Updates
The honest account of building Proof Systems in public.
Project Update — May 30, 2026
Six weeks on, the go-live timeline slipped — on purpose. Here's what the last six weeks actually taught me about the gap between detecting an opportunity and capturing it, and why $50,000/month is still the goal.
Read more →The eleven-night freeze, found and fixed
For eleven straight nights the bot stalled at midnight UTC, then recovered on its own. The cause was a nightly report query blocking the event loop on a 17-million-row table. One index fixed it.
Read more →First execution path completed — in paper
At 01:00 UTC, unattended, seven opportunities ran the full detect-to-execute path end to end for the first time. No alert, no fanfare — just different rows in the database the next morning.
Read more →The edge is proven
12 real execution attempts on DOT/USDT — genuine 0.18–0.35% spreads between OKX and Kraken. Too slow to capture, but the question that mattered was answered: the opportunity is real.
Read more →Running is not the same as working
For 14 days the bot ran clean and logged hundreds of thousands of detections — and executed zero real trades. The cause was a fee set 7.5× too high. The lesson rewired how the whole project is built.
Read more →Day 14: Refactor complete
bot.py reduced from 25,189 to 2,233 lines. 47 modules. 6 exchanges. Every signal now anchored to the Bitcoin blockchain.
Read more →Bitfinex added — 6 exchanges now live
Bitfinex connected and scanning. 6 exchanges now active across 9 pairs. Detection latency: 219ms mean.
Read more →Day 13: First signal verified
MREV-20260404-070729-0001 was the first signal posted to the channel. Verify it at proofsystems.ai/signals.
Read more →Phase 0: Latency tiering deployed
Each exchange now routes to strategies that match its execution speed. XARB requires Tier 1 only. The system scans ~550 combinations across 6 exchanges. Infrastructure cost: unchanged at $26/month.
Read more →Day 1: I was bored
Started with $1,021 and a question. What can one person build with an AI in their spare time?
Read more →