Project Enduring Pulse — v1.2.2 Charter / website release 10 explanatory edition. Constitutional source text is unchanged. Explanatory prose and editorial notes clarify evidence limits. PULSE ranks fourth, not second, in the six weak-institution comparisons. Reduced-form counts are parameter samples, not simulated Markov paths. The superseded concise overview mixed a working-paper DOI with journal details; the current paper and reference register distinguish those editions. Historical alternatives do not supersede the Constitution.

PROJECT ENDURING PULSE
Methodology, Provenance & Evidence Status
How the framework was developed and what its testing can—and cannot—show
v1.2.2 Publication & Advancement Clarification | September 2026
1. Development method
Project Enduring Pulse is human-directed systems design developed with extensive frontier-model assistance. The project steward supplied the normative premises, architectural judgments, practical intuitions and accept/reject decisions. Language models assisted with formalization, drafting, literature synthesis, adversarial critique, consistency checking, simulation design and code generation. This provenance should be disclosed plainly. Internal model critique is useful red teaming; it is not external peer review.
2. Evidence hierarchy
Targeted mechanism tests — narrow questions about voting rules, thresholds, quorum compromise, correlated failures, demographic caps, personhood panels, resilience architectures and similar mechanisms. These are the strongest quantitative claims when assumptions are explicit and the outcome follows closely from the tested mechanism.
Conditional civilization models — integrated long-horizon simulations in which architecture-effectiveness parameters are hypotheses. They explore consequences, interactions, breakpoints and failure conditions; they do not independently validate that an architecture will deliver the parameter values assigned to it.
External empirical analogues — human evidence from intergroup contact, constitutional governance, infrastructure resilience, wealth dynamics, demography and related fields. These constrain direction and plausibility but cannot directly predict a future human-synthetic society.
Normative doctrine — equal personhood, personal sovereignty and related moral commitments are arguments and constitutional choices, not empirical findings.
3. The long-horizon model critique
A hostile review correctly noted that PULSE receives high baseline values on several institutional dimensions it was designed to provide. The 114/120 result is therefore partly the parameterization echoing through an integrated model. It remains useful conditional analysis: if the mechanisms deliver approximately their intended effects, the model explores what follows and where the package still fails. It is not evidence that those effects have been independently established.
[Editorial model context] These results assume PULSE's institutions work well; the model doesn't independently establish that. In the skeptic test, moving selected PULSE institutional effects roughly 15–20% closer to the model's neutral setting removed its average lead over the strongest liberal-democratic comparator. This does not mean reducing every PULSE assumption by that amount. It is not an empirical confidence interval, a measured safety margin or evidence that PULSE is 15–20% better.
The skeptic sweep is the required counterweight. When selected PULSE institutional-effect coefficients moved roughly 15–20% toward the model’s neutral value of 0.5, its aggregate modeled advantage over the strongest liberal-democratic comparator disappeared. Public summaries should include that sensitivity whenever they mention the headline ranking.
4. Why simulation counts are secondary
Large Monte Carlo counts reduce sampling noise inside a model; they do not repair uncertain assumptions. Trial counts belong in reproducibility and methods sections. The most credible public evidence is the specific failure discovered, mechanism compared, design change made, sensitivity exposed or uncertainty acknowledged.
5. Publication standard
Disclose AI assistance.
Do not call internal frontier-model critique peer review.
Do not call scenario outputs empirical forecasts.
Do not market total simulation counts as proof.
Publish assumptions, code and negative results.
Invite independent reproduction and expert attack.
Change the framework if external evidence or better designs warrant it.