Surprisal-Timed Collapse as filed is underdetermined: a proposition-level audit and the conditions for decidability

Surprisal-Timed Collapse (STC) was proposed as a deterministic modification of the Diósi–Penrose (DP) gravitational reduction model: a superposition reduces when the accumulated relative gravitational phase reaches unity, with no stochastic field, and the Born rule is said to follow from stationary phase of an informational path integral. We audit the filed proposal claim by claim, with each claim typed as physical, derivational, self-descriptive or mathematical, and each verdict drawn from a closed vocabulary fixed before the audit began. Six consistency obligations are tested: a defined reduction energy, operational no-signalling, an outcome rule, a time and frame standard, consistency with data, and consistency with radiation bounds. As filed, STC discharges none of the first four and not the sixth; the fifth is discharged only for a binding model of the free inputs. The reduction energy depends on a smearing length, a profile and a normalisation the text does not fix (published normalisations span a factor 8π); the two rate formulas the proposal quotes are mutually inconsistent by a factor of two; one natural reading of its locality referent permits signalling, conditional on a locally observable record of the reduction; it supplies no outcome selector, and the only recovered written version of its Born-rule argument is circular; and its exemption from spontaneous-radiation bounds is an assumption rather than a calculation. Under the audited assumptions, no qualifying experiment excludes the whole admissible domain, so the proposal is not falsified by data; the one exclusion found inside that domain is conditional on a rigid-nucleus density prescription, the normalisation and the coherent interval, and disappears if the zero-point spread of the nuclei enters the density. This is not a finding that the proposal agrees with experiment. For its proposed interferometric test, the forecasts made with the filed per-trace estimator are not reproduced, and that test is not viable as filed; other analyses of the same apparatus were not adjudicated. Three central ideas of the proposal have prior art. The filed theory is therefore underdetermined. We identify requirements that a decidable version of this framework would need to satisfy. We do not claim a falsification of the Born rule, a no-go theorem for deterministic collapse, or any experimental support for STC. The author of this audit is also the author of the original STC proposal. This is a self-audit conducted with adversarial AI assistance, not an independent external peer review. Preprint version 1, 29 October 2026 (version label). The audited text is the recovered filed version of Surprisal-Timed Collapse, identified by the reserved DOI 10.5281/zenodo.23772944. This paper was prepared with substantial assistance from AI systems, as described in its Sec. III; the author is responsible for its content. File SHA-256 ad5b63c88503b3ee61264c30bd47bd96ca20415e64e4611e36aa2ba32a380367; source SHA-256 0cf82a6de5d976d4988a6f973727e4fb68d19ade622f28f3cb36ee37a32e80f0. Known presentation defects: loose interword spacing in reference [8], and a single closing line at the top of the left column of page 2. The hashes identify the archived bytes; they do not independently verify the history the audit records.