Research Watch

Set up a recurring or deferred research watch on a topic, company, paper area, or product surface.

The research watch workflow sets up ongoing monitoring of a research topic. It takes a baseline sweep, then uses schedule_prompt to create a recurring or deferred follow-up — rather than merely promising to check later. New papers, articles, and developments are surfaced as they appear.

Invocation

CLI

feynman watch "<topic>"

REPL

/watch <topic>

Examples

feynman watch "new developments in state space models for sequence modeling"
feynman watch "mechanistic interpretability papers 2025"

/watch speculative decoding research
/watch "Anthropic safety papers"

Workflow stages

1. Plan

   Before starting, the lead agent outlines the watch plan: what to monitor, what signals matter, what counts as a meaningful change, and the check frequency. The plan is written to outputs/.plans/<slug>.md and presented to you for confirmation before proceeding.

2. Baseline sweep

   A baseline sweep of the topic is run immediately, capturing the current state of the research landscape. This gives the watch a reference point so that future checks surface only genuinely new material.

   The baseline is saved to outputs/<slug>-baseline.md and ends with a Sources section containing direct URLs for every source found.

3. Schedule

   schedule_prompt is used to create the recurring or delayed follow-up. This means the watch is real and scheduled — not a promise to check later. The schedule is configured according to the check frequency agreed in the plan.

   The recurring follow-up runs the same researcher logic as the baseline sweep, comparing new results against the baseline to surface only new material.

Outputs

Artifact	Path
Watch plan	outputs/.plans/<slug>.md
Baseline research snapshot	outputs/<slug>-baseline.md

Managing watches

List all active watches and their status:

/jobs

The /jobs command shows active watches along with their schedule, last check time, and number of new items found. You can pause, resume, or delete watches from within the REPL.

Watch check output

Each scheduled check produces a summary of new material since the last run:

• New papers — titles, authors, and summaries of newly discovered papers relevant to your topic
• New articles — relevant blog posts, documentation updates, or news articles
• Relevance notes — why each item was flagged as relevant to your watch topic

When to use /watch

Use /watch to stay current on a research area without manually searching every day. It is particularly useful for:

• Fast-moving fields where new papers appear frequently
• Tracking specific research groups or topics related to your own work
• Monitoring the literature while focused on other tasks
• Keeping up with a competitor’s publication record or a product area you care about

Combine /watch with Deep Research to set up a watch after an initial deep dive — the baseline sweep will align with what you already know, so only genuinely new material is surfaced.

Related

• Deep Research — run a one-time deep investigation before setting up ongoing monitoring
• Literature Review — produce a structured snapshot of the current state of a field