The Power of AI

Pyrite already holds about 480 billion precalculated metrics. No table can display that, and few people can query it directly.

You type a question. The AI turns it into database queries, runs them, and returns the answer in seconds: a short text, and a table or chart.

You set the frame before you ask:

• The scope: one selected fund, or all funds you are entitled to see.
• The horizon: 2 (default), 3 or 5 years.
• The model: the AI model of your choice (see below).

• What are this fund's style probabilities over 3 years? Show them in a bar chart.
• Which academic factors are detected in this strategy? Show the top 10 and bottom 10.
• Compare this fund's Sharpe ratio with all funds in the same style.
• Which assets does this strategy trade (based on filtered betas by default)?
• Among CTAs exposed to copper, where does this fund's Sharpe ratio rank?
• Show the cumulative returns of these funds over the last three years, with this one in bold.

For allocators, the same approach screens the whole universe. Say you believe oil is underpriced, rates will fall and the euro is too cheap against sterling. Ask which funds carry those three exposures, sort them by Sharpe ratio, and chart their performance. One question replaces a day of spreadsheet work or emails with managers.

The 480 billion metrics are the starting point, not the limit. When the number you need is not stored, the AI computes it from the underlying data.

• The average Sharpe ratio of CTAs that trade copper.
• A fund's percentile within a peer group you define.
• A ratio or statistic the platform does not list.

Each answer states the metric, the period and the group of funds it used, so you can see exactly what was compared.

Since we are not charging for calculations, the permitted calculations remain within simple arithmetic.

Each answer comes with its own table or chart. One click saves the text, the chart, or both into your report.

You can also ask for a specific chart type, or build a chart step by step with the AI.

Pyrite offers the leading models and providers: Anthropic, Google, OpenAI, xAI, DeepSeek and Moonshot. We set the fastest, most stable model by default; you can change your default or switch for a single query.

Models differ in speed, depth and cost. In our test of nine typical queries (October 5, 2026), the median query takes 1–2 seconds and costs ~3 cents across all models. The average query takes ~6 seconds and costs ~4 cents. We train our AI to better understand challenging questions and improve efficiency.

The Research Preview includes about 100 free queries. Compare models and pricing.

• No investment advice. The AI describes funds and measures risks. It does not recommend buying, selling or holding any fund.
• No guessing. It answers from Pyrite's data. If the data cannot answer the question, it says so.
• No identification without entitlement. Research Preview and Manager users see anonymized funds only. The AI will not reveal a fund's name, manager or other identifying details to users who are not entitled to them. No comparative query will be run on a custom set of fewer than 20 funds.
• No bulk download. Our data providers and managers do not permit it. Results remain viewable in tables.
• Compliance: Investing in securities, especially in alternative investments, is regulated. Data has permissions. The AI shows only what your verified status permits.
• Your strategies stay yours. Strategies you upload remain private unless you choose to publish them. Allocators cannot reshare them. We do not manage assets, replicate strategies, or advise; we offer tools to understand them.
• Your queries stay yours. Pyrite monitors speed, token costs and any fallback by model, not the content of your analysis. Your questions are processed by the AI provider you select, under its own terms.

Like any AI, it can misread a question. When an answer looks wrong, rephrase it or narrow the frame.