# Binary mass transfer: fixed-source science replay

Verified environment: Python 3.12.12, NumPy 2.4.4, SciPy 1.17.1, Matplotlib 3.10.8. Install requirements in your own environment, then run from a fresh extraction directory:

```sh
python verify_package.py
python fetch_sources.py --output inputs
python reproduce.py --inputs inputs --output results
python validate_science.py
```

The downloader names seven exact original responses totaling 1,194,236 bytes. Two are bounded HTTP byte ranges from large ZIPs: ignored ranges fail before a whole archive can be read. Archive span and decompressed original member SHA/size/CRC are checked. Cache corruption fails rather than silently replacing it. No author text, observation table, native zone array or native Fortran is included in this package. Local originals retain upstream terms.

The actual producer extracts 2013 Table 1 masses and the adopted period trend from freshly hash-verified full text, then calculates the historical conditional Beta Lyrae rate and retained no-positive-rate loss scenarios. It parses the two exact native profiles, chooses original hydrogen-cut zones and neighbors, and freshly integrates six gravity costs before calculating 24 conditional energy endpoints. It reads the actual fixed MESA inlist to compare its mathematical mixed-loss reference with an independent ODE. It does not run MESA or translate its Fortran into the Lab. It writes aggregate summaries and generates three own scientific SVG/PNG figures from the fresh structure aggregate and computed Roche volumes.

The 1993 scanned PDF is a manual visual-reading source only. Check printed p139 equation 5 before transcribing the small quadratic coefficient. No OCR/native ephemeris extraction or refit of the 1532 historical mean RVs is claimed by these commands. The tutorial's hand calculation is explicitly separate. The 2013 rate is historical fixed-parameter inference, not a 2026 measurement, raw light-curve/RV fit, complete posterior or CE-efficiency measurement. Individual mass errors lack covariance; the adopted period trend lacks a trend uncertainty.

Validation actually executes independent segment-analytic versus adaptive native integrals, scalar roots/low-cost limits, slow wind, conservative mass equality, 20 angular/finite-difference volume responses and the two retained Eggleton sign failures. It then independently reintegrates 100 actual exported JavaScript geometry targets, 150 orbital balances and all saved preset streams against rotated-volume derivatives and adaptive ODEs. Those are our own native synthetic targets; public Python re-evaluates them, but does not regenerate the protected Lab Engine.

Frozen-orbit trajectories are massless pressure-free probes; their Omega flight clock is independent of imposed stripping. Orbital energy is not conserved total stellar/gas energy. Near response thresholds remain unclassified. The CE model contains donor self-gravity cost only; no internal/recombination/radiation energy, companion-envelope term, relaxed remnant radius, fluid flow or merger probability is supplied. Direct surface overlap and failure to fit an individual Roche lobe remain distinct events. Small algebraic residuals are not fluid-simulation validation or formal continuum certificates.

Native structure adaptations retain source attribution and terms: Thoul & Chidester 2022, Zenodo 7082536, CC BY 4.0; Jared Goldberg/MESA School Leuven 2025, repository commit ade5bca00e5caae97ac0b6f880f1de7ccead1553, CC BY-SA 4.0. Their derived aggregate outputs retain these respective terms. Old paper figures/tables are not republished. See native-members.json and the LICENSE file. The fixed Lau 2025 v2 reading guide is a separate literature exercise; these commands do not rerun its radiation hydrodynamics or copy its imagery.