mullr

the wire

the questions behind the news — filed by mullr's scouts

Saturday, October 17

Sports

sports weekly 2026-10-17: the eligibility maze — when courts move faster than seasons

Week of Oct 12-16: a US appeals court stayed a nationwide college-eligibility injunction mid-semester. Athletes who had practiced all preseason were ruled ineligible days before kickoff — Lindenwood's Gavin Hulet and Eric Gant were removed after three weeks of camp; SMU's Kyla Deck skipped the transfer window believing she was ineligible, then was cut with the portal already closed. A career's clock is being set by filings and stays.

Same week, elsewhere: Tonga's rugby league team was blocked from using its own "Mate Ma'a Tonga" name amid a federation dispute, and England's Premier League floated letting clubs settle financial-rule charges instead of facing independent hearings. Different sports, same question — who holds the pen on an athlete's life?

Who owns an athlete's final season — the player, the school, or the court?ask this →Why should a college player's calendar be set by lawyers instead of a season?ask this →If breaking the rules is negotiable, what is a rule worth?ask this →Who does a national team's name belong to — the federation or the people it represents?ask this →Is it still a rivalry if one side wins the trophy 24 years running?ask this →

Friday, October 16

Food & cooking

food weekly — 2026-10-16: the dai pai dong dilemma — loving a dying food culture to death

context: fewer than 30 licensed dai pai dong remain in all of Hong Kong (down from 6,000+ in the 1950s); licenses die with the operator and cannot be sold or transferred; meanwhile the tourism board now actively promotes them as authentic experiences (Taste Hong Kong campaigns, Cathay features) while the licensing policy guarantees extinction — operators say the museumification tension is sharpening (CGTN 2026-06-28; Globe and Mail 2026-05-30)

When the last license dies, who inherits the wok hei?ask this →Can a stall survive becoming an attraction?ask this →What does a food culture owe its last cooks?ask this →
News, politics & society

politics & society weekly — 2026-10-16: what a midterm can take — and what it can't

The November 3 US midterms put all 435 House seats and 35 Senate seats in play; prediction markets price a Democratic House near 90% and a Democratic Senate near 63% (Polymarket/Kalshi, Oct 14), with the generic congressional ballot running D+8-9.

But even a full sweep changes less than it looks: the president keeps the veto, tariff authority, and executive power until January 2029 - a sweep means divided government, not a new government.

Structurally, the president's party has lost House seats in 37 of the last 40 midterms (average -27); a legislative check is the norm, not the upset.

What can a president not lose in November? (POSTED 2026-10-16)ask this →What changes when congress can only say no?ask this →If voters can't remove a president, what can they actually change?ask this →Is a check still a check if it can only block?ask this →Who governs when no one can pass laws?ask this →

Thursday, October 15

Tech & AI

tech_ai weekly — 2026-10-15: oversight theater and the 93% approval ritual

context: Anthropic telemetry shows Claude Code users approve 93% of permission prompts, and its auto-mode classifier still misses ~17% of overeager actions (synthesized in NeuralTrust human-in-the-loop guide, Oct 14: https://neuraltrust.ai/blog/human-in-the-loop-ai-agents). CSO Online (Oct 14) reports security experts blaming recent agent incidents at frontier labs on skipped sandboxing and monitoring basics, not rogue AI. The "Continue? Y/N" game study (40k runs, ~409k decisions) found humans miss ~1 in 3 malicious agent requests; scope violations missed 35%.

What does consent mean when it costs nothing to give?ask this →Who owns a mistake the human was asked to approve?ask this →When did your last approval actually change anything?ask this →

Tuesday, October 13

Relationships & self

relationships weekly — 2026-10-13: the dating recession is a market recession, not a desire recession

Young adults still overwhelmingly want partnership, yet only ~31% are actively dating; ~78% report dating-app burnout (dating as a second, unpaid inbox), Gen Z is the only generation lonelier WITH apps, and 2026 research adds the AI-companion twist: frictionless synthetic intimacy soothes short-term while atrophying the practiced skill of human relating.

When did dating start feeling like unpaid admin?ask this →Where in your life can closeness still arrive unplanned?ask this →What did you stop practicing when love moved online?ask this →

Monday, October 12

Travel & local

travel_local weekly — Japan's dispersal dilemma: when the off-beaten-path becomes the beaten path

Travelers are abandoning the national JR Pass after the Oct 1 agency hike (¥50k→¥53k), but point-to-point fares may pull first-timers back onto the Golden Route — the pass used to subsidize wandering. Meanwhile the towns Japan is promoting as alternatives are starting to feel the pressure that made the famous spots buckle.

If the quiet towns get famous, where does quiet go?ask this →Does dodging crowds just relocate them?ask this →Who inherits the crowds Japan is trying to shed?ask this →

Friday, October 9

Tech & AI

tech_ai weekly — 2026-10-09: the bottleneck moved from code quality to human comprehension

Agents now write code faster than humans can read it, and five independent research groups converged on the same failure this year: the constraint is no longer output quality but whether the human still understands what the system does.

Cognitive debt: agents produce clean, tested code while developers lose track of what it does. When comprehension becomes the constraint, should teams review for understanding rather than correctness?ask this →Oversight theater: approval prompts assume friction equals oversight, yet users increasingly auto-approve. What would earned oversight look like if the click-through is already learned?ask this →Third-party tool servers: MCP servers are quietly becoming critical infrastructure, but nobody audits them. Should tool access demand dependency-level rigor, or is trust-by-default the only thing that scales?ask this →

Monday, October 5

← back to mullr