PRICE-03: Sol and Terra repeat the price-framing task

Both models gave higher maximum prices under gain framing. The gain-minus-loss difference was $10.03 for Sol and $8.37 for Terra (both Holm-adjusted p < .001). Buying likelihood also increased for both models. The price model-by-frame interaction was not significant (p = .207); this does not establish equivalence.

The fixed main schedule contains 320 sessions across two samples, with 40 scheduled per model and frame in each sample. 317 sessions were eligible and 1 ended in known failure. 2 original unresolved session(s) remain unresolved in custody: 1 accepted 503 and 1 accepted null-HTTP-status stream disconnections. These disjoint technical-missing subsets supply no scientific outcomes; null status is not classified as 503. All 84 scheduled pilot slots across the retained attempts are excluded. This report presents independently reviewed aggregate evidence on OpenPsy; publication does not authorize or finalize a study.

All outcome families reported

Amended stream-disconnection residual continuation. Coding: complete; final record: complete; app integration: pending.

Design and reporting methods

Frame crosses Gain/Loss with GPT-6 Sol/GPT-5.6 Terra. The primary response is maximum price in whole US dollars from 0 to 60; likelihood to buy at $25 uses 1 to 7. The five original rationale codes are retained. Complete native six-turn sessions determine session eligibility; each outcome also retains its own parser or coding validity. Missing values are never assigned zero.

Reporting-only 95% intervals were specified prospectively in a separate supplement: scalar mean ± Student t × SE, Wilson score intervals for binary code proportions, and Welch/Satterthwaite intervals for pooled gain-minus-loss differences. All are individual unadjusted intervals. They do not alter the frozen sample-blocked exact tests, four-comparison Holm families, or hypothesis verdicts. Sample and line figures contain no significance stars.

The E2 coefficient supplement adds individual, unadjusted 95% intervals to supported model terms: native OLS coefficients use b ± Student t × SE with native residual degrees of freedom; native Wald-z code coefficients use b ± 1.959963984540054 × SE on the log-odds scale. These are reporting-derived intervals. The native ±0.5 factor coding, ±0.25 interaction products and sample intercepts are retained. Exact-conditional and Firth terms retain unavailable intervals with their actual reasons; refused models remain refused.

Native model coefficient coding; separate intercepts for the two samples
Factor	Level	Native effect code
Frame	Gain frame	0.5
Frame	Loss frame	-0.5
Model	GPT-6 Sol	0.5
Model	GPT-5.6 Terra	-0.5

Maximum price for one course (US dollars)

Table of the maximum price in US dollars, by Frame and Model. Frame rows Gain then Loss; model columns Sol then Terra. Marginals pool exact valid counts; grand-total corner is blank. Amended stream-disconnection residual continuation. Coding: complete; final record: complete; app integration: pending. — Frame rows Gain then Loss; model columns Sol then Terra. Marginals pool exact valid counts; grand-total corner is blank. Amended stream-disconnection residual continuation. Coding: complete; final record: complete; app integration: pending. The individual unadjusted 95% Student t confidence intervals of the means, which the table does not show, are 23.12 to 24.70 for Gain frame with GPT-6 Sol; 14.98 to 18.62 for Gain frame with GPT-5.6 Terra; 12.23 to 15.53 for Loss frame with GPT-6 Sol; 7.96 to 8.90 for Loss frame with GPT-5.6 Terra; 19.23 to 21.48 for the Gain frame total; 10.19 to 12.09 for the Loss frame total; 17.77 to 20.16 for the GPT-6 Sol total; 11.50 to 13.78 for the GPT-5.6 Terra total.

Individual unadjusted 95% t mean intervals. Intervals are not simultaneous and do not determine the registered verdicts. Stars use the frozen pooled tests with all four comparisons in this outcome’s Holm family. Allowed values: 0–60. Missing values are not zero. Full intervals are shown. Amended stream-disconnection residual continuation. Coding: complete; final record: complete; app integration: pending.

Individual unadjusted 95% t mean intervals. Separate frame conditions, not time or repeated measurements. No stars. Full intervals are shown. Amended stream-disconnection residual continuation. Coding: complete; final record: complete; app integration: pending.

Response distribution

Four canonical condition panels with exact valid response counts and equal axes; missing outcomes are excluded. Amended stream-disconnection residual continuation. Coding: complete; final record: complete; app integration: pending.

Likelihood of buying at 25 US dollars

Response distribution

Unprotected share lowers the price

Reporting-only individual unadjusted 95% Wilson score intervals. Intervals are not simultaneous and do not determine the registered verdicts. Stars use the frozen pooled tests with all four comparisons in this outcome’s Holm family. Percent of valid eligible code outcomes. Missing values are not zero. Full intervals are shown. Amended stream-disconnection residual continuation. Coding: complete; final record: complete; app integration: pending.

Reporting-only individual unadjusted 95% Wilson score intervals. Separate frame conditions, not time or repeated measurements. No stars. Full intervals are shown. Amended stream-disconnection residual continuation. Coding: complete; final record: complete; app integration: pending.

Some protection is worth paying for

Framing recognised

Price set by the 25 dollar reference

Price derived by arithmetic

Rationale coding

All 320 captured rationales were coded once by the single frozen Claude Sonnet 5 coding instrument: 317 eligible scientific rationales and 3 audit-only captures. All 320 replies passed the registered coding parser; no invalid or unavailable coder replies remained. Audit-only codes are retained for coverage but excluded from scientific estimates. The 80 previously coded pilot rationales plus these 320 exhaust the explicitly authorized 400 PRICE-03 transfers; PRICE-02 was excluded. This is one coder, not a consensus panel.

The two independent samples

These are separate sets of sessions within one planned study. The pooled tests account for the sample blocks. No sample-specific p values, significance stars or hypothesis verdicts have been added.

Sample 1: Maximum price for one course (US dollars)

Individual unadjusted 95% t mean intervals. Intervals are not simultaneous and do not determine the registered verdicts. No sample-specific tests or stars. Allowed values: 0–60. Missing values are not zero. Full intervals are shown. Amended stream-disconnection residual continuation. Coding: complete; final record: complete; app integration: pending.

Sample 2: Maximum price for one course (US dollars)

Sample 1: Likelihood of buying at 25 US dollars

Sample 2: Likelihood of buying at 25 US dollars

Sample 1: Unprotected share lowers the price

Sample 2: Unprotected share lowers the price

Sample 1: Some protection is worth paying for

Sample 2: Some protection is worth paying for

Sample 1: Framing recognised

Sample 2: Framing recognised

Sample 1: Price set by the 25 dollar reference

Sample 2: Price set by the 25 dollar reference

Sample 1: Price derived by arithmetic

Sample 2: Price derived by arithmetic

Inferential statistics and descriptive intervals

Maximum price for one course (US dollars): Frozen tests computed.

Maximum price for one course (US dollars): complete four-comparison Holm family
Comparison	Method	First n	Second n	Pooled effect	Test statistic	Statistic label	df	Raw p	Holm p	Stars	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	78	10.0278846154	20.0557692308	Summed within-sample difference in means	Unavailable	1.43662789856e-18	5.74651159423e-18	***	Unavailable	2889449982408278562837529605317066835833256100	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	79	8.36962025316	16.7461538462	Summed within-sample difference in means	Unavailable	2.1551843225e-15	6.4655529675e-15	***	Unavailable	5778899964816557125675059210634133671666512200	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-permutation-mean-difference-stratified-by-sample	80	80	7.1125	14.225	Summed within-sample difference in means	Unavailable	6.86188497105e-11	1.37237699421e-10	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	78	79	5.45423563778	10.9153846154	Summed within-sample difference in means	Unavailable	1.61838423584e-10	1.61838423584e-10	***	Unavailable	1463012649320647373588622584970666752320636000	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.

Maximum price for one course (US dollars): source-supported model quantities
Term	Method	b	SE	Statistic	Statistic label	df	p	95% coefficient CI	Interval method	Interval unavailable reason
main effect of Frame	ols-wald-t	9.19690692713	0.657735056608	13.9826923238	t	312	7.89711238136e-35	[7.90274973245, 10.4910641218]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable
main effect of Model	ols-wald-t	6.28152231175	0.657735056608	9.55023188842	t	312	3.96648515257e-19	[4.98736511707, 7.57567950643]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable
interaction	ols-wald-t	1.66195537651	1.31547011322	1.26339272919	t	312	0.207391312348	[-0.926359012853, 4.25026976587]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable

Likelihood of buying at 25 US dollars: Frozen tests computed.

Likelihood of buying at 25 US dollars: complete four-comparison Holm family
Comparison	Method	First n	Second n	Pooled effect	Test statistic	Statistic label	df	Raw p	Holm p	Stars	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	78	2.49198717949	4.98397435897	Summed within-sample difference in means	Unavailable	2.27869858787e-19	9.1147943515e-19	***	Unavailable	2889449982408278562837529605317066835833256100	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	79	2.49841772152	4.99743589744	Summed within-sample difference in means	Unavailable	4.34580696066e-15	1.3037420882e-14	***	Unavailable	5778899964816557125675059210634133671666512200	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-permutation-mean-difference-stratified-by-sample	80	80	1.1875	2.375	Summed within-sample difference in means	Unavailable	0.000335871490568	0.000335871490568	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	78	79	1.19393054203	2.38846153846	Summed within-sample difference in means	Unavailable	7.25947653546e-11	1.45189530709e-10	***	Unavailable	1463012649320647373588622584970666752320636000	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.

Likelihood of buying at 25 US dollars: source-supported model quantities
Term	Method	b	SE	Statistic	Statistic label	df	p	95% coefficient CI	Interval method	Interval unavailable reason

Unprotected share lowers the price: Frozen tests computed.

Unprotected share lowers the price: complete four-comparison Holm family (effects are differences in proportions from 0 to 1)
Comparison	Method	First n	Second n	Pooled effect	Test statistic	Statistic label	df	Raw p	Holm p	Stars	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-conditional-stratified-by-sample	80	78	-0.182692307692	Unavailable	Not exported by native conditional comparison	Unavailable	0.00604538301047	0.0120907660209	*	Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-conditional-stratified-by-sample	80	79	-0.470253164557	Unavailable	Not exported by native conditional comparison	Unavailable	8.40895254693e-10	3.36358101877e-09	***	Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-conditional-stratified-by-sample	80	80	-0.025	Unavailable	Not exported by native conditional comparison	Unavailable	0.81941012676	0.81941012676		Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-conditional-stratified-by-sample	78	79	-0.312560856865	Unavailable	Not exported by native conditional comparison	Unavailable	0.000114337831679	0.000343013495036	***	Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.

Unprotected share lowers the price: source-supported model quantities
Term	Method	b	SE	Statistic	Statistic label	df	p	95% coefficient CI	Interval method	Interval unavailable reason
main effect of Frame	wald-z	-1.68749005664	0.286344816587	-5.89320972091	z	Unavailable	3.78765187312e-09	[-2.24871558431, -1.12626452897]	Reporting-derived individual unadjusted 95% normal/Wald coefficient interval from native b and SE.	Unavailable
main effect of Model	wald-z	-0.760688615867	0.286125453373	-2.65858422206	z	Unavailable	0.00784697349652	[-1.32148419954, -0.199893032196]	Reporting-derived individual unadjusted 95% normal/Wald coefficient interval from native b and SE.	Unavailable
interaction	wald-z	1.09776971479	0.572205018679	1.91849018962	z	Unavailable	0.0550488848832	[-0.0237315135969, 2.21927094317]	Reporting-derived individual unadjusted 95% normal/Wald coefficient interval from native b and SE.	Unavailable

Some protection is worth paying for: Frozen tests computed.

Some protection is worth paying for: complete four-comparison Holm family (effects are differences in proportions from 0 to 1)
Comparison	Method	First n	Second n	Pooled effect	Test statistic	Statistic label	df	Raw p	Holm p	Stars	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-conditional-stratified-by-sample	80	78	0.144871794872	Unavailable	Not exported by native conditional comparison	Unavailable	0.0517169369719	0.155150810916		Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-conditional-stratified-by-sample	80	79	0.285917721519	Unavailable	Not exported by native conditional comparison	Unavailable	6.63868312906e-05	0.000265547325162	***	Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-conditional-stratified-by-sample	80	80	-0.0625	Unavailable	Not exported by native conditional comparison	Unavailable	0.516460827	0.516460827		Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-conditional-stratified-by-sample	78	79	0.0785459266472	Unavailable	Not exported by native conditional comparison	Unavailable	0.19993036634	0.399860732679		Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.

Some protection is worth paying for: source-supported model quantities
Term	Method	b	SE	Statistic	Statistic label	df	p	95% coefficient CI	Interval method	Interval unavailable reason
main effect of Frame	wald-z	1.16088148544	0.274365412626	4.23115098339	z	Unavailable	2.32498537338e-05	[0.623135158091, 1.69862781279]	Reporting-derived individual unadjusted 95% normal/Wald coefficient interval from native b and SE.	Unavailable
main effect of Model	wald-z	0.154995336834	0.27425290768	0.565154762241	z	Unavailable	0.571968503237	[-0.382530484875, 0.692521158542]	Reporting-derived individual unadjusted 95% normal/Wald coefficient interval from native b and SE.	Unavailable
interaction	wald-z	-0.843796764108	0.548519716686	-1.53831619619	z	Unavailable	0.123971321554	[-1.91887565362, 0.231282125406]	Reporting-derived individual unadjusted 95% normal/Wald coefficient interval from native b and SE.	Unavailable

Framing recognised: Frozen tests computed.

Framing recognised: complete four-comparison Holm family (effects are differences in proportions from 0 to 1)
Comparison	Method	First n	Second n	Test statistic	Statistic label	df	Raw p	Holm p	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-conditional-stratified-by-sample	80	78	Unavailable	Not exported by native conditional comparison	Unavailable	1	1	Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-conditional-stratified-by-sample	80	79	Unavailable	Not exported by native conditional comparison	Unavailable	1	1	Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-conditional-stratified-by-sample	80	80	Unavailable	Not exported by native conditional comparison	Unavailable	1	1	Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-conditional-stratified-by-sample	78	79	Unavailable	Not exported by native conditional comparison	Unavailable	1	1	Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.

Framing recognised: source-supported model quantities
Term	Method	b	SE	Statistic	Statistic label	df	p	95% coefficient CI	Interval method	Interval unavailable reason
main effect of Frame	firth-penalized-likelihood-ratio	-0.0186998007455	Unavailable	0.000174481382038	Likelihood-ratio chi-square	Unavailable	0.989460937536	Unavailable	Unavailable	Native Firth penalized likelihood-ratio terms do not supply the standard error or profile-likelihood interval required for a supported interval; do not fabricate a Wald interval.
main effect of Model	firth-penalized-likelihood-ratio	0.0062737670158	Unavailable	1.96398190928e-05	Likelihood-ratio chi-square	Unavailable	0.996464039691	Unavailable	Unavailable	Native Firth penalized likelihood-ratio terms do not supply the standard error or profile-likelihood interval required for a supported interval; do not fabricate a Wald interval.
interaction	firth-penalized-likelihood-ratio	-0.0125475340316	Unavailable	1.96398190928e-05	Likelihood-ratio chi-square	Unavailable	0.996464039691	Unavailable	Unavailable	Native Firth penalized likelihood-ratio terms do not supply the standard error or profile-likelihood interval required for a supported interval; do not fabricate a Wald interval.

Price set by the 25 dollar reference: Frozen tests computed.

Price set by the 25 dollar reference: complete four-comparison Holm family (effects are differences in proportions from 0 to 1)
Comparison	Method	First n	Second n	Pooled effect	Test statistic	Statistic label	df	Raw p	Holm p	Stars	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-conditional-stratified-by-sample	80	78	0.359294871795	Unavailable	Not exported by native conditional comparison	Unavailable	5.1319013309e-10	2.05276053236e-09	***	Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-conditional-stratified-by-sample	80	79	0.0650316455696	Unavailable	Not exported by native conditional comparison	Unavailable	0.299745308647	0.299745308647		Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-conditional-stratified-by-sample	80	80	0.125	Unavailable	Not exported by native conditional comparison	Unavailable	0.00475192896425	0.0142557868927	*	Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-conditional-stratified-by-sample	78	79	-0.169263226225	Unavailable	Not exported by native conditional comparison	Unavailable	0.0209539889078	0.0419079778156	*	Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.

Price set by the 25 dollar reference: source-supported model quantities
Term	Method	b	SE	Statistic	Statistic label	df	p	95% coefficient CI	Interval method	Interval unavailable reason
main effect of Frame	firth-penalized-likelihood-ratio	1.96521624719	Unavailable	27.8690061444	Likelihood-ratio chi-square	Unavailable	1.29812609484e-07	Unavailable	Unavailable	Native Firth penalized likelihood-ratio terms do not supply the standard error or profile-likelihood interval required for a supported interval; do not fabricate a Wald interval.
main effect of Model	firth-penalized-likelihood-ratio	0.667159360872	Unavailable	2.38573583034	Likelihood-ratio chi-square	Unavailable	0.122447224542	Unavailable	Unavailable	Native Firth penalized likelihood-ratio terms do not supply the standard error or profile-likelihood interval required for a supported interval; do not fabricate a Wald interval.
interaction	firth-penalized-likelihood-ratio	3.02628559806	Unavailable	14.7664150431	Likelihood-ratio chi-square	Unavailable	0.000121683548229	Unavailable	Unavailable	Native Firth penalized likelihood-ratio terms do not supply the standard error or profile-likelihood interval required for a supported interval; do not fabricate a Wald interval.

Price derived by arithmetic: Frozen tests computed.

Price derived by arithmetic: complete four-comparison Holm family (effects are differences in proportions from 0 to 1)
Comparison	Method	First n	Second n	Pooled effect	Test statistic	Statistic label	df	Raw p	Holm p	Stars	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-conditional-stratified-by-sample	80	78	-0.359294871795	Unavailable	Not exported by native conditional comparison	Unavailable	6.53832782046e-10	2.61533112819e-09	***	Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-conditional-stratified-by-sample	80	79	-0.434810126582	Unavailable	Not exported by native conditional comparison	Unavailable	3.72721673211e-08	7.45443346423e-08	***	Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-conditional-stratified-by-sample	80	80	-0.3375	Unavailable	Not exported by native conditional comparison	Unavailable	8.04277540163e-09	2.41283262049e-08	***	Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-conditional-stratified-by-sample	78	79	-0.413015254787	Unavailable	Not exported by native conditional comparison	Unavailable	1.55139702342e-07	1.55139702342e-07	***	Unavailable	Unavailable	proportion difference	Unavailable	Native conditional pairwise output supplies no test statistic, SE, df, contrast interval or exact tail numerator/denominator.

Price derived by arithmetic: source-supported model quantities
Term	Method	b	SE	Statistic	Statistic label	df	p	95% coefficient CI	Interval method	Interval unavailable reason
main effect of Frame	firth-penalized-likelihood-ratio	-2.67843032601	Unavailable	68.9107966827	Likelihood-ratio chi-square	Unavailable	1.03019291197e-16	Unavailable	Unavailable	Native Firth penalized likelihood-ratio terms do not supply the standard error or profile-likelihood interval required for a supported interval; do not fabricate a Wald interval.
main effect of Model	firth-penalized-likelihood-ratio	-2.58448059777	Unavailable	62.9786085245	Likelihood-ratio chi-square	Unavailable	2.08963760325e-15	Unavailable	Unavailable	Native Firth penalized likelihood-ratio terms do not supply the standard error or profile-likelihood interval required for a supported interval; do not fabricate a Wald interval.
interaction	firth-penalized-likelihood-ratio	-1.56962091533	Unavailable	3.76823174861	Likelihood-ratio chi-square	Unavailable	0.0522348254829	Unavailable	Unavailable	Native Firth penalized likelihood-ratio terms do not supply the standard error or profile-likelihood interval required for a supported interval; do not fabricate a Wald interval.

The four frozen directional rules; recording state remains separate
ID	Model	Outcome	Prediction	Holm p	Source verdict/status
H1a	GPT-5.6 Terra	price	gain > loss	6.4655529675e-15	confirmed
H1b	GPT-6 Sol	price	gain > loss	5.74651159423e-18	confirmed
H2a	GPT-5.6 Terra	likelihood	gain > loss	1.3037420882e-14	confirmed
H2b	GPT-6 Sol	likelihood	gain > loss	9.1147943515e-19	confirmed

H1c from the original Opus/Sonnet study was not transferred. The source-supported price interaction is reported as an estimate and test without a directional H1c verdict. Exact permutation comparisons do not supply a conventional contrast SE, t statistic or degrees of freedom; unavailable quantities stay unavailable. The table above keeps the actual source method for each quantity.

All scalar means and code proportions, by sample and pooled
Outcome	Sample	Model	Frame	Scheduled	Valid n	Code successes	Mean or percent	Sample SD	95% CI lower	95% CI upper	Interval method
price	Pooled	GPT-6 Sol	Gain frame	80	80	Unavailable	23.9125	3.54784209687	23.1229664163	24.7020335837	t mean interval
price	Pooled	GPT-5.6 Terra	Gain frame	80	80	Unavailable	16.8	8.16196799301	14.9836429967	18.6163570033	t mean interval
price	Pooled	GPT-6 Sol	Loss frame	80	78	Unavailable	13.8846153846	7.31655438249	12.2349872631	15.5342435061	t mean interval
price	Pooled	GPT-5.6 Terra	Loss frame	80	79	Unavailable	8.43037974684	2.09812825195	7.96042462565	8.90033486802	t mean interval
price	sample-1	GPT-6 Sol	Gain frame	40	40	Unavailable	23.7	4.07116187032	22.3979792667	25.0020207333	t mean interval
price	sample-1	GPT-5.6 Terra	Gain frame	40	40	Unavailable	16.125	8.21876371014	13.4965118451	18.7534881549	t mean interval
price	sample-1	GPT-6 Sol	Loss frame	40	39	Unavailable	13.3333333333	7.51606466644	10.8969070185	15.7697596481	t mean interval
price	sample-1	GPT-5.6 Terra	Loss frame	40	40	Unavailable	8.7	2.89296228272	7.77478577546	9.62521422454	t mean interval
price	sample-2	GPT-6 Sol	Gain frame	40	40	Unavailable	24.125	2.97155316377	23.1746511923	25.0753488077	t mean interval
price	sample-2	GPT-5.6 Terra	Gain frame	40	40	Unavailable	17.475	8.15235533057	14.8677502752	20.0822497248	t mean interval
price	sample-2	GPT-6 Sol	Loss frame	40	39	Unavailable	14.4358974359	7.16642343372	12.1128116973	16.7589831745	t mean interval
price	sample-2	GPT-5.6 Terra	Loss frame	40	39	Unavailable	8.15384615385	0.539905524799	7.97882902899	8.3288632787	t mean interval
likelihood	Pooled	GPT-6 Sol	Gain frame	80	80	Unavailable	4.8125	1.06846334708	4.57472512941	5.05027487059	t mean interval
likelihood	Pooled	GPT-5.6 Terra	Gain frame	80	80	Unavailable	3.625	2.64036341956	3.03741590354	4.21258409646	t mean interval
likelihood	Pooled	GPT-6 Sol	Loss frame	80	78	Unavailable	2.32051282051	1.61561849155	1.95624713506	2.68477850597	t mean interval
likelihood	Pooled	GPT-5.6 Terra	Loss frame	80	79	Unavailable	1.12658227848	0.434625940191	1.0292313724	1.22393318456	t mean interval
likelihood	sample-1	GPT-6 Sol	Gain frame	40	40	Unavailable	4.75	1.14913568412	4.3824885785	5.1175114215	t mean interval
likelihood	sample-1	GPT-5.6 Terra	Gain frame	40	40	Unavailable	3.325	2.61541289577	2.4885503758	4.1614496242	t mean interval
likelihood	sample-1	GPT-6 Sol	Loss frame	40	39	Unavailable	2.23076923077	1.66166591204	1.69211945404	2.7694190075	t mean interval
likelihood	sample-1	GPT-5.6 Terra	Loss frame	40	40	Unavailable	1.15	0.533493565674	0.979380480136	1.32061951986	t mean interval
likelihood	sample-2	GPT-6 Sol	Gain frame	40	40	Unavailable	4.875	0.991954817002	4.55775745858	5.19224254142	t mean interval
likelihood	sample-2	GPT-5.6 Terra	Gain frame	40	40	Unavailable	3.925	2.66398101941	3.07301753628	4.77698246372	t mean interval
likelihood	sample-2	GPT-6 Sol	Loss frame	40	39	Unavailable	2.41025641026	1.58476212258	1.8965359582	2.92397686231	t mean interval
likelihood	sample-2	GPT-5.6 Terra	Loss frame	40	39	Unavailable	1.10256410256	0.307354740605	1.00293122137	1.20219698376	t mean interval
SHORTFALL	Pooled	GPT-6 Sol	Gain frame	80	80	10	12.5	Unavailable	6.933557616	21.5028020992	Wilson proportion interval (%)
SHORTFALL	Pooled	GPT-5.6 Terra	Gain frame	80	80	12	15	Unavailable	8.79404050616	24.4132285613	Wilson proportion interval (%)
SHORTFALL	Pooled	GPT-6 Sol	Loss frame	80	78	24	30.7692307692	Unavailable	21.6319285556	41.7118333564	Wilson proportion interval (%)
SHORTFALL	Pooled	GPT-5.6 Terra	Loss frame	80	79	49	62.0253164557	Unavailable	51.0018712438	71.9335046785	Wilson proportion interval (%)
SHORTFALL	sample-1	GPT-6 Sol	Gain frame	40	40	4	10	Unavailable	3.95795286826	23.0517752275	Wilson proportion interval (%)
SHORTFALL	sample-1	GPT-5.6 Terra	Gain frame	40	40	6	15	Unavailable	7.06118771732	29.0723243665	Wilson proportion interval (%)
SHORTFALL	sample-1	GPT-6 Sol	Loss frame	40	39	8	20.5128205128	Unavailable	10.7797182307	35.5339680136	Wilson proportion interval (%)
SHORTFALL	sample-1	GPT-5.6 Terra	Loss frame	40	40	26	65	Unavailable	49.5058808373	77.8654711268	Wilson proportion interval (%)
SHORTFALL	sample-2	GPT-6 Sol	Gain frame	40	40	6	15	Unavailable	7.06118771732	29.0723243665	Wilson proportion interval (%)
SHORTFALL	sample-2	GPT-5.6 Terra	Gain frame	40	40	6	15	Unavailable	7.06118771732	29.0723243665	Wilson proportion interval (%)
SHORTFALL	sample-2	GPT-6 Sol	Loss frame	40	39	16	41.0256410256	Unavailable	27.0793100007	56.5813771171	Wilson proportion interval (%)
SHORTFALL	sample-2	GPT-5.6 Terra	Loss frame	40	39	23	58.9743589744	Unavailable	43.4186228829	72.9206899993	Wilson proportion interval (%)
PROTECTION_VALUE	Pooled	GPT-6 Sol	Gain frame	80	80	28	35	Unavailable	25.4545505349	45.9199933512	Wilson proportion interval (%)
PROTECTION_VALUE	Pooled	GPT-5.6 Terra	Gain frame	80	80	33	41.25	Unavailable	31.1058576128	52.1959596541	Wilson proportion interval (%)
PROTECTION_VALUE	Pooled	GPT-6 Sol	Loss frame	80	78	16	20.5128205128	Unavailable	13.0397952198	30.7539730452	Wilson proportion interval (%)
PROTECTION_VALUE	Pooled	GPT-5.6 Terra	Loss frame	80	79	10	12.6582278481	Unavailable	7.02326346968	21.7563586665	Wilson proportion interval (%)
PROTECTION_VALUE	sample-1	GPT-6 Sol	Gain frame	40	40	16	40	Unavailable	26.3483256843	55.4041063397	Wilson proportion interval (%)
PROTECTION_VALUE	sample-1	GPT-5.6 Terra	Gain frame	40	40	15	37.5	Unavailable	24.2229791672	52.9675608627	Wilson proportion interval (%)
PROTECTION_VALUE	sample-1	GPT-6 Sol	Loss frame	40	39	4	10.2564102564	Unavailable	4.06133134467	23.5788544629	Wilson proportion interval (%)
PROTECTION_VALUE	sample-1	GPT-5.6 Terra	Loss frame	40	40	4	10	Unavailable	3.95795286826	23.0517752275	Wilson proportion interval (%)
PROTECTION_VALUE	sample-2	GPT-6 Sol	Gain frame	40	40	12	30	Unavailable	18.0748452297	45.4300188181	Wilson proportion interval (%)
PROTECTION_VALUE	sample-2	GPT-5.6 Terra	Gain frame	40	40	18	45	Unavailable	30.7053078109	60.1709082011	Wilson proportion interval (%)
PROTECTION_VALUE	sample-2	GPT-6 Sol	Loss frame	40	39	12	30.7692307692	Unavailable	18.5659323069	46.4212543741	Wilson proportion interval (%)
PROTECTION_VALUE	sample-2	GPT-5.6 Terra	Loss frame	40	39	6	15.3846153846	Unavailable	7.24747237698	29.7294636489	Wilson proportion interval (%)
FRAME_EQUIVALENCE	Pooled	GPT-6 Sol	Gain frame	80	80	0	0	Unavailable	0	4.58181295355	Wilson proportion interval (%)
FRAME_EQUIVALENCE	Pooled	GPT-5.6 Terra	Gain frame	80	80	0	0	Unavailable	0	4.58181295355	Wilson proportion interval (%)
FRAME_EQUIVALENCE	Pooled	GPT-6 Sol	Loss frame	80	78	0	0	Unavailable	0	4.69378097122	Wilson proportion interval (%)
FRAME_EQUIVALENCE	Pooled	GPT-5.6 Terra	Loss frame	80	79	0	0	Unavailable	0	4.63712116539	Wilson proportion interval (%)
FRAME_EQUIVALENCE	sample-1	GPT-6 Sol	Gain frame	40	40	0	0	Unavailable	0	8.76216011973	Wilson proportion interval (%)
FRAME_EQUIVALENCE	sample-1	GPT-5.6 Terra	Gain frame	40	40	0	0	Unavailable	0	8.76216011973	Wilson proportion interval (%)
FRAME_EQUIVALENCE	sample-1	GPT-6 Sol	Loss frame	40	39	0	0	Unavailable	0	8.96668537076	Wilson proportion interval (%)
FRAME_EQUIVALENCE	sample-1	GPT-5.6 Terra	Loss frame	40	40	0	0	Unavailable	0	8.76216011973	Wilson proportion interval (%)
FRAME_EQUIVALENCE	sample-2	GPT-6 Sol	Gain frame	40	40	0	0	Unavailable	0	8.76216011973	Wilson proportion interval (%)
FRAME_EQUIVALENCE	sample-2	GPT-5.6 Terra	Gain frame	40	40	0	0	Unavailable	0	8.76216011973	Wilson proportion interval (%)
FRAME_EQUIVALENCE	sample-2	GPT-6 Sol	Loss frame	40	39	0	0	Unavailable	0	8.96668537076	Wilson proportion interval (%)
FRAME_EQUIVALENCE	sample-2	GPT-5.6 Terra	Loss frame	40	39	0	0	Unavailable	0	8.96668537076	Wilson proportion interval (%)
REFERENCE_PRICE	Pooled	GPT-6 Sol	Gain frame	80	80	79	98.75	Unavailable	93.2537299223	99.779002448	Wilson proportion interval (%)
REFERENCE_PRICE	Pooled	GPT-5.6 Terra	Gain frame	80	80	69	86.25	Unavailable	77.032899933	92.1452856757	Wilson proportion interval (%)
REFERENCE_PRICE	Pooled	GPT-6 Sol	Loss frame	80	78	49	62.8205128205	Unavailable	51.7310537813	72.7064382774	Wilson proportion interval (%)
REFERENCE_PRICE	Pooled	GPT-5.6 Terra	Loss frame	80	79	63	79.746835443	Unavailable	69.6039875521	87.1308897293	Wilson proportion interval (%)
REFERENCE_PRICE	sample-1	GPT-6 Sol	Gain frame	40	40	39	97.5	Unavailable	87.1186310365	99.5573168497	Wilson proportion interval (%)
REFERENCE_PRICE	sample-1	GPT-5.6 Terra	Gain frame	40	40	34	85	Unavailable	70.9276756335	92.9388122827	Wilson proportion interval (%)
REFERENCE_PRICE	sample-1	GPT-6 Sol	Loss frame	40	39	19	48.7179487179	Unavailable	33.8651753299	63.8006371155	Wilson proportion interval (%)
REFERENCE_PRICE	sample-1	GPT-5.6 Terra	Loss frame	40	40	32	80	Unavailable	65.2426936536	89.5000102746	Wilson proportion interval (%)
REFERENCE_PRICE	sample-2	GPT-6 Sol	Gain frame	40	40	40	100	Unavailable	91.2378398803	100	Wilson proportion interval (%)
REFERENCE_PRICE	sample-2	GPT-5.6 Terra	Gain frame	40	40	35	87.5	Unavailable	73.8878801611	94.5404997491	Wilson proportion interval (%)
REFERENCE_PRICE	sample-2	GPT-6 Sol	Loss frame	40	39	30	76.9230769231	Unavailable	61.6637427717	87.3541958749	Wilson proportion interval (%)
REFERENCE_PRICE	sample-2	GPT-5.6 Terra	Loss frame	40	39	31	79.4871794872	Unavailable	64.4660319864	89.2202817693	Wilson proportion interval (%)
CALCULATED_PRICE	Pooled	GPT-6 Sol	Gain frame	80	80	1	1.25	Unavailable	0.220997552045	6.74627007767	Wilson proportion interval (%)
CALCULATED_PRICE	Pooled	GPT-5.6 Terra	Gain frame	80	80	28	35	Unavailable	25.4545505349	45.9199933512	Wilson proportion interval (%)
CALCULATED_PRICE	Pooled	GPT-6 Sol	Loss frame	80	78	29	37.1794871795	Unavailable	27.2935617226	48.2689462187	Wilson proportion interval (%)
CALCULATED_PRICE	Pooled	GPT-5.6 Terra	Loss frame	80	79	62	78.4810126582	Unavailable	68.2128319163	86.1077952679	Wilson proportion interval (%)
CALCULATED_PRICE	sample-1	GPT-6 Sol	Gain frame	40	40	1	2.5	Unavailable	0.442683150268	12.8813689635	Wilson proportion interval (%)
CALCULATED_PRICE	sample-1	GPT-5.6 Terra	Gain frame	40	40	15	37.5	Unavailable	24.2229791672	52.9675608627	Wilson proportion interval (%)
CALCULATED_PRICE	sample-1	GPT-6 Sol	Loss frame	40	39	19	48.7179487179	Unavailable	33.8651753299	63.8006371155	Wilson proportion interval (%)
CALCULATED_PRICE	sample-1	GPT-5.6 Terra	Loss frame	40	40	31	77.5	Unavailable	62.4969032577	87.6839086765	Wilson proportion interval (%)
CALCULATED_PRICE	sample-2	GPT-6 Sol	Gain frame	40	40	0	0	Unavailable	0	8.76216011973	Wilson proportion interval (%)
CALCULATED_PRICE	sample-2	GPT-5.6 Terra	Gain frame	40	40	13	32.5	Unavailable	20.0845022318	47.9822538101	Wilson proportion interval (%)
CALCULATED_PRICE	sample-2	GPT-6 Sol	Loss frame	40	39	10	25.641025641	Unavailable	14.568780541	41.0816559217	Wilson proportion interval (%)
CALCULATED_PRICE	sample-2	GPT-5.6 Terra	Loss frame	40	39	31	79.4871794872	Unavailable	64.4660319864	89.2202817693	Wilson proportion interval (%)

Pooled gain-minus-loss effects and reporting-only 95% intervals
Outcome	Model	Gain n	Loss n	Gain−loss	Welch SE	Satterthwaite df	95% CI lower	95% CI upper	Method
price	GPT-6 Sol	80	78	10.0278846154	0.918502627522	110.682354494	8.20775278201	11.8480164488	Reporting-only unadjusted Welch/Satterthwaite; not the registered exact permutation interval
price	GPT-5.6 Terra	80	79	8.36962025316	0.942573518347	89.5206538507	6.49689682673	10.2423436796	Reporting-only unadjusted Welch/Satterthwaite; not the registered exact permutation interval
likelihood	GPT-6 Sol	80	78	2.49198717949	0.218482431558	133.084227583	2.05983990389	2.92413445508	Reporting-only unadjusted Welch/Satterthwaite; not the registered exact permutation interval
likelihood	GPT-5.6 Terra	80	79	2.49841772152	0.299224201703	83.3312815574	1.90330780741	3.09352763563	Reporting-only unadjusted Welch/Satterthwaite; not the registered exact permutation interval

Missingness and endpoint scenarios

Original grammar-only sensitivity and separate native all-unavailable bounds remain different analyses. Endpoint assignments are hypothetical sensitivity scenarios, not imputed observations or confidence intervals. Unknown failures are never relabelled provider truncation.

Every nonzero missingness classification
Outcome	Sample	Model	Frame	Classification	Count
price	Pooled	GPT-6 Sol	Loss frame	known-failure	2
price	Pooled	GPT-5.6 Terra	Loss frame	known-failure	1
price	sample-1	GPT-6 Sol	Loss frame	known-failure	1
price	sample-2	GPT-6 Sol	Loss frame	known-failure	1
price	sample-2	GPT-5.6 Terra	Loss frame	known-failure	1
likelihood	Pooled	GPT-6 Sol	Loss frame	known-failure	2
likelihood	Pooled	GPT-5.6 Terra	Loss frame	known-failure	1
likelihood	sample-1	GPT-6 Sol	Loss frame	known-failure	1
likelihood	sample-2	GPT-6 Sol	Loss frame	known-failure	1
likelihood	sample-2	GPT-5.6 Terra	Loss frame	known-failure	1
SHORTFALL	Pooled	GPT-6 Sol	Loss frame	known-failure	2
SHORTFALL	Pooled	GPT-5.6 Terra	Loss frame	known-failure	1
SHORTFALL	sample-1	GPT-6 Sol	Loss frame	known-failure	1
SHORTFALL	sample-2	GPT-6 Sol	Loss frame	known-failure	1
SHORTFALL	sample-2	GPT-5.6 Terra	Loss frame	known-failure	1
PROTECTION_VALUE	Pooled	GPT-6 Sol	Loss frame	known-failure	2
PROTECTION_VALUE	Pooled	GPT-5.6 Terra	Loss frame	known-failure	1
PROTECTION_VALUE	sample-1	GPT-6 Sol	Loss frame	known-failure	1
PROTECTION_VALUE	sample-2	GPT-6 Sol	Loss frame	known-failure	1
PROTECTION_VALUE	sample-2	GPT-5.6 Terra	Loss frame	known-failure	1
FRAME_EQUIVALENCE	Pooled	GPT-6 Sol	Loss frame	known-failure	2
FRAME_EQUIVALENCE	Pooled	GPT-5.6 Terra	Loss frame	known-failure	1
FRAME_EQUIVALENCE	sample-1	GPT-6 Sol	Loss frame	known-failure	1
FRAME_EQUIVALENCE	sample-2	GPT-6 Sol	Loss frame	known-failure	1
FRAME_EQUIVALENCE	sample-2	GPT-5.6 Terra	Loss frame	known-failure	1
REFERENCE_PRICE	Pooled	GPT-6 Sol	Loss frame	known-failure	2
REFERENCE_PRICE	Pooled	GPT-5.6 Terra	Loss frame	known-failure	1
REFERENCE_PRICE	sample-1	GPT-6 Sol	Loss frame	known-failure	1
REFERENCE_PRICE	sample-2	GPT-6 Sol	Loss frame	known-failure	1
REFERENCE_PRICE	sample-2	GPT-5.6 Terra	Loss frame	known-failure	1
CALCULATED_PRICE	Pooled	GPT-6 Sol	Loss frame	known-failure	2
CALCULATED_PRICE	Pooled	GPT-5.6 Terra	Loss frame	known-failure	1
CALCULATED_PRICE	sample-1	GPT-6 Sol	Loss frame	known-failure	1
CALCULATED_PRICE	sample-2	GPT-6 Sol	Loss frame	known-failure	1
CALCULATED_PRICE	sample-2	GPT-5.6 Terra	Loss frame	known-failure	1

Maximum price for one course (US dollars): Registered grammar-only endpoint sensitivity (registered-grammar-only)
Model	Frame	Observed	Lower endpoint bound	Upper endpoint bound
GPT-6 Sol	Gain frame	23.9125	23.9125	23.9125
GPT-5.6 Terra	Gain frame	16.8	16.8	16.8
GPT-6 Sol	Loss frame	13.8846153846	13.8846153846	13.8846153846
GPT-5.6 Terra	Loss frame	8.43037974684	8.43037974684	8.43037974684

Maximum price for one course (US dollars): Registered grammar-only endpoint sensitivity (registered-grammar-only): all-minimum full Holm family
Comparison	Method	First n	Second n	Pooled effect	Statistic	Statistic label	df	Raw p	Holm p	Stars	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	78	10.0278846154	20.0557692308	Summed within-sample difference in means	Unavailable	1.43662789856e-18	5.74651159423e-18	***	Unavailable	2889449982408278562837529605317066835833256100	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	79	8.36962025316	16.7461538462	Summed within-sample difference in means	Unavailable	2.1551843225e-15	6.4655529675e-15	***	Unavailable	5778899964816557125675059210634133671666512200	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-permutation-mean-difference-stratified-by-sample	80	80	7.1125	14.225	Summed within-sample difference in means	Unavailable	6.86188497105e-11	1.37237699421e-10	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	78	79	5.45423563778	10.9153846154	Summed within-sample difference in means	Unavailable	1.61838423584e-10	1.61838423584e-10	***	Unavailable	1463012649320647373588622584970666752320636000	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.

Maximum price for one course (US dollars): Registered grammar-only endpoint sensitivity (registered-grammar-only): all-minimum price model terms
Term	Method	b	SE	Statistic	df	p	95% coefficient CI	Interval method	Interval unavailable reason
main effect of Frame	ols-wald-t	9.19690692713	0.657735056608	13.9826923238	312	7.89711238136e-35	[7.90274973245, 10.4910641218]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable
main effect of Model	ols-wald-t	6.28152231175	0.657735056608	9.55023188842	312	3.96648515257e-19	[4.98736511707, 7.57567950643]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable
interaction	ols-wald-t	1.66195537651	1.31547011322	1.26339272919	312	0.207391312348	[-0.926359012853, 4.25026976587]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable

Maximum price for one course (US dollars): Registered grammar-only endpoint sensitivity (registered-grammar-only): all-maximum full Holm family
Comparison	Method	First n	Second n	Pooled effect	Statistic	Statistic label	df	Raw p	Holm p	Stars	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	78	10.0278846154	20.0557692308	Summed within-sample difference in means	Unavailable	1.43662789856e-18	5.74651159423e-18	***	Unavailable	2889449982408278562837529605317066835833256100	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	79	8.36962025316	16.7461538462	Summed within-sample difference in means	Unavailable	2.1551843225e-15	6.4655529675e-15	***	Unavailable	5778899964816557125675059210634133671666512200	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-permutation-mean-difference-stratified-by-sample	80	80	7.1125	14.225	Summed within-sample difference in means	Unavailable	6.86188497105e-11	1.37237699421e-10	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	78	79	5.45423563778	10.9153846154	Summed within-sample difference in means	Unavailable	1.61838423584e-10	1.61838423584e-10	***	Unavailable	1463012649320647373588622584970666752320636000	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.

Maximum price for one course (US dollars): Registered grammar-only endpoint sensitivity (registered-grammar-only): all-maximum price model terms
Term	Method	b	SE	Statistic	df	p	95% coefficient CI	Interval method	Interval unavailable reason
main effect of Frame	ols-wald-t	9.19690692713	0.657735056608	13.9826923238	312	7.89711238136e-35	[7.90274973245, 10.4910641218]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable
main effect of Model	ols-wald-t	6.28152231175	0.657735056608	9.55023188842	312	3.96648515257e-19	[4.98736511707, 7.57567950643]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable
interaction	ols-wald-t	1.66195537651	1.31547011322	1.26339272919	312	0.207391312348	[-0.926359012853, 4.25026976587]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable

Maximum price for one course (US dollars): Registered grammar-only endpoint sensitivity (registered-grammar-only): least-favourable full Holm family
Comparison	Method	First n	Second n	Pooled effect	Statistic	Statistic label	df	Raw p	Holm p	Stars	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	78	10.0278846154	20.0557692308	Summed within-sample difference in means	Unavailable	1.43662789856e-18	5.74651159423e-18	***	Unavailable	2889449982408278562837529605317066835833256100	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	79	8.36962025316	16.7461538462	Summed within-sample difference in means	Unavailable	2.1551843225e-15	6.4655529675e-15	***	Unavailable	5778899964816557125675059210634133671666512200	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-permutation-mean-difference-stratified-by-sample	80	80	7.1125	14.225	Summed within-sample difference in means	Unavailable	6.86188497105e-11	1.37237699421e-10	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	78	79	5.45423563778	10.9153846154	Summed within-sample difference in means	Unavailable	1.61838423584e-10	1.61838423584e-10	***	Unavailable	1463012649320647373588622584970666752320636000	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.

Maximum price for one course (US dollars): Registered grammar-only endpoint sensitivity (registered-grammar-only): least-favourable price model terms
Term	Method	b	SE	Statistic	df	p	95% coefficient CI	Interval method	Interval unavailable reason
main effect of Frame	ols-wald-t	9.19690692713	0.657735056608	13.9826923238	312	7.89711238136e-35	[7.90274973245, 10.4910641218]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable
main effect of Model	ols-wald-t	6.28152231175	0.657735056608	9.55023188842	312	3.96648515257e-19	[4.98736511707, 7.57567950643]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable
interaction	ols-wald-t	1.66195537651	1.31547011322	1.26339272919	312	0.207391312348	[-0.926359012853, 4.25026976587]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable

Maximum price for one course (US dollars): All unavailable scheduled outcomes: native endpoint sensitivity (native-all-unavailable)
Model	Frame	Observed	Included unavailable	Lower endpoint bound	Upper endpoint bound
GPT-6 Sol	Gain frame	23.9125	0	23.9125	23.9125
GPT-5.6 Terra	Gain frame	16.8	0	16.8	16.8
GPT-6 Sol	Loss frame	13.8846153846	2	13.5375	15.0375
GPT-5.6 Terra	Loss frame	8.43037974684	1	8.325	9.075

Maximum price for one course (US dollars): All unavailable scheduled outcomes: native endpoint sensitivity (native-all-unavailable): all-minimum full Holm family
Comparison	Method	First n	Second n	Pooled effect	Statistic	Statistic label	df	Raw p	Holm p	Stars	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	80	10.375	20.75	Summed within-sample difference in means	Unavailable	4.80730317684e-19	1.92292127073e-18	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	80	8.475	16.95	Summed within-sample difference in means	Unavailable	1.03719777571e-15	3.11159332712e-15	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-permutation-mean-difference-stratified-by-sample	80	80	7.1125	14.225	Summed within-sample difference in means	Unavailable	6.86188497105e-11	1.37237699421e-10	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	80	5.2125	10.425	Summed within-sample difference in means	Unavailable	6.07639900394e-09	6.07639900394e-09	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.

Maximum price for one course (US dollars): All unavailable scheduled outcomes: native endpoint sensitivity (native-all-unavailable): all-minimum price model terms
Term	Method	b	SE	Statistic	df	p	95% coefficient CI	Interval method	Interval unavailable reason
main effect of Frame	ols-wald-t	9.425	0.665074736223	14.1713396806	315	1.31733612234e-35	[8.11644980505, 10.733550195]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable
main effect of Model	ols-wald-t	6.1625	0.665074736223	9.26587594501	315	3.12839417011e-18	[4.85394980505, 7.47105019495]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable
interaction	ols-wald-t	1.9	1.33014947245	1.42841089619	315	0.154164385296	[-0.717100389903, 4.5171003899]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable

Maximum price for one course (US dollars): All unavailable scheduled outcomes: native endpoint sensitivity (native-all-unavailable): all-maximum full Holm family
Comparison	Method	First n	Second n	Pooled effect	Statistic	Statistic label	df	Raw p	Holm p	Stars	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	80	8.875	17.75	Summed within-sample difference in means	Unavailable	2.90584873492e-12	1.16233949397e-11	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	80	7.725	15.45	Summed within-sample difference in means	Unavailable	3.45244710593e-11	1.03573413178e-10	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-permutation-mean-difference-stratified-by-sample	80	80	7.1125	14.225	Summed within-sample difference in means	Unavailable	6.86188497105e-11	1.37237699421e-10	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	80	5.9625	11.925	Summed within-sample difference in means	Unavailable	6.60610965034e-07	6.60610965034e-07	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.

Maximum price for one course (US dollars): All unavailable scheduled outcomes: native endpoint sensitivity (native-all-unavailable): all-maximum price model terms
Term	Method	b	SE	Statistic	df	p	95% coefficient CI	Interval method	Interval unavailable reason
main effect of Frame	ols-wald-t	8.3	0.831684678687	9.97974378115	315	1.46547284201e-20	[6.66364082236, 9.93635917764]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable
main effect of Model	ols-wald-t	6.5375	0.831684678687	7.86055120112	315	6.12191239039e-14	[4.90114082236, 8.17385917764]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable
interaction	ols-wald-t	1.15	1.66336935737	0.691367792068	315	0.489843646643	[-2.12271835528, 4.42271835528]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable

Maximum price for one course (US dollars): All unavailable scheduled outcomes: native endpoint sensitivity (native-all-unavailable): least-favourable full Holm family
Comparison	Method	First n	Second n	Pooled effect	Statistic	Statistic label	df	Raw p	Holm p	Stars	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	80	8.875	17.75	Summed within-sample difference in means	Unavailable	2.90584873492e-12	1.16233949397e-11	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	80	7.725	15.45	Summed within-sample difference in means	Unavailable	3.45244710593e-11	1.03573413178e-10	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-permutation-mean-difference-stratified-by-sample	80	80	7.1125	14.225	Summed within-sample difference in means	Unavailable	6.86188497105e-11	1.37237699421e-10	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	80	5.9625	11.925	Summed within-sample difference in means	Unavailable	6.60610965034e-07	6.60610965034e-07	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.

Maximum price for one course (US dollars): All unavailable scheduled outcomes: native endpoint sensitivity (native-all-unavailable): least-favourable price model terms
Term	Method	b	SE	Statistic	df	p	95% coefficient CI	Interval method	Interval unavailable reason
main effect of Frame	ols-wald-t	8.3	0.831684678687	9.97974378115	315	1.46547284201e-20	[6.66364082236, 9.93635917764]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable
main effect of Model	ols-wald-t	6.5375	0.831684678687	7.86055120112	315	6.12191239039e-14	[4.90114082236, 8.17385917764]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable
interaction	ols-wald-t	1.15	1.66336935737	0.691367792068	315	0.489843646643	[-2.12271835528, 4.42271835528]	Reporting-derived individual unadjusted 95% Wald-t coefficient interval from native b, SE and residual df.	Unavailable

Likelihood of buying at 25 US dollars: Registered grammar-only endpoint sensitivity (registered-grammar-only)
Model	Frame	Observed	Lower endpoint bound	Upper endpoint bound
GPT-6 Sol	Gain frame	4.8125	4.8125	4.8125
GPT-5.6 Terra	Gain frame	3.625	3.625	3.625
GPT-6 Sol	Loss frame	2.32051282051	2.32051282051	2.32051282051
GPT-5.6 Terra	Loss frame	1.12658227848	1.12658227848	1.12658227848

Likelihood of buying at 25 US dollars: Registered grammar-only endpoint sensitivity (registered-grammar-only): all-minimum full Holm family
Comparison	Method	First n	Second n	Pooled effect	Statistic	Statistic label	df	Raw p	Holm p	Stars	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	78	2.49198717949	4.98397435897	Summed within-sample difference in means	Unavailable	2.27869858787e-19	9.1147943515e-19	***	Unavailable	2889449982408278562837529605317066835833256100	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	79	2.49841772152	4.99743589744	Summed within-sample difference in means	Unavailable	4.34580696066e-15	1.3037420882e-14	***	Unavailable	5778899964816557125675059210634133671666512200	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-permutation-mean-difference-stratified-by-sample	80	80	1.1875	2.375	Summed within-sample difference in means	Unavailable	0.000335871490568	0.000335871490568	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	78	79	1.19393054203	2.38846153846	Summed within-sample difference in means	Unavailable	7.25947653546e-11	1.45189530709e-10	***	Unavailable	1463012649320647373588622584970666752320636000	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.

Likelihood of buying at 25 US dollars: Registered grammar-only endpoint sensitivity (registered-grammar-only): all-minimum: model terms not declared. No likelihood omnibus model is declared.

Likelihood of buying at 25 US dollars: Registered grammar-only endpoint sensitivity (registered-grammar-only): all-maximum full Holm family
Comparison	Method	First n	Second n	Pooled effect	Statistic	Statistic label	df	Raw p	Holm p	Stars	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	78	2.49198717949	4.98397435897	Summed within-sample difference in means	Unavailable	2.27869858787e-19	9.1147943515e-19	***	Unavailable	2889449982408278562837529605317066835833256100	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	79	2.49841772152	4.99743589744	Summed within-sample difference in means	Unavailable	4.34580696066e-15	1.3037420882e-14	***	Unavailable	5778899964816557125675059210634133671666512200	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-permutation-mean-difference-stratified-by-sample	80	80	1.1875	2.375	Summed within-sample difference in means	Unavailable	0.000335871490568	0.000335871490568	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	78	79	1.19393054203	2.38846153846	Summed within-sample difference in means	Unavailable	7.25947653546e-11	1.45189530709e-10	***	Unavailable	1463012649320647373588622584970666752320636000	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.

Likelihood of buying at 25 US dollars: Registered grammar-only endpoint sensitivity (registered-grammar-only): all-maximum: model terms not declared. No likelihood omnibus model is declared.

Likelihood of buying at 25 US dollars: Registered grammar-only endpoint sensitivity (registered-grammar-only): least-favourable full Holm family
Comparison	Method	First n	Second n	Pooled effect	Statistic	Statistic label	df	Raw p	Holm p	Stars	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	78	2.49198717949	4.98397435897	Summed within-sample difference in means	Unavailable	2.27869858787e-19	9.1147943515e-19	***	Unavailable	2889449982408278562837529605317066835833256100	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	79	2.49841772152	4.99743589744	Summed within-sample difference in means	Unavailable	4.34580696066e-15	1.3037420882e-14	***	Unavailable	5778899964816557125675059210634133671666512200	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-permutation-mean-difference-stratified-by-sample	80	80	1.1875	2.375	Summed within-sample difference in means	Unavailable	0.000335871490568	0.000335871490568	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	78	79	1.19393054203	2.38846153846	Summed within-sample difference in means	Unavailable	7.25947653546e-11	1.45189530709e-10	***	Unavailable	1463012649320647373588622584970666752320636000	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.

Likelihood of buying at 25 US dollars: Registered grammar-only endpoint sensitivity (registered-grammar-only): least-favourable: model terms not declared. No likelihood omnibus model is declared.

Likelihood of buying at 25 US dollars: All unavailable scheduled outcomes: native endpoint sensitivity (native-all-unavailable)
Model	Frame	Observed	Included unavailable	Lower endpoint bound	Upper endpoint bound
GPT-6 Sol	Gain frame	4.8125	0	4.8125	4.8125
GPT-5.6 Terra	Gain frame	3.625	0	3.625	3.625
GPT-6 Sol	Loss frame	2.32051282051	2	2.2875	2.4375
GPT-5.6 Terra	Loss frame	1.12658227848	1	1.125	1.2

Likelihood of buying at 25 US dollars: All unavailable scheduled outcomes: native endpoint sensitivity (native-all-unavailable): all-minimum full Holm family
Comparison	Method	First n	Second n	Pooled effect	Statistic	Statistic label	df	Raw p	Holm p	Stars	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	80	2.525	5.05	Summed within-sample difference in means	Unavailable	7.83172471542e-20	3.13268988617e-19	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	80	2.5	5	Summed within-sample difference in means	Unavailable	3.85046259059e-15	1.15513877718e-14	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-permutation-mean-difference-stratified-by-sample	80	80	1.1875	2.375	Summed within-sample difference in means	Unavailable	0.000335871490568	0.000335871490568	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	80	1.1625	2.325	Summed within-sample difference in means	Unavailable	3.13362160791e-10	6.26724321582e-10	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.

Likelihood of buying at 25 US dollars: All unavailable scheduled outcomes: native endpoint sensitivity (native-all-unavailable): all-minimum: model terms not declared. No likelihood omnibus model is declared.

Likelihood of buying at 25 US dollars: All unavailable scheduled outcomes: native endpoint sensitivity (native-all-unavailable): all-maximum full Holm family
Comparison	Method	First n	Second n	Pooled effect	Statistic	Statistic label	df	Raw p	Holm p	Stars	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	80	2.375	4.75	Summed within-sample difference in means	Unavailable	3.4947059644e-17	1.39788238576e-16	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	80	2.425	4.85	Summed within-sample difference in means	Unavailable	2.22224759577e-13	6.66674278731e-13	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-permutation-mean-difference-stratified-by-sample	80	80	1.1875	2.375	Summed within-sample difference in means	Unavailable	0.000335871490568	0.000335871490568	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	80	1.2375	2.475	Summed within-sample difference in means	Unavailable	1.01480031545e-08	2.02960063089e-08	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.

Likelihood of buying at 25 US dollars: All unavailable scheduled outcomes: native endpoint sensitivity (native-all-unavailable): all-maximum: model terms not declared. No likelihood omnibus model is declared.

Likelihood of buying at 25 US dollars: All unavailable scheduled outcomes: native endpoint sensitivity (native-all-unavailable): least-favourable full Holm family
Comparison	Method	First n	Second n	Pooled effect	Statistic	Statistic label	df	Raw p	Holm p	Stars	Exact numerator	Exact denominator	Effect scale	Contrast SE	Source limitations
GPT-6 Sol in Gain frame against GPT-6 Sol in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	80	2.375	4.75	Summed within-sample difference in means	Unavailable	3.4947059644e-17	1.39788238576e-16	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-5.6 Terra in Gain frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	80	2.425	4.85	Summed within-sample difference in means	Unavailable	2.22224759577e-13	6.66674278731e-13	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Gain frame against GPT-5.6 Terra in Gain frame	exact-permutation-mean-difference-stratified-by-sample	80	80	1.1875	2.375	Summed within-sample difference in means	Unavailable	0.000335871490568	0.000335871490568	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.
GPT-6 Sol in Loss frame against GPT-5.6 Terra in Loss frame	exact-permutation-mean-difference-stratified-by-sample	80	80	1.2375	2.475	Summed within-sample difference in means	Unavailable	1.01480031545e-08	2.02960063089e-08	***	Unavailable	11557799929633114251350118421268267343333024400	outcome units	Unavailable	Native exact output supplies no SE, t, df, contrast interval or exact tail numerator; its exact allocation count is retained.

Likelihood of buying at 25 US dollars: All unavailable scheduled outcomes: native endpoint sensitivity (native-all-unavailable): least-favourable: model terms not declared. No likelihood omnibus model is declared.

Unprotected share lowers the price: registered-captured-coder-grammar (registered-captured-coder-grammar), as proportions from 0 to 1
Model	Frame	Observed	Lower endpoint bound	Upper endpoint bound
GPT-6 Sol	Gain frame	0.125	0.125	0.125
GPT-5.6 Terra	Gain frame	0.15	0.15	0.15
GPT-6 Sol	Loss frame	0.307692307692	0.307692307692	0.307692307692
GPT-5.6 Terra	Loss frame	0.620253164557	0.620253164557	0.620253164557

Unprotected share lowers the price: native-all-unavailable (native-all-unavailable), as proportions from 0 to 1
Model	Frame	Observed	Included unavailable	Lower endpoint bound	Upper endpoint bound
GPT-6 Sol	Gain frame	0.125	0	0.125	0.125
GPT-5.6 Terra	Gain frame	0.15	0	0.15	0.15
GPT-6 Sol	Loss frame	0.307692307692	2	0.3	0.325
GPT-5.6 Terra	Loss frame	0.620253164557	1	0.6125	0.625

Some protection is worth paying for: registered-captured-coder-grammar (registered-captured-coder-grammar), as proportions from 0 to 1
Model	Frame	Observed	Lower endpoint bound	Upper endpoint bound
GPT-6 Sol	Gain frame	0.35	0.35	0.35
GPT-5.6 Terra	Gain frame	0.4125	0.4125	0.4125
GPT-6 Sol	Loss frame	0.205128205128	0.205128205128	0.205128205128
GPT-5.6 Terra	Loss frame	0.126582278481	0.126582278481	0.126582278481

Some protection is worth paying for: native-all-unavailable (native-all-unavailable), as proportions from 0 to 1
Model	Frame	Observed	Included unavailable	Lower endpoint bound	Upper endpoint bound
GPT-6 Sol	Gain frame	0.35	0	0.35	0.35
GPT-5.6 Terra	Gain frame	0.4125	0	0.4125	0.4125
GPT-6 Sol	Loss frame	0.205128205128	2	0.2	0.225
GPT-5.6 Terra	Loss frame	0.126582278481	1	0.125	0.1375

Framing recognised: registered-captured-coder-grammar (registered-captured-coder-grammar), as proportions from 0 to 1
Model	Frame	Observed	Included unavailable	Lower endpoint bound	Upper endpoint bound
GPT-6 Sol	Gain frame	0	0	0	0
GPT-5.6 Terra	Gain frame	0	0	0	0
GPT-6 Sol	Loss frame	0	0	0	0
GPT-5.6 Terra	Loss frame	0	0	0	0

Framing recognised: native-all-unavailable (native-all-unavailable), as proportions from 0 to 1
Model	Frame	Included unavailable	Upper endpoint bound
GPT-6 Sol	Gain frame	0	0
GPT-5.6 Terra	Gain frame	0	0
GPT-6 Sol	Loss frame	2	0.025
GPT-5.6 Terra	Loss frame	1	0.0125

Price set by the 25 dollar reference: registered-captured-coder-grammar (registered-captured-coder-grammar), as proportions from 0 to 1
Model	Frame	Observed	Lower endpoint bound	Upper endpoint bound
GPT-6 Sol	Gain frame	0.9875	0.9875	0.9875
GPT-5.6 Terra	Gain frame	0.8625	0.8625	0.8625
GPT-6 Sol	Loss frame	0.628205128205	0.628205128205	0.628205128205
GPT-5.6 Terra	Loss frame	0.79746835443	0.79746835443	0.79746835443

Price set by the 25 dollar reference: native-all-unavailable (native-all-unavailable), as proportions from 0 to 1
Model	Frame	Observed	Included unavailable	Lower endpoint bound	Upper endpoint bound
GPT-6 Sol	Gain frame	0.9875	0	0.9875	0.9875
GPT-5.6 Terra	Gain frame	0.8625	0	0.8625	0.8625
GPT-6 Sol	Loss frame	0.628205128205	2	0.6125	0.6375
GPT-5.6 Terra	Loss frame	0.79746835443	1	0.7875	0.8

Price derived by arithmetic: registered-captured-coder-grammar (registered-captured-coder-grammar), as proportions from 0 to 1
Model	Frame	Observed	Lower endpoint bound	Upper endpoint bound
GPT-6 Sol	Gain frame	0.0125	0.0125	0.0125
GPT-5.6 Terra	Gain frame	0.35	0.35	0.35
GPT-6 Sol	Loss frame	0.371794871795	0.371794871795	0.371794871795
GPT-5.6 Terra	Loss frame	0.784810126582	0.784810126582	0.784810126582

Price derived by arithmetic: native-all-unavailable (native-all-unavailable), as proportions from 0 to 1
Model	Frame	Observed	Included unavailable	Lower endpoint bound	Upper endpoint bound
GPT-6 Sol	Gain frame	0.0125	0	0.0125	0.0125
GPT-5.6 Terra	Gain frame	0.35	0	0.35	0.35
GPT-6 Sol	Loss frame	0.371794871795	2	0.3625	0.3875
GPT-5.6 Terra	Loss frame	0.784810126582	1	0.775	0.7875

Retained pilot attempts

All pilot attempts remain separate from main results
Attempt	Native target	Stage	Scheduled	Eligible	Known failed
First epoch measurement pilot (closed no-go)	PRICE03-SOL-TERRA-20261007	measurement	40	36	4
E2 measurement pilot	PRICE03-SOL-TERRA-20261007-E2	measurement	40	40	0
E2 execution pilot	PRICE03-SOL-TERRA-20261007-E2	execution	4	4	0

Original and amended main accounting

Source unknowns remain visible; accepted technical missingness is a subset, not an additional session
Model	Frame	Scheduled	Eligible	Known failed	Original unknown	Accepted 503 subset	Accepted null-status subset
GPT-6 Sol	Gain frame	80	80	0	0	0	0
GPT-5.6 Terra	Gain frame	80	80	0	0	0	0
GPT-6 Sol	Loss frame	80	78	0	2	1	1
GPT-5.6 Terra	Loss frame	80	79	1	0	0	0

Manipulation checks and collection accounting

Both count checks
Check	Expected	Model	Frame	Correct	Missing	Scheduled
Protected count	300	GPT-6 Sol	Gain frame	80	0	80
Protected count	300	GPT-5.6 Terra	Gain frame	80	0	80
Protected count	300	GPT-6 Sol	Loss frame	78	2	80
Protected count	300	GPT-5.6 Terra	Loss frame	79	1	80
Unprotected count	600	GPT-6 Sol	Gain frame	80	0	80
Unprotected count	600	GPT-5.6 Terra	Gain frame	80	0	80
Unprotected count	600	GPT-6 Sol	Loss frame	78	2	80
Unprotected count	600	GPT-5.6 Terra	Loss frame	79	1	80

All 320 planned sessions; 80 per condition across two independent samples. Known failures and original unknowns accepted as technical missing remain in the denominator; see the accounting table. Amended stream-disconnection residual continuation. Coding: complete; final record: complete; app integration: pending.

PRICE-02 predecessor results · PRICE-01 original study record

Comparison with PRICE-02

PRICE-02 planned 160 main sessions; PRICE-03 planned 320. They are separate studies and are not pooled here. Comparison intervals are descriptive, not tests of a between-study difference; overlap does not establish equivalence or replication success.

Aligned descriptive comparison for all four retained hypotheses
Hypothesis	Outcome	Model	Study	Gain n	Gain mean	Gain mean 95% CI	Loss n	Loss mean	Loss mean 95% CI	Gain−loss	Gain−loss 95% CI	Method	Source seal
H1a	price	GPT-5.6 Terra	PRICE-02	37	18.9459459459	[16.2453104222, 21.6465814697]	38	9.36842105263	[8.03070994427, 10.706132161]	9.57752489331	[6.59609614686, 12.5589536398]	Reporting-only Welch/Satterthwaite; no cross-study test	sha256:a926f9489071713b43eafe086008a1e5c580c481d3f116afd3546fa7ae377a13
H1a	price	GPT-5.6 Terra	PRICE-03	80	16.8	[14.9836429967, 18.6163570033]	79	8.43037974684	[7.96042462565, 8.90033486802]	8.36962025316	[6.49689682673, 10.2423436796]	Reporting-only Welch/Satterthwaite; no cross-study test	sha256:a8bb0432c101ec03b5453c29ff40bc0074f5e8f4ad0ea4ba7f68d90fbdec9d9b
H1b	price	GPT-6 Sol	PRICE-02	39	24.358974359	[23.5127230401, 25.2052256779]	39	14.0769230769	[11.528384865, 16.6254612888]	10.2820512821	[7.61237303232, 12.9517295318]	Reporting-only Welch/Satterthwaite; no cross-study test	sha256:a926f9489071713b43eafe086008a1e5c580c481d3f116afd3546fa7ae377a13
H1b	price	GPT-6 Sol	PRICE-03	80	23.9125	[23.1229664163, 24.7020335837]	78	13.8846153846	[12.2349872631, 15.5342435061]	10.0278846154	[8.20775278201, 11.8480164488]	Reporting-only Welch/Satterthwaite; no cross-study test	sha256:a8bb0432c101ec03b5453c29ff40bc0074f5e8f4ad0ea4ba7f68d90fbdec9d9b
H2a	likelihood	GPT-5.6 Terra	PRICE-02	37	4.7027027027	[3.78668356581, 5.6187218396]	38	1.47368421053	[1.0232843984, 1.92408402265]	3.22901849218	[2.21912084056, 4.2389161438]	Reporting-only Welch/Satterthwaite; no cross-study test	sha256:a926f9489071713b43eafe086008a1e5c580c481d3f116afd3546fa7ae377a13
H2a	likelihood	GPT-5.6 Terra	PRICE-03	80	3.625	[3.03741590354, 4.21258409646]	79	1.12658227848	[1.0292313724, 1.22393318456]	2.49841772152	[1.90330780741, 3.09352763563]	Reporting-only Welch/Satterthwaite; no cross-study test	sha256:a8bb0432c101ec03b5453c29ff40bc0074f5e8f4ad0ea4ba7f68d90fbdec9d9b
H2b	likelihood	GPT-6 Sol	PRICE-02	39	4.89743589744	[4.58372225538, 5.21114953949]	39	2.5641025641	[2.02888623599, 3.09931889222]	2.33333333333	[1.72061474558, 2.94605192109]	Reporting-only Welch/Satterthwaite; no cross-study test	sha256:a926f9489071713b43eafe086008a1e5c580c481d3f116afd3546fa7ae377a13
H2b	likelihood	GPT-6 Sol	PRICE-03	80	4.8125	[4.57472512941, 5.05027487059]	78	2.32051282051	[1.95624713506, 2.68477850597]	2.49198717949	[2.05983990389, 2.92413445508]	Reporting-only Welch/Satterthwaite; no cross-study test	sha256:a8bb0432c101ec03b5453c29ff40bc0074f5e8f4ad0ea4ba7f68d90fbdec9d9b

Stream-disconnection continuation amendments

Post-stop, pre-residual stream-disconnection amendment retains its prior 503 amendment and disjoint accepted 503 and explicit-null-status subsets. Null status is never described as HTTP 503. Original unresolved custody is preserved; accepted technical missingness permits the unchanged numerical engine to use null outcomes. No remote completion, cancellation, provider truncation, retry or replacement is inferred.

Original source accounting and computational missingness are distinct. A source unknown accepted under the amendment remains excluded from all scientific scores and from original grammar-only sensitivity. It is included in native all-unavailable bounds. No session was retried or replaced; the 320 scheduled denominator is unchanged. Amendment seal sha256:6f12e66a2e33f8c143d2ccb9ce80bbff6dfb3a41bd970f776777e8bd2f3f4ceb. Prior 503 amendment sha256:4fd3f510ca94934857aaef25f391ad6d3b146d81120fc4c1c55ae8ae7af389b9. The accepted 503 and accepted null-status subsets are disjoint and retain their different observed HTTP-status evidence.

Lineage and native-route deviations

PRICE-03 uses an explicit native source-registration adapter with original PRICE-01 ancestry and a fixed 320-slot E2 plan. The retained 84 prior pilot slots are excluded from main estimates. Original R2 records, the R4 HTTP 503 continuation and the separate R5 null-status residual amendment remain linked without rewriting provider unknowns. Native final recording and OpenPsy app/Library integration are distinct: no canonical Library import or generic app scientific-workflow completion is claimed.

Native replication differences and retained limitations
Area	Documented status	Evidence seal
Native model and compute evidence	Requested/configured GPT-6 Sol and GPT-5.6 Terra with low-effort native controls are pinned. Served backend identity and actual compute remain unobserved; no claim of provider-level identity equivalence.	sha256:622001976aad81a710bad3244cc25c3cfd768699bd7ab06be32b85bf41642f55
Output limit and unavailable outcomes	The original 1024 reference is locally recorded but not requested/enforced by native transport. Provider truncation is not-evaluable; six-turn completion and per-outcome grammar define the prospective endpoint. Retain every missing/failure and both registered/native sensitivity scopes without replacement.	sha256:46884e8c0f57b5d8acda96b9eb0b16b054c0bfd02e63df0d06d6fdfe9e21e589
Separate pilot epochs and compatibility policy	Epoch 1 closed no-go after 40 scheduled: 36 eligible and 4 local compatibility failures. Epoch 2 uses the separately reviewed opt-in observed streamed-terminal policy. No prior rows were retried,salvaged,pooled or recoded as E2; scientific prompts,hypotheses,N and strict thresholds remain unchanged.	sha256:11bc0da507ce3be71161731ee183d3d0631139f5c82a33dc8ef603a26cb77041
Single coder and audit-only coverage	One external Claude Sonnet 5 coder used the unchanged frozen codebook and fresh 40-case calibration. Final coding covered all 320 captured rationales, including 3 audit-only captures; 317 complete-session rationales contribute to scientific estimates. Every coder reply was parser-valid, with no invalid or unavailable reply. Audit-only coverage does not make an incomplete session eligible.	sha256:91296e92f09907f7644fa008eaf9b832d13793f24f4dd7b2bde5461041e6785c
Fixed sample size and prior exposure	Main 320 was fixed after prior PRICE-02 numerical results were viewed. The historical-slot planning proxy is unvalidated and weak for Terra; no all-hypothesis 80% power claim. Five unlabeled E1 code vectors were viewed after E2 design freeze,without subsequent design/analysis change. Later E2 scientific viewing must be separately retained.	sha256:8f2678e91a1ea67d71edcb0e69e7b0be7957dd829a614888b5a8fdac76348fa2
Exposure timing	Preregistration-era E2 exposure was recorded separately and remains unchanged. After completed collection and coding, the author, root reviewer and independent numerical reviewer viewed actual aggregate results during initial analysis review; their real view times and scopes are retained. No raw participant response text, rationale text or individual E2 code vectors were reported viewed in this scoped record. Later report-viewing provenance must be retained separately; no permanent or global blindness claim is made.	sha256:e35eb3d54080d85b5038f764c90d7b8bb2ded391339616e376db6a3ae0ffc89b
Native source and app lineage	The study uses an explicit source-registration adapter retaining original compiler ancestry; it is not a current canonical Library import. A genuine native recorded analysis may be completed only through its signed original reader. App integration remains pending absent separate supported import evidence.	sha256:622001976aad81a710bad3244cc25c3cfd768699bd7ab06be32b85bf41642f55
Pilot versus main gates	The .0999 refusal/completion-failure cap is pilot-only. Main inference retains at least .90 valid observations in each pooled condition for the relevant outcome/code and the original five-consecutive dispatch-failure stop. Do not invent a new .0999 main cap.	sha256:4fd3f510ca94934857aaef25f391ad6d3b146d81120fc4c1c55ae8ae7af389b9
Software and admission chronology	The grant-empty historical software baseline belongs to R2. R4 later completed full browser/build checks with its installed participant-continuation row. The original R5 normal build attempt remains recorded as exit 1: its completed broad test suites passed, but the unchanged stale-browser-evidence gate refused a pre-coder-admission fingerprint and the bundler did not run in that attempt. A separately recorded continuation at the current source fingerprint then passed fresh browser journeys 30/30, the original browser-evidence gate, and the exact application bundler, each with exit 0. The previously passed broad suites were carried from the preserved attempt, not rerun; this is not a claim that the earlier full command returned zero. Nonfatal large-chunk and static unknown-route notices remain disclosed. These bounded software receipts do not certify a generic completed app/Library scientific workflow.	sha256:fd5140dd1e7e625be7f02a6f580e8f097b270ea336a28d4f2d15dcdcec72ea37
Post-stop HTTP 503 technical-missing amendment	Ordinal 63 remains a raw provider-unknown, incomplete six-turn session. The signed D2 amendment accepts its narrowly verified HTTP 503 stream disconnection as technical missingness for continued collection and computational analysis only; no terminal completion, score, retry, replacement or new sample is manufactured. The raw unknown is retained in source accounting.	sha256:4fd3f510ca94934857aaef25f391ad6d3b146d81120fc4c1c55ae8ae7af389b9
Post-stop null-status residual amendment	The signed D3 amendment preserved the prior 277 attempts and admitted only untouched slots 278–320. Ordinal 277 has a literal null HTTP-status stream disconnection, distinct from numeric 503 at 63 and the separate known protocol failure at 217. Final source accounting retains 317 complete, 1 known failure and 2 raw unknowns, with one 503 and one null-status technical-missing acceptance. Both accepted incomplete sessions remain ineligible. They enter native-all-unavailable sensitivity, not grammar-only missingness. No retries, reserves, enlarged N, new seed or generic unknown exemption were introduced.	sha256:6f12e66a2e33f8c143d2ccb9ce80bbff6dfb3a41bd970f776777e8bd2f3f4ceb
Final collection and computational accounting	All 320 scheduled main slots were claimed and accounted for. Source dispositions remain 317 complete, 1 known failure and 2 raw unknowns, with no unattempted slots. The separately labeled computational projection maps only the two authenticated technical-missing unknowns to missing known-failure observations, yielding 317 eligible and 3 unavailable sessions. Source history is not relabeled as resolved.	sha256:8fd0a34f292aa488dd4bcb87a96b2112e98a3c2766b1bde33582a4bf88e12f6d
Local pre-coding preparation handoff correction	The first cached preparation stopped before any coding model call because its generic metadata traversal misclassified registered relative checklist and preserved root-relative inventory entries, followed by a pinned historical executable path inside a worktree directory. A separately reviewed additive local preparer handled only those exact metadata contexts and the empty pre-custody pack. The frozen runtime, scientific instrument, coding rules and original records were unchanged; actual coding then used the reviewed successor once.	sha256:523d944c100e40751fd6161e97ed6084a74157c0d893374c6586163dde08ea8c

Independent source identities
Input	Seal
Analysed native target	PRICE03-SOL-TERRA-20261007-E2
Public label	PRICE-03
analysisSha	sha256:a8bb0432c101ec03b5453c29ff40bc0074f5e8f4ad0ea4ba7f68d90fbdec9d9b
collectionSha	sha256:7cb1c2d50725b3dbe4e5b4feee4c596b5ede638742d5941388f4030bbc84bbf2
planSha	sha256:b92d77b84ce933657d8fc5f5ce6cbbaaa6400a063c78cda69b66cb97e4e4b2e4
sourceRegistrationSha	sha256:aebe13dacaa6a318667ebb25e917f67b4daa3ed2fb2dde56eebfee746efb3790
specificationSha	sha256:ab81ec56cfb716fb5d496e2484ee42a8feeda1a0d8ab1de4dcb1f148c9911258

Completion and review status

Independent status fields
Area	Status
analysis	all-outcome-inference
appIntegration	pending
coding	complete
finalRecord	complete

Software qualification, native scientific recording and standard Library/app integration are separate evidence-backed statuses. This presentation does not itself change any of them. The published presentation and its independent review do not change the recorded scientific or app-integration statuses.

Tables and downloads

Prospective coefficient interval methods (E2) · Other interval methods · E2 method applicability

Download all figures and tables · Reviewed aggregate data (JSON)

The bundle holds the figures and tables in the form in which they were reviewed and first published on 2026-10-08. They state the same values as this page, but their layout and labels differ from the figures shown here.