Research Workflows

Mahout is designed for exploratory research. These workflows are good starting points.

Asset analysis

Use asset analysis for crypto, equities, macro-linked instruments, or broad market themes. Mahout can pull historical bars from Databento, centralized-exchange history from Binance, Bybit, and OKX, and crypto market data from CoinGecko.

Ask for:

• trend and momentum
• volatility and drawdowns
• important dates or catalysts
• comparable assets
• charts and summary tables
• a view of what evidence would change the analysis

Example:

Analyze BTC from January 2023 through today. Focus on regime changes, volatility, major drawdowns, and whether the current setup looks more like trend continuation or mean reversion.

Good follow-ups:

• “Add a 50/200 moving average chart.”
• “Compare this to ETH over the same period.”
• “Summarize the three strongest risks to the thesis.”
• “Turn the key observations into a saved research note.”

Macro and rates research

Use macro research for inflation, rates, employment, fiscal data, and event timing. Mahout can query FRED (Federal Reserve Economic Data), US Treasury fiscal data, and a macro-economic calendar with point-in-time reconstruction, so “what did we know at the time” questions are answerable.

Example:

Summarize the last twelve months of US CPI and core PCE, note the major revisions, and list the macro releases scheduled for the next two weeks.

Good follow-ups:

• “Chart the series with the release dates marked.”
• “How did the dollar and BTC behave around the last four CPI prints?”
• “What did the series look like as of March, before the revisions?”

News and filings research

Mahout can search GDELT news and fetch a URL you paste into the chat. There is no general web search, so paste a link when you want a specific page read.

For US public companies, Mahout can pull SEC EDGAR filings.

Example:

Search recent news for this asset, group the coverage by theme, and fetch this article I pasted to compare its claims against the filing.

Good follow-ups:

• “Which of these claims are supported by the filing?”
• “What is the earliest date this story appeared?”
• “What should I verify manually before acting on this?”

Backtests and code-backed analysis

For quantitative checks, ask Mahout to generate and run a small analysis in its isolated code sandbox.

Example:

Backtest a simple RSI mean-reversion rule on BTC hourly candles from 2021 through 2023. Return assumptions, metrics, an equity curve, and the biggest failure modes.

Good prompts include:

• instrument or market
• time range
• frequency
• rule definition
• transaction cost assumption
• output format

Good follow-ups:

• “Rerun with 10 bps costs.”
• “Add max drawdown and win rate.”
• “Compare RSI 14 to RSI 21.”
• “Save the backtest output.”

Agents

When a question becomes a rule you want watched continuously, ask Mahout to create an agent. Describe the rule in plain language:

Create an agent that alerts me when BTC hourly realized volatility doubles versus its 30-day average while funding is negative.

The workflow:

1. Draft — Mahout interprets your description and drafts the agent, including the instruments, data sources, and schedule it will use.
2. Compile — the rule is validated and compiled into a machine-executable form. If the rule is ambiguous, uses an unsupported source, or asks for trade execution, Mahout explains the rejection so you can reword it.
3. Preview — before deploying, you can preview the agent against historical time to see when it would have fired.
4. Deploy — a deployed agent runs on its schedule, evaluates the rule against live data, and sends you an in-app notification each time it fires. You can ask Mahout to analyze a fire in chat.
5. Manage — pause, resume, or archive agents at any time. Edits to a deployed agent create a new version.

Agents watch and notify. They never place orders or move funds.

Combining sources in one ask

You do not need one prompt per data source. Mahout can merge a mixed ask — for example price history, macro releases, and news — into a single answer:

Compare BTC's reaction to the last three FOMC meetings: price action around each date, what the macro calendar showed going in, and how news coverage framed each decision.

When an answer combines sources, ask Mahout to list which sources each part came from.

Synthesis and research notes

Use synthesis when you want a concise artifact from a longer conversation.

Example:

Create a one-page research note from this session. Include thesis, evidence, risks, unresolved questions, and sources.

Ask Mahout to preserve uncertainty. Good research notes should separate facts, model interpretation, assumptions, and open questions.

Source checks

For any high-stakes answer, ask Mahout to show source provenance.

Useful follow-ups:

• “Which data sources did you use?”
• “What is stale or missing?”
• “What should I verify manually before acting?”
• “What would change your answer?”