🕳️When Two Rabbit Holes Collide: Building an AI Trading Bot as Bitcoin Enters Free Fall
Making sense of the future, one obsession at a time. 🕳️
Monday, November 24, 2025
There’s something darkly comic about spending weeks teaching AI to build you a crypto trading bot just as Bitcoin sheds $40,000 from its peak. It’s like finally learning to fly right as you’re running out of sky.
But that’s exactly where I found myself in late 2025—neck-deep in two simultaneous rabbit holes that would collide in the most instructive way possible.
PART I: The AI Agent Gauntlet
The question: Could I, someone with zero coding background, build a functional cryptocurrency trading bot using nothing but AI agents?
I tested them systematically: ChatGPT, Grok, Tasklet AI. Each followed the same pattern—confident code generation, implementation failure, endless debugging loops. The scripts looked professional but broke in practice. They were like three students giving me reworded versions of the same incomplete answer.
Frustrated, I pivoted to Hummingbot—a professional-grade trading bot with a GUI. Surely this would work? It didn’t. After two weeks of failed configuration attempts, I decided to build my own from scratch.
Enter Claude
Claude was different. Not because it worked immediately—the first version also failed. But the underlying architecture was sound. Each debugging session built forward instead of starting over. Claude explained why approaches wouldn’t work, not just what to change.
We went through many rounds. Error messages became learning opportunities. Crashed scripts taught me about exception handling. But we were making progress.
The VPS Odyssey
Building the bot was half the battle. Running it 24/7 meant learning VPS, SSH, Linux terminals, screen sessions—technologies I’d never touched. Every acronym led to another rabbit hole. Claude guided me through all of it.
After weeks: a fully operational trading bot running on a remote server.
Did it make money? No. But it worked. It executed real trades based on logic I’d built. For someone who’d never coded, this was profound—not because of profit, but because it proved AI can genuinely teach you to build functional systems through sustained, iterative engagement.
The key insight: different AI models approach problem-solving fundamentally differently. ChatGPT, Grok, and Tasklet gave surface-level solutions. Claude helped build something that actually worked—through struggle, not shortcuts.
PART II: The Bitcoin Mystery
While I was SSH-ing into Linux servers, Bitcoin was experiencing its own system failure.
October 10th, 2025: The Largest Liquidation Event in Crypto History
$19.36 billion in leveraged positions wiped out in 24 hours.
Bitcoin crashed from $126,000 to $104,000. Ethereum dropped 21%. Some altcoins lost 50-90%. Total crypto market cap: down $370 billion almost instantly.
My trading bot, which had just gone live, executed trades right into this chaos.
Six Weeks of Sustained Decline
But October 10th wasn’t the story. What followed was: Bitcoin has fallen for six consecutive weeks—over 35% peak to trough, now around $85,000.
And there’s no clear consensus on why.
The Competing Theories
Theory #1: The Technical Glitch
Tom Lee: A stablecoin misprice on an exchange triggered cascading liquidations in an over-leveraged market. When thousands use 100x leverage, small moves trigger margin calls, forcing more selling—a self-reinforcing liquidation spiral. This was a “mini black swan event”—a plumbing issue, not a fundamental problem.
Theory #2: The MicroStrategy Delisting Risk
JPMorgan warned that MicroStrategy could be delisted from MSCI USA and Nasdaq 100 indexes by January 15th, 2025. This creates structural selling pressure—institutional funds forced to sell not by choice but by mandate. When MSTR falls, BTC follows.
Theory #3: The Great OG Whale Exodus
Owen Gunden, an early Bitcoin trader from 2011, sold his entire holding—$1.3 billion worth of Bitcoin—between late October and mid-November. He held through 2013, 2018, 2022. Then liquidated everything.
He’s not alone. Early holders who’ve sat on illiquid positions for years now have mature markets to exit. When someone dumps $1.3 billion of BTC even over weeks, it creates measurable pressure. If this represents a broader trend of OG exits, the impact compounds.
Theory #4: The Macro Regime Shift
Real yields at 2.32%—highest since mid-2024. Dollar strong at 107.8. Fed not cutting as aggressively as hoped. Trade war fears. Risk-off sentiment.
Why take crypto volatility when you can earn 5%+ risk-free in Treasury bills? Bitcoin historically correlates positively with liquidity and negatively with real yields. When liquidity tightens, Bitcoin suffers.
Theory #5: The ETF Reversal
Throughout 2024-2025, spot Bitcoin ETFs were structural buyers. In November, that reversed. BlackRock’s IBIT: $523M single-day outflow. Total ETF outflows: $2.8B+.
When the structural bid becomes a structural seller, it removes key support. As one analyst noted: “The convergence of forced liquidations and structural ETF selling has pushed the market into a particularly vulnerable state.”
Theory #6: The Four-Year Cycle Isn’t Broken
Bitcoin follows roughly four-year cycles (2013, 2017, 2021, 2025). Every cycle includes 30-40% drawdowns during bull markets. By this logic, nothing unusual is happening—just healthy consolidation before the next leg up. Zoom out, and the pattern holds.
Theory #7: The Zcash Flight—Privacy Coins as the New Safe Haven
While Bitcoin crashed 35%, Zcash surged approximately 15x since September—from ~$40 to over $600.
The correlation: Zcash has a -0.78 correlation with Bitcoin. When BTC dumps, ZEC pumps.
Why? Quantum computing fears. Zcash developers have spent years building “quantum recoverability” while Bitcoin remains vulnerable. Vitalik Buterin warned quantum computers could break Bitcoin’s cryptography by 2028.
As Zcash engineer Sean Bowe explained: “With Bitcoin, even if the quantum risk is low, its ability to respond is poor... In Zcash, we have been thinking about this for so long... the remaining changes do not feel daunting.”
Add rising privacy demand (30% of ZEC now in shielded addresses vs 10% earlier), institutional backing from the Winklevoss twins’ Digital Asset Treasury ($150M+ accumulated), and a philosophical shift—maybe Bitcoin’s institutional adoption made it too transparent, too tracked, too establishment.
Privacy coins represent a return to crypto’s cypherpunk ethos. The irony: Bitcoin crashes as institutions exit, while a privacy coin surges as both institutions and cypherpunks pile in.
The Convergence
My trading bot went live right as Bitcoin entered sustained decline. Meanwhile, Zcash—a coin I’d never considered—surged 15x. The crypto world fractured: those selling Bitcoin in panic, and those rotating into privacy coins.
What I Learned
1. Complex Systems Fail in Cascades When my bot crashed, it was never one thing—it was chains of failures. Bitcoin’s crash is the same: not just the glitch, not just ETF outflows, not just whale selling—all of them interacting, amplifying each other.
2. Iteration Beats Prediction I didn’t build successfully by planning perfectly upfront. I built through cycles of failure and adjustment. Markets reward adaptability over prediction.
3. The Gap Between Theory and Practice is Everything You can read about API authentication forever, but until you debug a 401 error at 2am, you don’t understand it. Theory gives the map. Practice shows the terrain.
4. AI is a Learning Accelerant, Not a Replacement My bot works because I went through the painful process of making it work. If it worked immediately, I’d have learned nothing. AI didn’t code for me—it coded with me.
5. Timing Doesn’t Matter If You’re Learning My bot launched at the worst possible time for crypto. But because the goal was learning, not profit, the timing was perfect. Building during chaos taught me to handle edge cases and system stress—far more valuable than calm-market lessons.
Where This Goes Next
Bitcoin sits around $85,000, down from $126,000. Jeff Park’s analysis suggests Bitcoin’s volatility regime may be shifting back toward its pre-ETF behavior—becoming option-driven again rather than ETF-flow driven.
The next major leg—up or down—may come from volatility markets remembering what Bitcoin is capable of, not from institutional flows.
Or maybe we’re just in the messy middle of a cycle, where narratives shift weekly before some new catalyst tips the scales.
I don’t know. Neither does anyone else.
What I do know: I built something I had no business building. I learned to navigate systems—technical and financial—that were completely foreign. And rabbit holes are never about the destination.
The bot isn’t profitable. Bitcoin is in free fall. None of this went according to plan.
And yet I can’t remember the last time I learned more.
The Questions That Linger
What makes one AI model effective for iterative problem-solving while others excel at one-shot tasks?
Why do markets punish leverage consistently, yet participants keep using it?
When early adopters exit en masse, does it signal maturity or exhaustion?
Can AI actually democratize technical expertise, or just create a new class of people who know how to talk to AI?
Is timing ever “right” for building things, or is the act of building itself what creates opportunity?
I don’t have answers. But I have a functioning trading bot, practical coding skills I didn’t have three months ago, and a front-row seat to one of the most fascinating market dislocations in crypto history.
Sometimes the rabbit hole chooses you right when you need it most—even if you don’t realize it until you’re deep underground, watching your bot execute orders into a market that’s forgotten how to go up.
What rabbit holes are you exploring? Drop your thoughts in the comments—I’m always looking for the next thread to pull.
RabbitAIPha • Exploring rabbit holes in science, tech & culture - where wandering ideas collide 🕳️
Kudos to you, Andrew. We’ll have this elephant mapped soon.
We have to remember that what we observe is not nature in itself, but nature exposed to our method of questioning.
— Werner Heisenberg
Great post Drew. I'd love to hear more about your trading system. What's your thesis, what types of trades are you doing, what types of signals are you looking for, what worked in testing, what didn't work. What services or software did you have to buy? What was surprisingly free? What limitations did you run into? What was surprisingly unlimited? Do you think you could apply this to other markets like equities? Go down that rabbit hole!