Published on Sun Sep 14 2025 19:00:00 GMT-0500 (Central Daylight Time) by Orkid Labs

Thirty days ago we set out to build ORKID — the MEV signal engine that actually sees. Not vibes. Not lagging charts. A principled system that measures, simulates, and learns. Today, we’re celebrating the first month with meaningful steps forward in scale, realism, and scientific rigor.

What changed since our last post

• Single‑process A/B harness for fair comparisons
  • ArbDetector and FMD (our physics‑informed detector) now alternate each detection cycle under the same mempool context. Apples‑to‑apples, tagged analytics, zero RPC thrash.
• Uniswap v3 + Algebra quoting on Polygon
  • New V3QuoteService with fee‑tier sweeps (500/3000/10000) for Uniswap v3 and dynamic‑fee QuickSwap v3 (Algebra). Bigger surface, tighter quotes.
• Multi‑venue arbitrage search (V2+V3)
  • Sampled‑size search across ordered venue pairs (Quick, Sushi, plus v3s), combining V2 reserves math with V3 quoters to select best net after gas.
• Mempool realism and gas sanity
  • Polygon gas = base + tipP90 with clamps; WMATIC→USDC on‑chain quotes to dollarize. No meme spikes; honest EV.
• Keen.io analytics with canonical event time
  • Every event carries an explicit keen.timestamp. Queries reflect when things happened, not when they arrived. Normalized fields for detector, route, size, gross/net USDC, spread, sim stats, and mempool snapshots where available.
• Token and venue expansion
  • Added SOL (wormhole, 9 decimals), BNB, AVAX, SHIB. Env‑driven v3 activation. Clean unknown‑token handling.

The north star: first‑principles physics

ORKID isn’t a bag of heuristics. It’s a physics‑informed engine that models pressure, flow, and transition paths — then validates with real execution costs. The A/B harness is our instrument: Arb vs FMD comparisons on a shared mempool snapshot, with event‑time analytics and clean safety gates:

• Minimum size/ROI filters to suppress noise
• Dedup windows to avoid repeats
• Sim‑derived gas usage to converge on honest EV

Technical highlights

• V3QuoteService
  • Uniswap v3 fee‑tier sweep with best‑quote selection
  • QuickSwap v3 (Algebra) dynamic‑fee quoter support
• Multi‑venue arb (sampled sizes)
  • Ordered pair search across V2 and V3 venues, pick best net after gas
• Mempool + gas integrity
  • Smart WS observer, feeHistory + tipP90, realistic MATIC dollar conversion
• Analytics with Keen.io
  • Explicit keen.timestamp, unified dimensions, configurable sampling

Why this matters

• More signal, less guesswork
  • V3 efficiency plus multi‑venue routing uncovers edges that v2‑only scanners miss.
• Science over superstition
  • Alternating detectors on a shared cadence isolates causality and accelerates learning.
• Operational confidence
  • Feature flags, safe defaults, and observer health provide a platform that can scale without regressions.

Charts and what we’re tracking

• Detection throughput and positive rate by detector
• Net and gross USDC by venue pair and token
• ROI percentiles and stage latencies
• Mempool overlays: tipP90Gwei, sandwich intensity, dup rates

Next 30 days

• Cross‑protocol MEV: Curve/Balancer, Aave; conservative minOuts + sim gating
• Private routing: protect‑first strategy, TTLs, relay health, replacement rules
• Adaptive size search: refinement around the best sampled size
• Analytics polish: saved queries and cached datasets for investor‑ready dashboards

Call for design partners

We’re opening a few design‑partner slots. If you value correctness and scientific rigor, ORKID is built for you. We’ll provision a private dashboard, tailor venue coverage, and deliver weekly action‑ready reports.

Thank you

Thank you to everyone who helped us reach Day 30. ORKID’s foundation is set. The next month is about depth and scale. We’re building the signal engine that actually sees.

Editor’s note: This post was updated on September 19, 2025 to include a custom Open Graph image for improved social media sharing. No content changes were made.

Written by Orkid Labs