← Playbook index

Watcher-driven outreach — one author, one paper detail, one narrow ask

Stage this page moves: → 1 and → 2. Buyer pages: workflow fit · methods supplement · the segment page that matches the paper.

The Watcher (page 18, scripts/the_watcher.py) finds recent papers and theses on digital holography and writes unreviewed candidate rows. This page turns one reviewed row into one email to one person about one specific detail of their paper, with one narrow test ask. Never a batch, never a sequence.

Procedure

  1. Monthly run (operator). The live fetch is operator-only: venv/bin/python scripts/the_watcher.py --live --out <dated>.csv. Reps never run --live; --check (fixture, no network) is for testing.
  2. Review every candidate row by hand. Drop it unless all are true:
  3. the paper or thesis is recent (last ~12 months) and uses their own holograms (not only simulation);
  4. the setup is described well enough to fill the fit check (geometry, wavelength, pitch or magnification, distance, format);
  5. the author is reachable through a published institutional address;
  6. not on the do-not-approach list, not a production-line organisation, not already owned by another rep.
  7. Pick one person — usually the corresponding author or the thesis author, not every co-author. Claim the row (one owner).
  8. Read the paper. Note one specific, checkable detail from the methods: e.g. "off-axis at 632.8 nm with a 3.45 µm sensor, reconstructed by angular spectrum after manual carrier selection". Use only what the paper states.
  9. Run the fit check with those parameters (page 24); save the response. unsupported → stop, log, no email.
  10. Send one email (template below). One follow-up after 10 days; then the row closes as no reply.

Artifact: reviewed CSV (kept/dropped with reason), the claimed row, the saved fit response, the sent email.

Template (plain text, claims-clean)

Subject: Your <journal/thesis> <year> holograms: one reconstruction-record test?

Dear Dr <Surname>,

I read "<paper title>" — in particular that you <one specific methods detail,
in the paper's own terms>. We run a hosted reconstruction service that returns,
for each frame and its settings, a result ID and an evidence record anyone can
recompute, plus a methods paragraph you can adapt.

Your setup as described falls in our <tested | needs-review> envelope
(checker: https://api.apertureapi.com/docs/workflow-fit). If useful, I could
run ONE of your frames with the parameters from the paper and send back the
result ZIP so you can judge it against your own reconstruction. No account,
no cost, and nothing is published or kept without your say.

It is research tooling: we do not claim accuracy for any physical setup yet,
which is exactly why your comparison would be informative.

<your name>
ApertureAPI

Do not add figures, comparisons with their method, or "better/faster" language. If they reply with a frame, it goes straight into the 48-hour loop (page 17).

Limits