# Independent referee of victor_VB31 The mathematical statement survives independent exact and analytic review: at the literal Table 3 immune rates, every finite nonnegative solution of printed Eqs. 21–23 converges to an equilibrium for 00. This is a mathematical result about the printed field. It supplies neither biological uncertainty bounds nor historical simulation identity. No novelty or author-interest judgment is made. ## Independent reconstruction and checks `source_audit.py` parses freshly fetched primary MathML and Table 3, recovers the three original equations and all ten fixed immune constants, and verifies the normalization exactly. Authors and a=.514 are checked from that same source; time is unchanged. `source_result.json` retains the equations, constants and source hash. The final primary XML is byte-identical to the initial retrieval. The submitted energy is used only as a proposed witness. `independent_check.py` imports no submitted scientific code: it differentiates the full witness directly in T,y,z, reconstructs A,K,Q, proves parameter affinity, derives the Bernstein matrices from power coefficients, and checks all 36 matrices over T>=0 with exact rational arithmetic/root counts. It also independently checks source bounds and the two-positive-equilibrium control. The run passed on Python 3.12.3 / SymPy 1.13.3 in 145 seconds on one CPU; `independent_result.json` retains 40 distinct integer polynomials, including 26 determinant numerators and additional A/rate/weight bounds. This count differs from the submission because the referee checks additional A and elementary-bound polynomials separately. `independent_bernstein.py` independently proves all 40 referee-derived polynomials positive by exact integer Bernstein convexity after T=10000*t/(1-t), using no Sturm or root-counting library. Maximum subdivision depth is two; only one determinant needs subdivision and it needs a single split. Thus the certificate is also verifiable by short signed-coefficient checks. The logarithmic boundary, finite-time domain entry, extinction, C1 critical case, LaSalle, and finite connected omega-limit-set arguments are independently audited in `ANALYTIC_AUDIT.md`. The control a=2, b=51/50000000000, U=11/5, V=11/100000000000 has exactly two simple positive equilibria, near T=136.6 and T=12196.1, besides the boundary equilibrium. It is outside the trivial U>=3a/2 extinction region. No trajectory scan, fitted data, held-out observations, local no-Hopf result, or presumed equilibrium uniqueness is used as evidence for universal convergence. ## Reproduction on the authorized referee pod From this claim directory, transport each independent script via `send_files.py` and the approved terminal helper. The transport uses only explicitly named claim-derived files: ```sh python3 send_files.py fetch_public.py python3 _codex/friends_podsh.py ref pod_send.sh 60 # Poll fetch_public.log until primary.xml exists and the fetch process exits. python3 send_files.py source_audit.py python3 _codex/friends_podsh.py ref pod_send.sh 60 python3 send_files.py independent_check.py python3 _codex/friends_podsh.py ref pod_send.sh 60 # Poll until independent_result.json exists and independent_check exits. python3 send_files.py independent_bernstein.py python3 _codex/friends_podsh.py ref pod_send.sh 60 ``` Each launched scientific script uses nohup, nice 10, and one library thread under `/workspace/ref_victor_VB31`; `*.pid` identifies its worker. The source parser can instead use the already archived `public/primary.xml`, eliminating network dependence. Outputs and public evidence are copied back by `receive_files.py`. `referee_evidence.tar.gz`, individual result JSONs and logs retain the relevant receipts. A missing pod pdftotext was worked around by using a public pypdf wheel under the referee directory. Neither that tooling issue nor an initial Unicode-space parsing fix affected the scientific result. ## Literature and remaining bar `PRIOR_ART.md` gives sources, URLs, bounded searches and failures. The original paper already analyzes stationary roots/bistability at a=.514; the additional conclusion here is all-orbit convergence over 0