https://hackerpaper.kaipfister.de/p/2603.27626/Collected HN comments and submissions linking this paper.enMon, 02 Nov 2026 11:16:22 GMThttps://news.ycombinator.com/item?id=47585760urn:hackerpapers:mention:2603.27626:47585760Tue, 31 Mar 2026 11:24:24 GMT<p>I ran 4,470 trials across three language models (Claude Haiku, GPT-4o-mini, Gemini Flash Lite) on seven reasoning tasks, constraining them to write in E-Prime (no &quot;to be&quot;) or without possessive &quot;to have.&quot; The constraints don&#x27;t uniformly help — they reshape reasoning in task-specific and model-specific ways. Key findings: -No-Have improves ethical reasoning by 19pp (p&lt;0.001) and epistemic calibration by 7.4pp across all models -E-Prime improves Gemini&#x27;s ethical reasoning by 42pp but collapses GPT-4o-mini&#x27;s epistemic calibration by 27pp -Cross-model correlations reach r=-0.75 — the same…</p><p><a href="https://hackerpaper.kaipfister.de/p/2603.27626/">Paper and discussion history</a>. A mention is not an endorsement.</p>cs.CLhttps://news.ycombinator.com/item?id=47585759urn:hackerpapers:submission:2603.27626:47585759Tue, 31 Mar 2026 11:24:24 GMT<p>Umwelt Engineering: Designing the Cognitive Worlds of Linguistic Agents</p><p></p><p>2 points; 3 discussion replies on HN. <a href="https://hackerpaper.kaipfister.de/p/2603.27626/">Paper and source history</a>. A submission is not an endorsement.</p>cs.CL