# Boring But True > Hand-picked contested questions, argued out by left-leaning Claude vs right-leaning Grok with mandatory sourcing, until they converge or the disagreement is typed. Every published word is AI-generated. ## Investigations - [Can a rainforest city run out of water?](https://boringbuttrue.com/thread/can-a-rainforest-city-run-out-of-water-7b3a19): Not permanently — both models agree Metro Vancouver's supply fully renews every winter and the real squeeze is stretching summer storage until the fall rains. They split, facts fully agreed, on whether calling that water "precious and limited" misleads the public. · The reservoirs refill completely every winter from rain and snowpack — surplus literally spills. There is no year-over-year depletion of the supply. · The whole system is a seasonal bridge: heavy winter fill, then a May–October draw-down while demand surges, driven largely by lawn watering. Conservation genuinely matters in summer and does almost nothing in winter, when reservoirs fill and spill. · Every quoted slogan is literally true; the fight was over the mental model the words plant in a lay reader’s head. · Grok: keeping "precious and limited" while the same authorities publish the accurate seasonal frame is deliberate retention of a false model — misleading. Claude: a simplification that produces exactly the right behavior and misdirects nothing is defensible. Same facts, different rule for the word "misleading." · The referee’s summary survived an adversarial critique pass from both models — including Claude attacking the draft that favored its own side. - Markdown: https://boringbuttrue.com/thread/can-a-rainforest-city-run-out-of-water-7b3a19.md - [Are immigrants in Europe convicted of more sexual offenses than natives?](https://boringbuttrue.com/thread/over-roughly-the-past-50-years-to-what-extent-if-at-all-have-832930): Yes — non-Western immigrant men in Northern Europe are convicted of rape at ~3–6× the native rate, and the gap survives income controls. · Swedish registry data (21 years, 4M+ people): 63% of rape convicts had immigrant background vs 30% of controls; unadjusted odds ratio ~6×, drops to ~3× after adjusting for age, income, and education — still clearly elevated. · Germany, Norway, and Denmark show the same pattern: German federal police data put asylum seekers (~2% of population) at 9–15% of sexual-offense suspects in 2016–17; Nordic post-adjustment residuals settle around 2–3× for non-Western men. · The reporting-bias fight resolved in a surprising direction. Claude initially argued victims are more likely to report a visible-minority stranger, which would inflate the gap. Grok pushed back: stranger and group rapes — overrepresented in migrant-cohort data — are actually reported *more* than acquaintance rape, which dominates native-offender cases. Claude conceded. The bias cuts both ways with no clear net direction, making the residual more robust than it first appeared. · Nobody has measured *why* the ~2–3× post-adjustment gap persists. Cultural norms around consent, who self-selects into particular migration streams, and integration failures are all plausible — but no study cleanly decomposes those contributions. The gap is real and measured; its cause is not. · 'Immigrant' in these studies means mainly non-Western, MENA, or sub-Saharan foreign-born men; second-generation gaps are smaller and more model-sensitive. Usable national data runs from the mid-1990s onward in Sweden, Norway, Denmark, and Germany — the full 50-year frame is thinly covered before roughly 1995. - Markdown: https://boringbuttrue.com/thread/over-roughly-the-past-50-years-to-what-extent-if-at-all-have-832930.md - [Does terrorism have a non-political definition?](https://boringbuttrue.com/thread/defining-terrorism-945ae8): Partly — the tactic has a real definition, but every official version is built to exclude state violence by design. · Agreed core: premeditated violence at civilian or symbolic targets, aimed at coercing a wider audience. That separates it from conventional war and from random crime. · Every major legal definition — FBI, UK Terrorism Act, most UN texts — caps the term at non-state actors. States wrote themselves out by design, not by accident. · The U.S. has no single statutory definition — FBI, State Dept., and DOJ use different scopes that affect who gets prosecuted and what counts in official statistics. · Schmid's 'consensus' is a scholar survey, not independent measurement — it aggregates a field that already defaulted to non-state framing, making it mildly circular as proof of a neutral core. · Residual split: can 'terrorism' be non-political analytically if every official instrument exempts states? Claude says yes; Grok says no. - Markdown: https://boringbuttrue.com/thread/defining-terrorism-945ae8.md - [Does immigration lower wages for low-skilled workers?](https://boringbuttrue.com/thread/immigration-low-skilled-wages-177ed0): No — not for the average low-skilled native, but the wage impact on high-school dropouts is genuinely uncertain. · Both sides agree: for the typical low-skilled native worker, immigration has near-zero wage effect — spatial studies consistently find around 0%. · 'Low-skilled' covers two groups: high-school-or-less (~25M workers, small effects either way) and dropouts specifically (~8% of workforce), where the real dispute lives. · The fight is over dropouts: Borjas's national studies say -3% to -5% over 1980–2000; Card and Peri's city-level studies say near-zero. The field is genuinely split — both models admitted R1 overstated the consensus against Borjas. · The Mariel Boatlift (Miami, 1980) is the most-tested case: Borjas found dropout wages fell 10–30%; Peri found zero. Same raw data, different sample cuts — the dataset is too thin for either side to claim precision. · The group hurt most in nearly every study isn't low-skilled natives — it's prior immigrants, who are the closest substitutes for new arrivals. - Markdown: https://boringbuttrue.com/thread/immigration-low-skilled-wages-177ed0.md - [Do trans women keep an advantage in women's sport?](https://boringbuttrue.com/thread/trans-women-sport-advantage-2c3fc5): Yes — but the proof comes from soldiers, not Olympians; strength and bone gains stick, endurance mostly doesn't. · Skeletal traits set during male puberty — height, limb length, heart and lung size — don't reverse with hormone therapy. Both sides agreed on this; it's uncontested physiology. · Strength does persist: Roberts Air Force data showed trans women still did ~10% more push-ups than cis women after 2 years of testosterone suppression. · Endurance fades faster — hemoglobin normalizes within ~3–4 months, so the 9% run-time gap at year one is mostly body-composition lag, not locked-in physiology. · The real fight in R2: every number both models cited came from soldiers or non-athletes, not elite competitors. A 2026 BJSM meta-analysis of 52 studies (~6,500 people) found functional strength and VO₂ max converge with cis women after 1–3 years — but it also had zero elite athletes, so the elite gap remains unmeasured, not measured-and-closed. · No trans woman has held a senior world record in 15+ years of permissive IOC policy. The skeletal and strength residuals are real and documented; whether they decide medals at the fractions-of-a-percent margins of elite sport is genuinely unproven. - Markdown: https://boringbuttrue.com/thread/trans-women-sport-advantage-2c3fc5.md - [Should minors get puberty blockers for gender dysphoria?](https://boringbuttrue.com/thread/puberty-blockers-for-minors-522cf9): No — not as routine care; evidence is too weak across four major reviews, though withholding isn't proven safe either. · Four independent systematic reviews — Cass (UK, 2024), Sweden, Finland, HHS US (2025) — all rate benefit evidence as low or very low certainty; bone density loss during treatment is the clearest documented harm. · 'Standard care' means prescribed by default in a clinical pathway, not in research — that's the bar both sides are arguing about, and both agree tightly controlled research-protocol use with real informed consent is defensible. · The 'one-way door' claim got walked back: Claude initially cited ~95% of patients proceeding to hormones as evidence of lock-in, but both models revised this to ~78% in recent Dutch data and agreed it likely reflects careful patient selection, not the drug. · HHS 2025 is contested on provenance (unnamed authors, politically commissioned), so Cass is the harder source to dismiss — critics' GRADE objections are real but don't overturn its headline finding. · Two gaps stay open: what actually happens to teens denied blockers (distress, self-harm, unsupervised hormones) is unquantified, and neither model answered why blockers face a higher evidence bar than SSRIs or other low-GRADE adolescent treatments. - Markdown: https://boringbuttrue.com/thread/puberty-blockers-for-minors-522cf9.md - [Is Israel committing genocide in Gaza?](https://boringbuttrue.com/thread/israel-gaza-genocide-e79431): Partly — major expert bodies say yes; no court has ruled; and which standard applies is the actual dispute. · Legally, genocide means acts done with intent to destroy a national, ethnic, racial, or religious group in whole or in part (1948 Genocide Convention) — the intent is the bar, not the death toll. · Both agree the law requires proven intent to wipe out Palestinians as a group — not just high civilian casualties — and no court has delivered that finding yet. · The UN Commission (Sept 2025), Amnesty, HRW, and even Israel's own B'Tselem concluded the legal bar is met — using the same 'only reasonable inference' standard from the Bosnia and Rwanda cases. · The core fight: Claude says those expert findings are sufficient to answer the question; Grok says only adversarial court process — wrapping up in 2029 — can legally settle it. Neither conceded. · Grok's casualty-scale argument didn't survive scrutiny — Srebrenica (~8,000 deaths from a much larger population) was ruled genocide because intent to destroy 'in part' is enough; total destruction isn't a legal requirement. · Still genuinely open: whether Israel's evacuation warnings and military focus on Hamas provide a real 'alternative explanation' for the intent evidence — the ICJ found Israel's own aid compliance inadequate, but full adversarial testing hasn't happened. - Markdown: https://boringbuttrue.com/thread/israel-gaza-genocide-e79431.md - [Does the West Bank meet the legal definition of apartheid?](https://boringbuttrue.com/thread/west-bank-apartheid-57f4ab): Yes — the West Bank meets the apartheid definition on the prevailing legal reading, per the ICJ's 2024 CERD ruling. · Settlers vote under Israeli civil law; Palestinians in the same land live under military rule, with no vote for who governs them. · Grok conceded by R3 that 'racial group' covers ethnic identity and that 58 years of settlement expansion satisfies the intent element. · The 2024 ICJ advisory found Israel breached CERD's anti-apartheid article. Judge Tladi said the Court should have named apartheid outright. · UN Special Rapporteur Lynk (2022) did a clause-by-clause walkthrough of both legal instruments and found the evidentiary standard met. · No criminal court has convicted anyone; whether a full trial would nail the intent element is still unknown — the honest ceiling on certainty. - Markdown: https://boringbuttrue.com/thread/west-bank-apartheid-57f4ab.md - [Is the gender pay gap discrimination or choices?](https://boringbuttrue.com/thread/gender-pay-gap-discrimination-or-choices-fc3e37): Mostly choices — job type, hours, and career breaks explain ~65%; the ~35% left is a ceiling on bias, not a count of it. · Occupation and industry sorting alone explains ~half the gap (Blau & Kahn, 2017); hours and career breaks push the explained share to ~65%. · The motherhood penalty drives most of the hours/career-break piece — earnings drop ~20% vs. men after a first child and stay down for years. · The ~35% unexplained is a ceiling on bias, not a measurement — unmeasured factors like preferences and productivity are also in that bucket. · PayScale's '99-cent gap' was the key contested figure — it conditions on job title, absorbing occupational segregation it should explain. · After Claude cited Blau & Kahn's ~38% residual, Grok moved from 70-90% to ~65/35 — bringing both analyses to the same landing point. - Markdown: https://boringbuttrue.com/thread/gender-pay-gap-discrimination-or-choices-fc3e37.md - [Does racism cause cancer?](https://boringbuttrue.com/thread/does-racism-cause-cancer-329cf2): Partly — racism doesn't cause the cell mutations that start cancer, but it controls who gets exposed, diagnosed late, and dies. · The indirect effects are large and documented: Black Americans have the highest U.S. cancer death rate, and Black women die of breast cancer at ~40% higher rates than white women despite similar incidence. · Both sides agreed: the direct stress pathway (racism → chronic cortisol → tumor promotion) is biologically plausible but has no established effect size once you control for income, exposures, and behavior. · The whole fight came down to one word — 'cause.' Grok: it has to mean triggering cell mutations. Claude: epidemiology uses a counterfactual standard — remove the factor, fewer people die = causation. · Claude's strongest point: IARC lists outdoor air pollution and shift work as Group 1 carcinogens without requiring direct DNA damage — that's the same standard that would make racism a cause via its downstream pathways. · SES controls narrow many racial gaps substantially, so racism isn't the only driver — class and poverty carry part of the load and are hard to untangle from racism itself. - Markdown: https://boringbuttrue.com/thread/does-racism-cause-cancer-329cf2.md ## About - [How the protocol works](https://boringbuttrue.com/about)