Draft LabLearn from posts you may use · publish what you approve

Sign in to your workspace

Paste the workspace token from WORKSPACE_TOKEN in the server's .env.local. It is stored as an 8-hour session cookie.

Radar

Find reference posts worth learning from. Candidates are ranked by how unusually well they did for their own author, with the limits of that ranking shown: never a viral score. Analyse one to see its pattern, then use it in Draft Lab.

  1. Load or refresh candidates
  2. Read the limitations
  3. Analyse a candidate
  4. Use it in Draft Lab

Watchlists

Editors create and pause monitored X topics/authors. Reads happen only when you explicitly poll; no automatic polling or draft copying.

Official X: search phrase or author handle. Treg author lists require a numeric user ID; topic lists need explicit post IDs.
1–10,000 read attempts per UTC day, including search and optional sample requests. A request can return multiple posts; paid cost caps apply separately.
Off by default. Enabling does not start a poll.
Optional comma-separated numeric IDs. Leave blank for a supported author/topic listing.

Live providers require configured credentials and human-verified prices. The local demo uses fake reads. Unknown prices/routes are refused, not free.

No watchlists yet.

Coverage gaps — advisory inspiration

Compare frozen owned-post pillar tags with exact pillar phrases in current watchlist posts. This is a lexical sample, not semantic classification or evidence.

Competitor text is read-only, reuse rights unknown. No evidence or draft is created here. Model consent is not requested here and is off by default on retrieved posts. Public visibility is not permission to copy.

Discovery inbox

Candidates collected from your watchlists, or a comparison sample to see how the inbox works.

Analyze saves this permission choice for the candidate you click, and loading a comparison sample uses it too. It resets when you reopen the page and does not show a candidate's saved consent. Local sample analysis works with this off; external model calls require it.

No candidates yet. Load the comparison sample to see how the inbox works, or refresh to show candidates collected from your X watchlists.