaw: https://openclaw.ai
• Claude Code: https://claude.ai/code
Want your own content pipeline built on AI agents? We help founders build their presence on X and YouTube without spending hours a day on it.
→ https://founderfunnel.com
Subscribe if you want more of this. New video every week.