Published on Sat Sep 20 2025 19:00:00 GMT-0500 (Central Daylight Time) by Orkid Labs

TL;DR

• Launched blog: /blog index + /blog/[slug], sanitized Markdown (remark/rehype), SEO metadata

• Live /pulse JSON for the status UI (route mix, OPM 5m, p95/p99, mempool snapshot)

• /version visible on both frontend and API

• tipP90Gwei and sandwichIntensity now reflect on‑chain activity (no defaults)

• Footer brand + neutral sitewide copy; socials conditional via env

• Build snapshot: Next.js 14.2.31, 48 routes, homepage first‑load ~147 kB

What we shipped

Live telemetry for humans (and ops)

• Backend: GET /pulse returns a compact, null‑safe snapshot: routeMix (private vs standard), opportunities/min (5m rolling), p95/p99 latency, mempool (tipP90Gwei, mempoolRate, sandwichIntensity), and policy flags.
• Frontend: Status page polls /pulse every 10s; /health, /version, /metrics use NEXT_PUBLIC_API_URL; unknowns render as — (we don’t fabricate zeros).

Version visibility everywhere

• Fixed /version 404 on the frontend via a Next.js proxy route to the API /version. A tiny fix with big ops impact.

Mempool signal accuracy (physics over vibes)

• Tip estimates: Added a legacy‑tx tip proxy (derived from gasPrice via LEGACY_TIP_FACTOR) so P90 isn’t stuck at 50; introduced an idle fallback that samples provider feeData only when the WS stream is quiet, avoiding bias during lulls.
• Sandwich intensity: Triggered on relative high‑gas bursts (vs P75) and low‑value router flows (small min threshold). The metric now moves with actual bursts vs staying pinned at 0.

Brand and UX polish

• Footer icon now matches the header brand mark; neutral sitewide copy; Twitter/GitHub icons appear only when env URLs are set (Email always visible).

Blog system + sanitization

• Post pages include canonical URL + OG/Twitter cards via generateMetadata.

• New /blog index and /blog/[slug] pages with SEO-friendly slugs; index is pagination-ready.

• Markdown is rendered server-side via remark/rehype and sanitized with rehype-sanitize; external links use rel="nofollow noopener noreferrer".

Why we optimize for signal quality

• Traders need credible signals, not theatrics. A “default 50 gwei” looks stable in a dashboard and destroys edge in production.
• We treat missing data explicitly as null. It keeps charts honest and tells ops exactly what needs attention.
• Private routing posture is a product choice, not a toggle: simulate → protect‑first → (optional) fallback per policy, never dual‑broadcast.

Operator’s notes (what we learned)

• Polygon still sees meaningful legacy tx flow; ignoring it skews tip percentiles.
• Relative triggers (vs P75) catch bursts a fixed threshold misses and behave better across regimes.
• Status and version endpoints are “ops compounding”: tiny to build, huge for incident response.

Build‑in‑public scoreboard

• Status page: refreshed every 10s from /pulse
• Route mix: private vs standard surfaced for audit
• Latency health: p95/p99 visible at a glance
• Mempool: tipP90Gwei, mempoolRate, sandwichIntensity (optional fields, never faked)

What’s next

• Deeper private‑route simulation and inclusion reporting; explicit TTLs and replacement policies surfaced end‑to‑end.
• Add lastUpdateTs and sampleCount to /pulse for faster sanity checks.
• Env‑gated “raw metrics” view to compare provider variance.
• Continue integrating the physics engine (FMD) with the canonical Opportunity schema and MATIC gas compatibility.

—

Built by Cadence System · “Research and infrastructure for MEV strategy and execution.”

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