Anatomy of a Great Prompt

1. Universe

The universe is your starting point—the pool of stocks the AI will draw from when building your index. Defining it upfront gives the evaluation swarm a meaningful scope to work within. Without it, the system has to make assumptions, which can produce results that feel too broad or off-thesis.

Your universe can be index-based, sector-based, or defined by a characteristic. The key is that it should meaningfully narrow the field before any screening begins.

Strong universe definitions:

Weak universe definitions:

Too vague—the AI has no meaningful starting pool.

Not screenable, too broad.

If you don’t specify a universe, the system will default to a broad US equity universe—which works, but often produces less focused results than if you define one yourself.

2. Constraints

Constraints are your hard rules. They tell the system what to exclude or require based on quantifiable financial characteristics. Think of them as the guardrails of your thesis—they filter out companies that don’t meet your baseline criteria before the system starts looking for the qualities you actually want.

Good constraints are specific and grounded in fundamentals. The more clearly you define the financial profile you’re looking for, the more precisely the AI can screen for it.

Strong constraint language:

Revenue Growth

Leverage

Profitability

Margin Stability

Cash Flow

Weak constraint language:

You can stack multiple constraints in a single prompt—the system will apply them all. Just keep each one specific enough that the AI can evaluate it against real financial data.

3. Signals

Signals are the qualitative or thematic layer of your prompt—the characteristics that define your investment thesis beyond the numbers. While constraints filter out what you don’t want, signals tell the system what you’re actively looking for. This is where your view of the market comes through.

Signals can be thematic (exposure to a trend or sector), structural (how a company is run), or behavioral (how it has performed relative to a pattern). They’re often harder to reduce to a single metric, which is exactly why the AI evaluation swarm is built to assess them.

What you can signal:

Thematic exposure

Capital efficiency

Volatility profile

Leadership Structure

Growth quality

Macro sensitivity

Strong signal language:

Weak signal language:

Putting it all together

The most effective prompts combine all three components into a single, coherent thesis. The universe sets the scope, constraints set the floor, and signals define the character of the index you’re building.

Example:

This structure gives the evaluation swarm a clear, coherent thesis to operationalize—a defined pool to draw from, clear rules to filter by, and a distinct investment angle to screen for. The result is an index that actually reflects your view, not just a generic basket of large-cap names.

Refining your index: The iterative loop

Once you’ve built an index, the real magic happens in the refinement loop. You can tighten constraints, strengthen signals, or reshape the universe entirely with follow-up prompts:

With each iteration, you’re guiding the swarm—like directing a research team toward a sharper interpretation of your thesis.