back to table · edit · history · where entries came from · files · download
6669 bytes, as of the version from 2026-10-30 04:30 (current). Recorded here, not run.
# Build handoff: T617
Built **Central pressure constants of Lane–Emden polytropes**,
https://numberdb.org/T617, as a private draft. Filled ten entries once, each
with 100 significant digits. The draft has not been offered, reviewed or
published. The controller must run the separate critique and repair stages.
No independent acceptance sidecar was written by this build.
The requested machine handoff is
`agents/runs/20261030T035316Z-build.table.json`, containing only `{"tid":"T617"}`.
All working scripts and evidence are in `agents/runs/20261030T035316Z-build-work`.
## Policy and family evidence
Read issue 246 first using `python3 agents/queue.py show 246`; its saved text
is `family-246.md`. Read both archived skills and compared them with the
canonical source snapshot. The controlling manifest is
`agents/runs/20261030T035316Z-build-skills/manifest.json`, copied to
`skill-manifest.json` and attached to T617. Both the local and live versions
have SHA-256
`28ecbeeccb0c5e5d6eb541b5a5b0f88aec47b154374d909085116d34f5629199`.
The source snapshot agrees. Read `campaign-gate.md`, `argument_review.py`
and `docs/design/guarding-generated-tables.md`.
Screening used `already_here`, `already_asked`, three `api/lookup` samples,
and live tag reads. No duplicate family was found. Pi/8 appears in other
contexts, including T29; this is a theorem relating different quantities and
is not grounds for declaring `repeats`. The family companions T613–T616 were
read from the website, including their definitions, ranges and normalizations.
Actual slugs were read from their pages, not constructed from titles.
The T29 `1/8` entry was checked explicitly. Source PDFs, their hashes, a
rendered Pols Table 4.1, and original solver attachments are retained.
## Decisions
Use W_n = P_c R^4/(G M^2), with the three-dimensional, spherical, nonrotating
Newtonian model, positive real Lane–Emden branch, and first zero-pressure
surface. Equivalent formula: xi_1^4/[4*pi*(n+1)*mu_n^2], with mu_n omitting
4*pi. This matches Pols (2011), equation (4.17), and Lu, equation (6).
The n=0 model is incompressible with a surface density jump. Its value is
3/(8*pi); the n=1 value is pi/8. These are irrational closed forms, evaluated
with Arb balls. No rational anchor was replaced by a decimal.
Keep the ten proposed indices in ascending order. The only free parameter n
is mixed: intrinsic {0,1}, documented_application {3/2,3}, reference_sample
{1/2,2,5/2,7/2,4,9/2}. The six reference inputs cover simple algebraic
density–enthalpy laws from stiff to concentrated finite profiles. Denominator
two and cutoff 9/2 are explicit noncanonical choices, independent of numeral
base. The attached `argument-selection.md` gives purpose, bounds, sources and
coverage rationale. No independently sampled radius, scale or density enters.
Index five is excluded because the radius is infinite. Later mass–radius and
tidal-response quantities remain separate proposals.
The draft uses existing tags physics and dimensionless. Astrophysics remains
absent; T613, T614, T615, T616 and T617 are specific destinations that would
support that proposed tag. No new tag was created by this run.
## Validation
* `compute.py`: fresh runs at 150 and 210 working decimal digits, orders
160/210, step caps 1/12 and 1/18, matching radii 1/2 and 2/3. Maximum
relative difference across all ten models was 1.70427e-153. Both analytic
profiles were also integrated as nonzero controls.
* `independent_checks.py` and `radial_check.py`: separate radial-coordinate
integration with exact rational central coefficients and accumulated
pressure integral. Recover W using the virial identity rather than only
the surface mass. All ten pass a relative tolerance of 1e-40; the observed
maximum difference was 9.30473e-61. Displaced controls are rejected.
* Pols Table 4.1: all seven overlapping printed constants round correctly;
the largest error is 0.43952 units of the printed last place.
* Prewrite identities against the live radius, mass and density companions
pass to 98 relative decimal places, including the physical scale formula.
* The repository `dry_run.py` passes exactness, prose and size checks:
10 entries, longest returned representation 159 characters, about 2 KB.
* Fill outcome: ten added, no replacements or removals. Reproducible sources,
argument selection and numerical evidence were attached. `verify()` reports
10/10 matched, no differing, missing or extra entries.
* `readback_checks.py`: every stored 100-digit interval contains both precision
computations; analytic balls fit inside their stored intervals. The live
radius/mass and concentration identities overlap with narrow finite balls.
The independent virial identity and displaced-interval controls pass.
* Site audit: `findings: []`, `clean: true`, saved as `audit.json`.
* An unauthenticated API read cannot access T617; see `private-status.json`.
The first readback check failed because its helper treated the exact companion
C_0=1 as the decimal interval [0,2]. The cause was identified, the helper now
parses exact rational/integer text exactly, and the whole readback check passes.
No numerical table entry changed. The failed log and initial fingerprint are
preserved. The corrected check and readback evidence are attached to the draft.
This was a validation-code bug covered by the skill's existing exactness rule,
not a new skill-policy lesson.
The rigour is **heuristic (agreement-checked)**. The main runs do not rigorously
bound ODE truncation or propagated error, and the independent endpoint tail is
also heuristic. Agreement substantially exceeds the stored precision but is
not a rigorous ODE enclosure. The exact-profile calculations use Arb.
## Independent critique inputs
`fingerprint.json` was obtained with the campaign fingerprint command against
live attachments, including generate.py, solver.py, compute.py,
independent_checks.py, radial_check.py, readback_checks.py and
argument-selection.md. Its selection SHA-256 is
`44a335ce0d5aef70df84284e19426e28172c8a5cabd6bb16a178a970bc271aa1`.
The critic must obtain its own current fingerprint, independently assess all
three argument groups, run numerical checks, and use its own NUMBERDB_RUN_ID
as reviewer_run in its required sidecar. This handoff does not accept the
selection on the critic's behalf.
The website database and attachments are authoritative. No published table
was edited, no Git actions or deployment occurred, and no batch-exhausted flag
or supervisor completion receipt was written. The family queue was not marked
built at this draft-only stage. All started computations have finished.