Whitepaper — ArtSignal

Print version Token details → Live roadmap →

Table of contents

Abstract
The problem
Methodology
Conviction scoring
Architecture
$SIGNAL token design
Governance
Accountability layer
Risks & mitigations
Future work

01Abstract

Crypto markets are saturated with noise: anonymous shills, paid promotion, and survivor-bias storytelling. Real edge comes from on-chain truth — what wallets do, not what mouths say. ArtSignal builds that edge into a transparent, verifiable signal protocol. Every signal is logged on-chain. Every miss is publicly counted. Every analyst is accountable.

02The problem

The asymmetric-information problem in crypto is severe. Insiders, market-makers, and well-connected funds operate hours or days ahead of retail. The "alpha" most retail traders consume is downstream content: by the time it arrives, the move is over.

Influencer bias: paid promotions disguised as conviction.
Selection bias: winners loud, losers silent.
Latency: social signals lag on-chain reality by 30-180 minutes on average.
Verifiability: no public, immutable record of who called what, when.

03Methodology

ArtSignal combines four primary inputs:

On-chain

Wallet flows, holder concentration, smart-money clustering, contract activity.

Social

Cross-platform sentiment (X, Telegram, Discord) weighted by wallet-verified authors.

Market

Liquidity, depth, slippage, DEX/CEX listings, derivatives positioning.

Narrative

Sector momentum, attention windows, comparable-asset performance.

04Conviction scoring

Each candidate asset receives a conviction score from 1.0 to 10.0. Scoring is a weighted ensemble of model outputs, calibrated against three years of historical Solana market data.

S = w₁·O(t) + w₂·M(t) + w₃·N(t) + w₄·R(t) where O = on-chain · M = momentum · N = narrative · R = risk-adjusted reversal weights are re-fit weekly via walk-forward cross-validation

A signal fires only when S ≥ τ, where τ is the tier-specific threshold. Threshold values are public.

Signal classes

Class	Trigger	Typical horizon
Bull	Score ≥ 7.5 with positive momentum	1–14 days
Watch	5.5 ≤ Score < 7.5	1–30 days
Caution	Score < 4.0 or risk flag elevated	Indefinite
Pre-signal	Insider tier · score 7.0+ and accumulation in early window	Variable

05Architecture

ArtSignal is built around a deterministic ingestion pipeline, a stateless scoring service, and an on-chain attestation layer.

Ingestion: Helius WebSockets + dedicated Geyser plugin parse every block within ~400ms.
Storage: append-only ClickHouse cluster; warm 90 days, cold to S3.
Inference: ONNX-runtime model servers, p99 latency 28ms.
Attestation: each fired signal hashed and recorded via memo-program on Solana within the same slot.
Distribution: tier-gated WebSocket fan-out, Telegram & Discord bots, REST/gRPC API.

06$SIGNAL token design

$SIGNAL is a Solana SPL token with a fixed supply of 100,000,000. It serves three purposes:

Tier access — staked balances unlock real-time signals, pre-signal intelligence, and API.
Revenue share — 70% of subscription, API, and integration revenue is distributed pro-rata to stakers each epoch (7 days).
Governance — stakers vote on signal-tier thresholds, fee schedules, treasury allocation, and partner integrations.

Distribution

Bucket	%	Vesting
Community rewards	35%	5-year emission curve
Ecosystem & grants	20%	Multi-sig, milestone-released
Development fund	20%	3-year linear
Team	15%	1-year cliff, 4-year linear
Reserve	10%	Locked, governance-released

Burn mechanism

10% of all monthly platform fees are programmatically burned via the SPL burn instruction. Burns are visible on-chain and reflected in real-time on the public dashboard.

07Governance

From Phase 06, $SIGNAL stakers govern via on-chain proposals using a quadratic-voting module. Quorum is 4% of circulating supply; supermajority (66%) is required for treasury motions and parameter changes affecting more than one tier simultaneously.

08Accountability layer

Every signal — bull, watch, caution, pre-signal — is hashed and stamped on-chain at the moment of emission. The hash includes asset, score, timestamp, and signer wallet. No retroactive editing is possible.

Performance is computed by deterministic, open-source scripts and re-published monthly. Bad calls are not removed from the record. Win rate, average return, max drawdown, and t-stat are all available via the public dashboard.

09Risks & mitigations

Model decay: rolling re-training plus champion/challenger evaluation.
Front-running of public signals: tiered release cadence; pre-signal channel is not broadcast until N blocks of inclusion confirmed.
Oracle/data integrity: redundant providers; consensus required across at least two independent sources.
Governance capture: quadratic voting + delegated voting + 7-day timelock on critical parameters.
Regulatory: ongoing legal review; geofencing where required.

10Future work

Cross-chain expansion (initially EVM L2s and Sui), formal verification of the attestation contract, an on-chain marketplace for community-published signal models, and an institutional offering (BAML-grade reports + co-located inference).