The latest articles on DEV Community by Voytek (@askvoytek). https://dev.to/askvoytek https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3908284%2F5bbc59f5-f3e7-4f6e-9d2c-52efcf8f9a2f.jpg https://dev.to/askvoytek en Voytek Wed, 12 Aug 2026 04:00:00 +0000 https://dev.to/askvoytek/i-got-tired-of-late-signals-so-i-built-a-7-agent-defi-bot-on-an-80-raspberry-pi-4ge8 https://dev.to/askvoytek/i-got-tired-of-late-signals-so-i-built-a-7-agent-defi-bot-on-an-80-raspberry-pi-4ge8 <p>I've been in crypto since before it was cool to have opinions about it.</p> <p>For years I watched signal channels on Twitter and Discord broadcast trades <em>after</em> the professional bots had already caught them. By the time a "call" hit my feed, the opportunity was gone. Somewhere in there I asked myself the obvious question: what if I built something that got the signal first, instead of reading about it secondhand?</p> <h2> The Build </h2> <p>Saturday night in April. Instead of scrolling another signal channel, I opened my laptop and started sketching Agent 1: a pool watcher for Aerodrome Finance on Base.</p> <p>First problem, immediately: The Graph API I planned to use had just gone paid-only. Pivoted to DeFiLlama instead. By 2am, Agent 1 was live — scanning 229 Aerodrome pools in real time, flagging 62 as HOT and 42 as WATCH based on yield and volume.</p> <p>That one agent turned into seven:</p> <ul> <li> <strong>Pool Watcher</strong> — scans pools, flags opportunities</li> <li> <strong>Yield Hunter</strong> — ranks pools by risk-adjusted APY</li> <li> <strong>Gap Calculator</strong> — coming in the next post</li> <li> <strong>Trade Coordinator</strong> — decides what actually gets acted on</li> <li>Plus a signal bus, alerting, and a paper-trading execution engine</li> </ul> <p>All seven run 24/7 on an $80 Raspberry Pi 5. Not a $50,000 rig — a Pi sitting in my apartment in Atlanta.</p> <h2> The Numbers (So Far) </h2> <p>Six weeks after that Saturday night, paper trading went live with a $1,000 simulated bankroll.</p> <p>Day 1: <strong>+$3.25 on 5 positions, 4 of 5 profitable.</strong></p> <p>Small number. But it's real proof the pipeline works end to end — signal detection, scoring, coordination, and execution, without a human in the loop.</p> <h2> The Honest Part </h2> <p>I'm not a career software engineer. 25 years in IT infrastructure, about a year of teaching myself Python and Linux seriously. "Always learning, never building" was the trap I was stuck in for a long time. This project was the fix — and it's still very much a work in progress, running on paper money, not real capital yet.</p> <h2> What's Next </h2> <p>This post is the concept and architecture. The next few posts in this series go agent-by-agent with the actual code: how the gap calculator scores opportunities, how the coordinator decides what to trade, how the signal bus ties it all together, and eventually — real money results, not paper.</p> <p>If you're building something similar or have opinions on the architecture, I'd genuinely like to hear them in the comments.</p> <p>Whatever you're building today, keep going. 🚀<br> — Voytek</p> python raspberrypi showdev cryptocurrency