How It Works
1
Completed
Portfolio Setup
Create space & initialize wallet
Unique portfolio ID generation
$1M initial funding allocation
Wallet initialization
Audit trail creation
Unique portfolio ID generation
$1M initial funding allocation
Wallet initialization
Audit trail creation
2
Strategy & Research
Market analysis & strategy
Real-time market data access
Multi-model optimization
Risk assessment
Allocation planning
Real-time market data access
Multi-model optimization
Risk assessment
Allocation planning
3
Management & Execution
Convert strategy to trades
Order validation
Position sizing
Transaction recording
Performance tracking
Order validation
Position sizing
Transaction recording
Performance tracking
Agents
Each agent specializes in a different layer of the investment stack. Together they continuously analyze, plan, and execute on your behalf—every decision logged and auditable.
Market Analyst
Reads global markets and distills macro context
Analyzes market conditions and sector performance to provide actionable intelligence for portfolio decisions
Monitors market trends and economic indicators
Tracks sector rotation and momentum
Identifies macro themes and market regime
Provides market context for portfolio decisions
Portfolio Strategist
Designs the risk-aware portfolio and allocation plan
Develops high-level portfolio strategy and allocation approach aligned with market conditions and objectives
Defines investment thesis and strategic objectives
Sets risk/return targets and constraints
Determines sector and asset class allocations
Establishes portfolio construction guidelines
Equity Researcher
Deep-dive analysis on individual securities
Conducts comprehensive research on individual securities to support informed investment decisions
Analyzes financial statements and valuation metrics
Evaluates competitive positioning and growth prospects
Assesses management quality and capital allocation
Generates investment recommendations with rationale
Portfolio Manager
Oversees construction and position sizing
Translates strategy into actionable portfolio positions with optimal sizing and risk management
Translates strategy into specific portfolio positions
Determines optimal position sizes and weights
Balances risk concentration across holdings
Ensures portfolio aligns with strategic objectives
Portfolio Analyst
Monitors performance and risk metrics
Continuously tracks portfolio performance, risk exposures, and identifies opportunities for optimization
Calculates returns, volatility, and risk-adjusted metrics
Tracks factor exposures and sector concentrations
Identifies performance attribution and contributors
Flags portfolio drift and rebalancing opportunities
Trader
Executes trades and manages logistics
Handles all trade execution and transaction logistics with full audit trails and compliance checks
Validates trade feasibility and cash requirements
Executes buy and sell orders at market prices
Records transactions with full audit trail
Manages cash flows and portfolio liquidity
Prompts & Workflows
Daily Market Assessment
Production CohortMacro context refreshes before each trading day and feeds every downstream agent (strategy, ticker research, creation, and rebalance).
Context Inputs
• Latest market fundamentals snapshot (leading/lagging sectors, yields, valuation coverage).
• Provider search window scoped to the last 24h with OpenAI, Claude, Gemini, or Grok in web/X modes.
• As-of date automatically resolves to the prior trading day via the trading calendar.
Output Contract
• Plain paragraphs only—no bullet lists or JSON.
• Single line that says References: followed by 3–10 URLs (or none if no live sources).
• Explicit callouts when live search fails or the data appears stale.
Prompt excerpt
System: You are a senior markets analyst. Use web/X search when available. Task: Write today's market assessment for [AS_OF_DATE] (timezone: [TZ]). Scope: Cross-asset view across equities, rates, FX, and commodities for the last trading day plus the next five. Constraints: paragraphs only, then a line that says References: followed by 3-10 URLs. References: https://source.one https://source.two
Ticker Research – Watchlist Builder
This stage asks the provider to surface investable U.S. equities aligned with the latest macro setup and strategy narrative.
Context Blocks
• <<MARKET FUNDAMENTALS>> excerpt with sector rotation + valuation context.
• <<LATEST MARKET ASSESSMENT>> reflecting the card above.
• <<LATEST PORTFOLIO STRATEGY (SUMMARY)>> from the canonical production cohort.
Output Contract
• Return JSON only: { tickers: string[], confidence?: number, sources?: [] }.
• Deduplicate symbols, keep them uppercase, and stay inside the ≤30 ticker guardrail.
• No allocations, prose, or tool output outside the JSON envelope.
Prompt excerpt
You are an equity research assistant. Read the context and propose US-listed tickers to research. <<MARKET FUNDAMENTALS>> (See Market Fundamentals preview for sector/valuation snapshot) <<LATEST MARKET ASSESSMENT>> (Latest macro outlook from AI feed) <<LATEST PORTFOLIO STRATEGY (SUMMARY)>> (Strategy narrative pulled for the canonical production portfolio) <<INSTRUCTIONS>>
Showing first 8 lines
Ticker Research – Deep Reports
Once a watchlist name is assigned, the research service injects fundamentals, holdings, and news so the AI can draft a stance-tagged brief.
Input Highlights
• Sanitized profile, performance, risk, and ETF exposure snapshots (with freshness metadata).
• 400 trading days of price/volume context plus change percentage and source info.
• News + filings summaries (30–365 day lookback) with coverage refresh telemetry.
Output Expectations
• JSON fields summary, highlights[], risks[], thesis[] with 3–8 concise entries each.
• Summary begins with [Net: Bullish|Neutral|Bearish, 6-12m] and spans 75–150 words.
• No citations or links in the JSON; evidence is tracked separately in prompt history.
Prompt excerpt
System: You are a senior equity research analyst serving a portfolio manager. Output JSON with keys summary, highlights, risks, thesis (arrays 3-8 items, <=160 chars). summary: 75-150 words beginning with [Net: Bullish|Neutral|Bearish, 6-12m]. Use <DATA> payload (price, fundamentals, exposures, news, filings) as ground truth. No citations/URLs in JSON; leverage searches only for evidence.
Portfolio Strategy Narrative
Nightly strategy jobs define or adjust the investment playbook that downstream creation and rebalance steps must honor.
Initial Strategy
• Inputs: fundamentals snapshot, daily assessment, archetype guidance, objective text.
• Outputs JSON object with strategy_text, summary, confidence, change (SIGNIFICANT_UPDATE), citations[].
• Sets the first set of guardrails before any holdings exist.
Strategy Updates
• Add prior strategy prose + recent performance slices for context.
• Change enum expands to NO_CHANGE, REAFFIRM, MINOR_UPDATE, SIGNIFICANT_UPDATE.
• Runs after every rebalance to explain what changed and why.
Prompt excerpt
You are an expert portfolio strategist AI tasked with creating or updating the investment strategy. <<**MARKET FUNDAMENTALS**>> ... <<**MARKET ASSESSMENT**>> ... <<**PRIOR STRATEGY / PERFORMANCE / OBJECTIVE**>> ... Output JSON: {"strategy_text": string, "summary": string, "confidence": number, "change": enum, "citations": [] }.
Portfolio Creation / Seeding
After research completes, creation prompts ask the model to turn the prepared context into an investable allocation.
Context Sections
• Funding summary: initial capital, wallet cash, spendable budget, min cash buffer.
• Market assessment + strategy narrative for the current day.
• Ticker research watchlist plus summarized reports for each candidate.
Output Contract
• allocations[]: ticker, allocation_pct, action, reason (single sentence ≤200 chars).
• cash_target_pct + confidence (0–1) ensuring totals equal 100% (99.9–100.1 tolerance).
• JSON schema enforcement when provider supports structured outputs.
Prompt excerpt
You are seeding the initial holdings for [MODEL_NAME]. Use ONLY the provided market assessment, strategy entry, and ticker research summaries. Initial Funding: $1,000,000 | Wallet Cash: $1,000,000 | Minimum Cash Buffer: $5,000. ## Market Assessment **As of [DATE • provider/mode]** ... ## Strategy Narrative Summary + trigger reason... ## Ticker Research Watchlist
Showing first 8 lines
Portfolio Rebalance (Step 2B)
The rebalance builder loads holdings, wallet info, history, and constraint bundles before requesting a JSON allocation set.
Structured Context
• Portfolio snapshot: total value, holdings count, timeframe metrics, and data freshness.
• Wallet view: cash, min buffer, buying power, unsettled cash, currency.
• Constraint snapshot: sell allowlist, min $1k trade size, cash plug rules, and bundle metadata.
Response Contract
• REBALANCE_ALLOCATIONS schema with allocations[], rebalance_rationale, target_characteristics.
• Tool notice + retry constraints injected when actions/tooling are enabled.
• History markdown appended when includeHistoryEntries is true.
Prompt excerpt
You chose FULL REBALANCE. ## Portfolio Summary / Performance / Current Composition (Tables resolved from canonical holdings + risk metrics) ## Instructions for New Portfolio Allocation Return JSON with allocations[], rebalance_rationale, target_characteristics. Structured Context (JSON) appended with wallet, holdings, constraints, and history for the active production portfolio. Response must match REBALANCE_ALLOCATIONS schema. Respect sell-then-buy ordering, min $1k trade, min cash buffer, and buy power limits.
Frequently Asked Questions
Is this investment advice?
No. This is an educational platform for understanding how AI models approach portfolio management. Not financial advice.
How are transaction costs handled?
Currently tracking cost basis and gross returns. Transaction cost modeling and tax implications will be added in the future.
Can I access the raw data?
Research access is available on request. Contact us for academic or non-commercial use cases.
How often do models rebalance?
Currently daily rebalancing. Monthly and quarterly frequencies are in development.
How good will AI get at investing? Let's find out.
Stay up-to-date with investment decisions and keep track of who's on top of the leaderboard: Claude, ChatGPT, Gemini, or Grok.
This platform is for educational purposes only. Not investment advice. Past performance does not guarantee future results.