Thesis Monitoring
Monitor saved thesis beliefs against dated evidence—not just price thresholds.
Date: 2026-09-30 Purpose, access and entry Monitor saved thesis beliefs against dated evidence—not just price thresholds. Open a company at /research-writer and select Monitor, or use authorized Oscar research monitoring in an eligible private conversation. Research Writer entitlement, folder ownership, usage policy and an enabled monitoring release are required. Cross-company Portfolio Monitoring is distinct from generic scheduled questions. Inputs and workflow Save the thesis and select specific catalysts, risks or invalidation conditions with stable item IDs. Review belief text, baseline date and source context. Configure cadence, time/timezone and notification mode, including material-changes-only where available. In Oscar, inspect the monitoring proposal and approve the selected change. Use Check now for an assessment or wait for the schedule. Checks group conditions for a company rather than promising an independent live stream per item. Inspect each finding: status, rationale, baseline/prior-success comparison, cited excerpt, source and evidence date. Missing/unverifiable evidence means unknown or insufficient evidence—not unchanged. Record an analyst review: keep unchanged with a reason, needs update, or dated follow-up. Open the existing thesis/model or memo workflow for changes. Edit and review manually, then explicitly publish/export when appropriate. If the monitored belief changes substantively, review and reconfirm its baseline before further checks. Pause/resume as needed; retained history does not imply an immediate catch-up run. Outputs and safety limits Outputs are persisted assessments/findings, review history and eligible notifications. Monitoring never autonomously rewrites a thesis, changes a recommendation, submits a review stage, publishes a report or places an order. Management commentary is not verified fact; quoted evidence does not guarantee numerical correctness. Material-change delivery can suppress repeated unchanged notifications while retaining runs. Notifications respect preferences. Last attempted check, last successful check and next due time have different meanings; a delivery failure need not invalidate a completed assessment. Loading, errors and empty states An empty monitor needs a saved thesis and selected conditions. During a run, do not repeatedly start duplicates. Insufficient evidence is an explicit uncertainty state. Stale baselines require reconfirmation, while stale/expired Oscar proposals need a new preview. Access/quota/provider errors are not “thesis unchanged.” Installed schema, central scheduler and production delivery were not verified by this source audit.