Research controls and limits
Find commands, backtest rules, exports, and retention behavior.
Commands and questions
| Use | To… |
|---|---|
| /screen | Create a screener from your criteria |
| /thesis | Start a thesis |
| @ | Attach an available ticker or screener to the message |
| “Run this screener” | Execute the current criteria |
| “Compare the last two runs” | Review changes |
| “Schedule this weekly” | Create a recurring signal |
| “List my signals” / “Stop this signal” | Review or stop scheduled work |
| “Search my documents for…” | Search ready document sources |
| “Chart…” | Request a supported visual comparison |
Available data includes prices, company financials, earnings, dividends, news, filings, macro data, fund information, and supported technical indicators. Coverage varies by instrument and source.
X/Twitter sentiment
Read the grounding disclosure:
- Grounded in recent X posts: the analysis used a live search and reports a sample size.
- Model assessment, not live X search: the answer is not evidence of current posts.
Scores, confidence, themes, and percentage breakdowns are interpretations. Use the linked posts and caveats to check them. This is a view of discussion about a name, not a measure of business value.
Run states
| State | What to do |
|---|---|
| Pending / Running | Return to this conversation to check progress |
| Awaiting review | Read and respond to the requested review |
| Completed | Inspect results and available deliverables |
| Insufficient data | Review the missing inputs or narrow the question |
| Failed | Read the error and retry when appropriate |
Deliverables
Fresh runs provide a report and results CSV. Available exports include Markdown, CSV, and PDF through printing. Reused scheduled results may have fewer files; check the deliverables panel.
Scheduling
Daily, weekly, and monthly schedules turn screeners into standing signals. Use the schedule control or ask in chat. Review the Scheduled group and your notification preferences.
Point-in-time backtest rules
Name the dates in your first request. After completion, use From, To, and Run again to change the interval.
- Start on or after 1 January 2005.
- End in the past or present, with start before end.
- Use a window of at least 90 days.
Results are grouped by quarter end. Historical field coverage varies; today's listing universe introduces survivorship bias. This is a screener replay, not a trading simulation or a guaranteed point-in-time reconstruction of every revised data field.
AI usage
An exhausted credit balance or daily ceiling can pause AI work. A scheduled screener may still run its data filters while skipping AI evaluation. Read the run notes before treating it as equivalent to a fully evaluated result.
Understand credits and limits.
Conversation history
Inactive conversations can be archived and later removed. Check your account's retention policy or ask support before relying on a conversation as a permanent record.
Export important research. An archived conversation is not a permanent record. If an older conversation cannot reopen, use the recovery message to start a new one; do not assume a blank transcript means its saved screener also disappeared.