Greed

Retail trading exploded across crypto, stocks, Kalshi, and sportsbooks, but research stayed fragmented. New traders think in narratives; platforms think in tickers. Greed builds one from your socials.

Comment

Greed — Signal-Driven Trading Intelligence Exchange

Inspiration

Most trading tools start with a ticker:

You already know what you want to buy, then hunt for narrative and data to justify it.

Real alpha often works the other way around:

A story appears on X, a macro thread, Reddit, or a prediction market before you know which assets express it.

We built Greed because we wanted an intelligence exchange where the primary object is not a symbol, but a signal — a hypothesis.

You bring the idea:

  • “ETF outflows crushing BTC beta”
  • “Circle vs Tether”
  • “Oil shock”
  • “AI infrastructure boom”

Greed discovers the branches:

  • Crypto
  • Equities
  • Prediction markets
  • DeFi
  • Sports markets

Then routes users to the appropriate venue to act.

Vision

Greed evolved from our earlier prototype Bibim and combines two projects we had already been building:

Minam

The perception and data layer.

Syuzhet

The narrative and belief-formation layer.

Greed

The shell that makes signals:

  • Tradeable
  • Verifiable
  • Cross-asset
  • Optionally private

What It Does

Greed is a signal-driven trading intelligence platform.

Users (or agents) create signals from:

  • Social corpus
  • Manual input
  • Research
  • Emerging narratives

The system then executes a multi-stage pipeline:

1. Ingest & Corroborate

Collects signal evidence from:

  • X (live or curated feed)
  • Reddit
  • Optional Nimble web/SEC-style corroboration
  • Rust backend services

2. AI Analysis

Transforms raw information into:

  • Structured hypotheses
  • Sentiment analysis
  • Evidence quotations
  • Ranked asset opportunities

Across:

  • Crypto
  • Stocks
  • Prediction markets
  • Sports
  • DeFi

3. Discovery & Routing

Greed surfaces related instruments and routes users toward:

  • Binance
  • Robinhood
  • Kalshi
  • Polymarket
  • DraftKings
  • Other platforms

4. Execution Planning

Per-asset strategy templates:

  • Buy & Hold
  • Momentum
  • Mean Reversion
  • Breakout

Including:

  • Entry framework
  • Exit framework
  • Risk considerations

5. Optional On-Chain Registration

Signals can be registered on:

Monad Testnet

Using:

SignalRegistry

Features:

  • Verifiable signal hashes
  • Optional privacy routing through Unlink

The Flagship Feature: Pipeline

/pipeline

Workflow:

Connect Socials
        ↓
Build Corpus
        ↓
Run Agentic Pipeline
        ↓
Discover Assets
        ↓
Generate Graph Relationships
        ↓
Plan Execution
        ↓
Trade on External Platforms

Features include:

  • Live graph visualization
  • Pipeline flow stepper
  • Evidence display
  • Strategy chips
  • One-click routing

Execution happens on real exchanges, not a simulated in-app ledger.

Architecture

Frontend

Tech Stack

  • React
  • TypeScript
  • Material UI
  • Framer Motion

Pipeline interface includes:

  • Live graph drawer
  • Flow stepper
  • Web evidence display
  • Strategy chips

OpenAI runs directly in browser analysis with local fallback support.

Backend

Rust + Axum

Responsibilities:

  • Signal API
  • X graph corpus
  • Reddit public corpus
  • Nimble corroboration
  • CORS handling

Additional features:

  • Environment-driven rate limits
  • Budget-aware caching
  • API cost management

Blockchain

Monad Testnet

Components:

  • Solidity SignalRegistry
  • Hardhat deployment scripts
  • Transaction orchestration layer

Flow:

Privacy Layer
      ↓
Signal Registration
      ↓
Platform Routing

Integrations

Minam

Normalized data feeds

Syuzhet

Narrative and thesis generation

Additional Sources

  • Wallet connectivity
  • DexScreener discovery
  • CoinGecko-style discovery
  • Nimble for niche assets

Challenges We Ran Into

Signal-First Discovery Is Difficult

LLMs naturally drift toward:

  • BTC
  • ETH
  • BNB

Solutions:

  • Keyword asset merging
  • Corpus scoring
  • Discovered asset ranking

Social Data Reliability

Problems:

  • API costs
  • Rate limits
  • CORS issues

Solution:

Built demo corpus datasets for consistent demonstrations.

Browser LLM vs Backend Observability

Pipeline LLM calls bypassed monitoring tools.

Potential solution:

  • Move LLM processing server-side
  • Add tracing

Git Hygiene

A committed Rust build artifact caused:

  • 800MB+ push
  • Large .pdb files

Lesson:

backend/target/

Local ignore ≠ removed from Git history.

UX vs Execution Scope

We removed simulated trading.

Instead:

Users open real exchange links.

Benefits:

  • Simpler UX
  • Better legal clarity
  • More realistic behavior

Accomplishments

✅ End-to-end signal → asset → execution flow

✅ Cross-asset intelligence routing

✅ Modular architecture:

  • Minam
  • Syuzhet
  • Greed

✅ Monad on-chain signal registration

✅ Privacy routing through Unlink

✅ Hackathon traction:

  • 🏆 Blockworks Permissionless IV — Winner
  • 🏆 NYC AI Tinkerers — Finalist

✅ Demo-friendly operator pipeline

What We Learned

Hypothesis-first UX changes products

Stories become as important as charts.

Agentic systems need guardrails

Required:

  • Corpus normalization
  • Evidence citation
  • Keyword fallback systems
  • Explicit demo/live separation

Cross-asset intelligence is mostly a data problem

Questions become:

  • Which symbols?
  • Which platforms?
  • Which routes?

Before:

  • Blockchain
  • Smart contracts

Rust backends are highly effective

Small Rust services:

  • Secure keys
  • Fast corroboration
  • Low overhead

But:

Build artifacts must never enter source control.

MVP execution is often simpler than expected

Deep links to real exchanges frequently provide more value than simulated trades.

What's Next

Server-Side AI + Tracing

Move analysis into Rust backend.

Potential additions:

  • Lapdog
  • Datadog spans

Richer Asset Discovery

Expand:

  • Non-major tokens
  • Prediction markets
  • Equity discovery
  • Narrative branches

Live Social Signals By Default

Improvements:

  • Better X budget allocation
  • Optional user handles

Deeper Minam + Syuzhet Integration

Goals:

  • Auto-bind feeds
  • Dynamic thesis updates
  • Signal quality scoring

Signal Markets & Performance Layers

Potential additions:

  • Bonding curves
  • Signal quality pricing
  • Historical signal performance
  • Postgres storage

Production Hardening

Move toward:

  • Mainnet deployments
  • Audits
  • Agent APIs
  • Autonomous signal creation and execution

Greed transforms narratives into structured, tradeable intelligence.

Signal → Discovery → Validation → Execution

Built With

Share this project:

Updates