# Hidden protostars: public research program

Publication date: 11 October 2026. The page introduces the accepted seven-day
question; the primary comparison has not been measured and its detailed protocol
is not yet frozen. No confirmed birth, identified population hidden-protostar
fraction, physical lifetime measurement or validated detector is asserted.

The blog post lives in `_posts/2026-10-11-hidden-protostars-research-program.md`.
The program entry is in `_data/series.yml`. Shared blog layout and other programs
are unchanged. The interactive example uses invented 20% clump occurrence and
a 0–25% control occurrence; it is explanatory subtraction, not a fitted model.

## Historical evidence

`public/data/starbirth/source-inventory.csv` binds nine byte-identical compact
outputs/reports to research repository `moiseevigor/starbirth-search`, revision
`3ae8c22bbd4bf73c68074e0cc44d63094c509e0e`.
The source reports retain their dates and historical publication notes; they do
not establish the current state of any private dashboard. This blog publication
is separate from that dashboard. No bulk surveys or archived candidate galleries
are included. The compact snapshot is not a backup of the native scientific data.

Re-export only after reviewing the source revision and every associated claim:

    python3 research/starbirth/export_evidence.py /path/to/starbirth-search

Validate the compact source bindings, quoted arithmetic, article links and
production inclusion using:

    python3 research/starbirth/check_publication.py _site_production

Render the site with its existing Docker Compose build service. Desktop/mobile,
light/dark and keyboard interaction checks supplement the source checks. No new
JavaScript library, image acquisition or model fitting is required.

## Interpretation

E15 counterpart counts are consumed historical evidence, not untouched validation.
SPICY candidate-YSO overlap shares infrared information and does not supply
independent birth labels. E11 estimator spreads and corrected-share scenarios
are descriptive sensitivity, not confidence bounds or measured lifetimes.
E27/E28 failed the two specified incremental-value checks, not every possible
image approach. N20 and c6828 remain stopped. E29–E160 remain a separate archive.

The seven-day program must register its footprint, band, match rules, quality
thresholds, estimator, spatial dependence treatment, uncertainty, decision
thresholds and cost caps before the new outcome is read. A positive, null or
insufficient-evidence conclusion is acceptable. A follow-up list contains at most
20 supported association candidates, never a promised list of newborn stars.