The research system

Multi-agent system

The specialized agents that collaborate to turn opportunities into measurable evidence.

The AI is not one model making a call — it is a team of specialized agents that collaborate. Each owns a narrow job, runs on the same kernel, and feeds evidence into the scoring engine. The Memory agent threads long-term knowledge through all of them.

The agents

Agent	Responsibility
Scout	Detects opportunities — new listings, launches, emerging momentum.
Research	Builds complete investment reports for a candidate.
Quant	Creates features, indicators and statistical models.
Security	Analyzes contracts and rug-pull risk.
On-chain	Studies wallets, liquidity and blockchain activity.
Social	Measures narratives and community momentum.
Strategy	Creates and proposes new strategies.
Backtesting	Tests every hypothesis on historical data.
Debate	Attempts to invalidate a proposed trade.
Risk	Protects capital.
Memory	Maintains long-term knowledge across decisions.
Execution	Executes validated, human-governed strategies.

Early-token specialists

For the early-token domain, additional specialists handle the first seconds of a launch:

Agent	Responsibility
Early Discovery	Detects new tokens, pool creation, launchpad activity and migrations; creates initial lifecycle events.
Token Identity	Resolves identity; links contracts/symbols/profiles; detects fake copies and symbol collisions.
Launch Analysis	Evaluates launch mechanics and bonding curves; analyzes migration state; flags unusual conditions.
Liquidity	Analyzes reserves, estimates slippage, evaluates liquidity growth, detects removal events.
Wallet Intelligence	Labels creator/related wallets; identifies whales, bots and snipers; detects coordinated behavior.

How they collaborate

Scout ─┐
        ├─► Research ─► Quant ─► Security ─► On-chain ─► Social
        │
Strategy ─► Backtesting ─► Debate ─► Risk ─► (scoring engine)
                                              │
                                  Memory ◄────┴────► Execution

The scout surfaces candidates; analysts enrich them; the debate agent tries to break the trade; the risk agent protects capital; only what survives is scored.

The debate agent is adversarial on purpose

Before any proposal is scored, the debate agent actively tries to invalidate it. A trade that survives a genuine attempt to refute it is worth far more than one nobody challenged — this is how the lab avoids plausible-but-wrong conclusions.

The agents participate in building the system itself: they generate hypotheses, propose new indicators and scoring models, challenge previous assumptions, and document every decision. The AI is a research partner, not a replacement for quantitative analysis.