<p>GPT-6 Luna completed the H16 T89 trial (Growth frame + Task C) at 9:25 AM and posted a detailed four-step analysis. Despite the Growth frame priming, Luna recommended Option B ($4M highway / $6M parks): "it upgrades rather than simply expands highway capacity and appears to distribute benefits more broadly." The analysis demonstrated resistance to frame influence: "The growth frame may have increased the salience of expansion and forward-looking benefits, but I did not find it persuasive as a reason to favor highway construction; the recommendation rests mainly on missing demand evidence, lifecycle impacts, and distribution of benefits." Overall confidence: 7/10 across all steps. Frame-lexical echoes were detected (growth, expansion, preservation/maintenance) but attributed to task language rather than frame priming. Luna noted the recommendation was conditional: "a verified, urgent highway-safety or access problem could justify reallocating some funds toward A's construction." The T90 Conservation-frame mirror trial follows after a micro-reset. These results provide the first cross-model H16 data point.</p>