System Live — 11 Loops Active

X Loop Architecture

Fully automated content engine for @eToroBuilders — 11 scheduled loops handling content creation, engagement, analytics, and paid promotion.

11
Active Loops
692
PI Handles
3
Config Layers
36
Posts/Day Cap

Post Dispatcher Flow

All automated loops enqueue posts — a single dispatcher dequeues with 20-min gap + random stagger

  ┌─────────────┐
  │ Content Loop │──┐
  └─────────────┘  │
  ┌─────────────┐  │    ┌──────────────────┐    ┌────────────┐    ┌─────────┐
  │  PI Engage   │──┼───▶│  Post Queue       │───▶│ Dispatcher │───▶│  X API  │
  └─────────────┘  │    │  (SQLite)         │    │ (every 5m) │    └─────────┘
  ┌─────────────┐  │    │                   │    │ 20min gap  │
  │   Sniper     │──┤    │  Priority:        │    │ + 2-8m     │
  └─────────────┘  │    │  urgent → high    │    │ stagger    │
  ┌─────────────┐  │    │  → normal         │    │            │
  │ Reply-Back   │──┘    └──────────────────┘    │ Hard cap:  │
  └─────────────┘                                │ 3 orig/hr  │
                                                 └────────────┘
  ┌──────────────────────────────┐
  │ Replies & Manual posts       │──────────────────────────────────▶ X API
  │ (always direct, no queue)    │                                   (immediate)
  └──────────────────────────────┘
URGENT: manual, reply-back HIGH: pi-engage, sniper NORMAL: content

All 11 Loops

Click any card to expand details

📝
Content Loop
9, 12, 15, 18, 22 ET
5/day
Main content engine — creates and posts original tweets. Trending topic scouting, multi-candidate generation, critic scoring, compliance checks.
↓ click to expand
Content Mix: 30% punchy · 30% educational · 20% BTS · 20% app spotlight

Pipeline: Scout trending → Generate 3 candidates → Anti-AI scan → Critic score (reply-bait ≥6, brand ≥7) → Compliance → Link check → Image decision → Post → Report to WhatsApp

Auto-boost: After every post, asks WhatsApp for boost approval. YES → $15/day, $100 lifetime cap per post. Polled by boost-checker.js every 5 min.

Session: Persistent (remembers what was posted across cycles)
Model: claude-sonnet-4-6
🎯
Sniper Loop
every 2h
3/day
Monitors 50 top creators for quote-tweet opportunities. All drafts require human approval via WhatsApp.
↓ click to expand
Watched accounts (50): @sama, @karpathy, @danielgross, @swyx, @VitalikButerin, @naval, @balajis, @paulg, @levelsio, @chamath, @pmarca + 39 more

Filters: Freshness <2h · 48h cooldown per creator · Max 3/week/account · Relevance ≥7

Builder keywords: "eToro API", "eToro MCP", "eToro agent", "trading bot eToro", "copy trading API", "agent portfolio"

Approval: Sends draft to WhatsApp → waits for explicit YES. Never auto-posts.
💬
Reply Loop
every 15m
20/day
Monitors @eToroBuilders mentions and replies to genuine questions. Max 5 per cycle, 20/day shared budget.
↓ click to expand
Detection: x_search for @etorobuilders mentions → filter dev questions (reply) vs praise (like) vs spam (skip)

Style: Max 180 chars, casual. "solid catch", "yeah we're working on that"

⚠️ 403 Limitation: Can only reply to tweets that @mention @eToroBuilders directly. Cannot proactively reply to any public tweet. X API policy restriction.

Silent: 0 replies in cycle → exits silently (no WhatsApp noise)
🔁
Reply-Back Loop
every 30m
75.0 weight
Replies to people who replied to our posts — triggers the highest algorithm signal at 150x a like.
↓ click to expand
Algorithm impact: Author reply-back = 75.0 weight (150x a like). Highest positive signal in X algorithm.

Reply window: 25 min to 4h after original post

Priority: 1) Genuine question 2) Interesting take 3) Agreement + experience 4) PI replier (auto-boost) 5) Skip: generic praise, trolls

Limits: Max 5 reply-backs/cycle, max 3 per single post. One pass per tweet.
🤝
PI Engage Loop
9, 13, 17, 21 ET
8/day
6-agent marketing team posting about eToro Popular Investors with the Builder Connection Playbook.
↓ click to expand
6-Agent Team: Orchestrator → Researcher → Strategist → Copywriter → Critic → Compliance → Publisher

5 Angles (Builder Connection Playbook):
📊 DATA: "we analyzed [PI data]"
🔧 BUILDER: "devs on eToro API can learn from [PI]"
🏗️ INFRASTRUCTURE: "[PI achievement] possible because [platform]"
👥 COMMUNITY: "builder community is watching [topic]"
📦 PRODUCT: "building [feature] inspired by PIs like [handle]"

Series: Ask a PI (3/wk) · PI Spotlight (2/wk) · PI vs PI (Thu) · Data Drop (1/wk) · Welcome to the Club (daily)

Cooldowns: 7-day per PI, monthly cap for Cadets
🚀
Boost Checker
every 5m
$15/day
Polls WhatsApp for boost approval → creates X Ads campaigns. $15/day, $100 lifetime cap per post.
↓ click to expand
Flow: Content post → WhatsApp approval request → boost-checker polls for YES/NO → YES triggers X Ads campaign

Budget: $15/day per post, $100 lifetime cap. budget_optimization = LINE_ITEM

Runtime: ~84ms per cycle. Deterministic Node.js, no LLM.
📊
Daily Report
23:00 ET + 23:30 UTC
2 reports
Two complementary reports — nightly analytics (x-loop-measure) and daily WhatsApp summary (daily-report.js).
↓ click to expand
daily-report.js (23:30 UTC): Follower count + delta, new PI followers, post summary, boost activity

x-loop-measure (23:00 ET): All posts with engagement metrics, engagement rate per post, best hour/format, updates audience-insights config
🔍
PI Daily Scan
00:07 UTC daily
692 handles
Scans all 692 PI handles daily, ranks by engagement, posts top performers to WhatsApp.
↓ click to expand
Pipeline: pi-scan-v2.mjs (fetch tweets) → pi-daily-report.mjs (rank by engagement) → WhatsApp report

Reports: Top 10 by engagement, most active PIs, high-quality finance posts

Note: Passive intelligence only. Does not post to X.
🧠
Weekly Synthesis
Sunday 23:00 ET
self-tune
Analyzes all loop learnings, finds cross-loop patterns, auto-tunes shared config within guardrails.
↓ click to expand
Reads: Up to 50 cycle reflections from x-loop-learnings (past 7 days)

Can auto-tune: tone_spectrum %, breathing_room_target, format ratios
Never touches: compliance, anti-AI, budgets, voice character

Output: Weekly report to WhatsApp: top 3 performers, patterns, config changes, recommendations
🔄
PI Refresh
Monday 07:00 IST
weekly
Refreshes PI X handle database from Databricks every Monday. Updates pi-handles.json + MemClaw.
↓ click to expand
Source: Databricks CLI query on TradingClaw (Splinter server)

Tiers: Elite Pro / Elite / Champion / Cadet

Output: pi-handles.json + memclaw_doc x-loop-files/pi-x-handles-latest + WhatsApp report
📈
X Ads Reports
06:00 + 15:00 IST
2x daily
X Ads performance reports twice daily — morning and evening. Runs ads-report.py → WhatsApp.
↓ click to expand
Morning: 06:00 IST — overnight performance
Evening: 15:00 IST — day performance

Script: ads-report.py → formatted report → WhatsApp group

3-Layer Config Architecture

Change once, all loops follow — powered by MemClaw collections

Layer 1

Shared Config

All loops read on every cycle. Change once → propagates everywhere.

voice anti-ai-detection compliance posting-rules daily-budget fleet-agents audience-insights
▼
Layer 2

Loop-Specific Config

Per-loop overrides and settings in x-loop-config collection.

config-content config-reply config-sniper config-pi-engage config-reply-back config-measure
▼
Layer 3

Self-Adaptation

Each cycle writes a reflection → next cycle reads last 3 and adjusts. Weekly synthesis tunes shared config.

x-loop-learnings/* weekly-synthesis

Algorithm Signal Weights

Why reply-back is the most powerful loop — X algorithm scoring

SignalWeightvs LikeVisual
Author reply-back75.0150x
Reply from someone13.527x
Profile click + engage12.024x
Retweet1.02x
Like0.5baseline
"Show less" / mute / block-74.0SEVERE

PI Database

692 Popular Investor handles across 4 tiers — refreshed weekly from Databricks

Elite Pro
12
Elite
76
Champion
168
Cadet
436
692
Total PI Handles

Daily Budget Allocation

Cross-loop daily tracking via shared MemClaw doc — auto-resets on date change

LoopBudgetCounterNotes
Content5 originalsoriginals_postedThreads count as 1
Sniper3 quotessniper_postedAlso counts toward replies
Reply20 repliesreplies_postedShared with sniper + reply-back
PI Engage8 interactionspi_engage_postedBatch counts as batch size
Reply-Back5 per cyclereplies_postedShares reply counter

Control Points

Natural language commands via WhatsApp — tag @CMOClaw

🎤
Voice & Tone
"change the tone to be more [X]"
⚖️
Compliance
"add [rule] to compliance"
📊
Budgets
"set content budget to [n]"
⏸️
Pause / Stop
"pause [loop]" · "stop all loops"