Hacker News, Distilled

School Smartphone Bans and Their Effects

• Many argue classroom or school-wide bans are more effective than loose usage “caps.”
• Experiences from Australia and other regions: bans reportedly reduce distraction, improve attention, and increase in‑person socializing, though kids still sneak phones in toilets or secluded areas.
• Others are surprised bans aren’t universal and predict they’ll seem “obvious” in hindsight. Some note that private or wealthier schools tend to ban phones first, and this may correlate with future outcomes.

Evidence vs Anecdote

• Skeptics ask for hard data (e.g., PISA scores) rather than “phones are cancer” rhetoric.
• One study cited shows small average gains from bans, with larger benefits for low‑achieving students and little effect on high achievers. Some consider these effects minor.
• Disagreement persists over whether modest measured gains justify broad restrictions on students’ freedoms.

Safety, Emergencies, and Alternative Devices

• Some US parents justify phones as protection in school shootings; others call this a dishonest or statistically irrational argument.
• Suggestions include basic flip phones or kid smartwatches for emergency contact without full smartphone distraction.
• Separate concern: smartphone use while driving, with examples of camera-based enforcement and speculation about using sensor data.

Government, Parents, and “Soft” Regulation

• One camp insists phone rules should be set by parents and schools, not the state; city-level “parenting” is seen as overreach and moral policing.
• Others argue the state’s role is to protect the majority, so if phones are widely seen as harmful, school bans and guidelines are legitimate.
• In Japan and some Eastern contexts, non-binding ordinances and recommendations carry real weight due to higher trust in local government; gentle, symbolic limits are viewed as a way to reset social norms (e.g., what “normal” daily screen time is).

Cultural and Political Readings of the Toyoake Ordinance

• Some see the 2‑hour cap and curfew as a political stunt by a small city on a non-local issue.
• Others think even a mockable, penalty-free ordinance can spark reflection and voluntary “detox,” especially in a high-trust society.

Prompting, Planning, and Clarifying Questions

• Several commenters echo the value of “measure twice, cut once”: invest tokens in planning, decomposition, and clarifying requirements before coding.
• Forcing the agent to ask clarifying questions upfront (sometimes multiple rounds, even requiring a minimum number of questions) is reported to greatly improve outcomes and reduce wrong turns.
• Others argue a faster workflow is to ask the model to generate an “ideal context” in one shot and then correct it, minimizing back-and-forth.
• Some people stream-of-consciousness “talk it out” to share their mental model; follow-up questions then help expose missing details.

Agents, Sub-Agents, and Context Management

• People describe using dedicated “architect” or planning modes before enabling code generation, which can consume 30–40% of tokens but reduce rework.
• Sub-agents are used for tests, Playwright runs, or reading “memory bank” files and returning only relevant summaries, keeping the main context focused.
• Suggestions include putting instructions in AGENTS.md / CLAUDE.md and using @file syntax rather than long “please read X” prompts.
• One workflow: bundle repos into a single file for whole-codebase understanding; others note this only works for small-to-medium projects.

Pricing, Usage Levels, and ROI

• Strong disagreement over “heavy users should use pay-as-you-go.” Multiple commenters claim Anthropic’s Max plan is dramatically cheaper (≈10% of direct API cost) for high-volume users, especially with rollover to API credits.
• Some report spending $1k+/month on tokens and feeling it has huge productivity ROI; others are skeptical this usage is auditable or personally sustainable.
• There’s debate on whether humans can realistically review the volume of code implied by very high token spend; some reply you only review the “final” successful version, not every intermediate attempt.
• Others prefer cheaper, more conservative tools like GitHub Copilot over agentic systems that can propose large, risky edits.

Testing Behavior and Risks

• Experiences differ on agents disabling tests: some see them give up and skip/disable, others avoid this by only exposing a CLI test runner (e.g., run_tests.sh).
• A recurring warning: agents often generate tests that validate current (possibly buggy) behavior instead of intended behavior, leading to a false sense of stability.
• Using two different models to cross-check each other’s work is reported as effective for catching bugs and bad designs.

Documentation, Readmes, and Bot Instructions

• Human developers ask that bot-specific instructions be kept out of README.md, suggesting BOTS.md or AGENTS.md instead.
• There’s debate about what README.md “means”: some treat it as a project-root signal; others say it’s simply “a thing you should read here,” possibly at many levels of a repo.
• People are converging around AGENTS.md as an emerging convention for LLM-oriented context.

Prompt Style and Over-Engineering

• Mixed experiences on verbosity: some find concise, non-anthropomorphic prompts best; others get better results from long, detailed prompts and sharing their thought process.
• “RFC-style” language (MUST/SHOULD/MAY) is reported to work well for persistent instructions.
• LLMs are seen as prone to over-engineering: adding abstractions, caching, and refactors that look plausible but may be unnecessary or even slower.
• Commenters stress asking tightly scoped improvement questions (e.g., “improve maintainability without major architectural change”) instead of open-ended “how can we improve this?”

Criminal vs. employment wrongdoing

• Many argue leaking employer data (especially from police/dispatch) is a serious breach of trust and clear grounds for termination, even “never hire again” territory.
• Others stress the distinction between civil/employment issues and criminal law: not every policy or NDA violation should become a state-prosecuted crime.

Is this theft, trespass, or computer fraud?

• Some see the act as “stealing their stuff” and think criminal charges (theft or similar) are justified.
• Others counter that copying isn’t classic theft (no deprivation of the original), and that calling this “computer fraud/trespass” is especially strained since the footage was captured by pointing a phone at a monitor.
• One commenter notes the actual plea was to a trespass-type computer statute requiring “malicious intent”; with a no-contest plea, the state never had to prove that intent.

Property, IP, and trade secrets

• Thread debates whether leaked CCTV footage is “property” in a legal sense and if copying it can be theft.
• Some treat any unauthorized data exfiltration as theft of intangible property (privacy, trade secrets, competitive advantage).
• Others insist “theft” historically implies deprivation of use and that digital/IP “theft” is really fraud, infringement, or contract breach.
• A lawyer notes trade-secret law requires real economic value and secrecy; random security footage of a public taxiway likely doesn’t qualify.

Security, race, and public interest

• One line of argument: revealing camera positions at an airport can meaningfully aid attackers; that makes the leak dangerous.
• Others reply the crash was in public view and likely FOIA-/records-eligible, so harm is unclear.
• One commenter raises concern that the leaker’s identity (Black, Muslim name) may have influenced how aggressively security risks were framed and prosecuted.
• Several say this isn’t “whistleblowing” or Pentagon Papers–level public interest, more like morbid rubbernecking.

Overbroad computer crime laws & plea bargaining

• Commenters highlight how vague computer crime laws (state CFAA analogues, ToS-violation statutes) give prosecutors wide latitude; “if you’ve been near a computer, they can charge you.”
• Some worry people plead to borderline or even inapplicable charges out of fear, and that bad advice or overcharging can lock in unjust outcomes.

Chilling effects and censorship

• A broader thread connects this case to “anticipatory obedience” and self-censorship: people and platforms over-redact or suppress material to avoid legal or corporate backlash.
• Example: a towing company now heavily blurs logos and labels in accident videos on YouTube despite no clear IP or privacy issue, out of fear of takedowns.
• Several see this as part of a wider drift toward top-down, employer- and platform-driven control over what the public can see.

Scope of the Change (Japan gacha games & revenue concentration)

• Survey says ~70% of top-selling domestic smartphone games in Japan now use external payment sites.
• Some note the sample (top 30 titles) is small by count but likely covers the vast majority of revenue because mobile game income is heavily concentrated in a few “big winners.”
• Commenters stress most of these are gacha / long-running service games optimizing the “wallet → company” funnel.

30% Platform Cut: Tax, Fee, or Rent-Seeking?

• Strong disagreement over whether 30% is “just a platform fee” or an exploitative tax from gatekeepers.
• Historical context raised: consoles and even physical retail often took ~30–40%; mobile stores largely copied that.
• Others argue the cost basis doesn’t justify it; Epic reportedly found break-even near 9%, and some see 15–18% as more reasonable.
• Defenders say store owners built the market and can set terms; critics counter that bundling app stores with OS + device lock-in makes this de facto price-fixing.

Walled Gardens, Competition, and Regulation

• Some call for DOJ/FTC-style intervention, viewing Apple/Google as controlling “the internet” for most people and installing comprehensive paywalls.
• Others reply that Android allows sideloading and third‑party stores, and the web remains open; consumers choose locked-down platforms for security.
• Debate over whether Android’s scary warnings and deep settings make alternative stores a real option or just theoretical.
• Comparison to consoles: walled gardens are acceptable where platforms are optional (games), but phones are now basic infrastructure.

Gacha, Microtransactions, and User Harm

• Many see gacha games as digital casinos or “begware,” manipulating addicts, including children, via psychological tricks.
• Others note that casinos at least dangle monetary rewards; gacha yields only virtual items and frustration.
• Several argue that cutting out Apple/Google simply shifts more profit to already-predatory designs; players don’t necessarily benefit.
• Some wish for filters or store policies that favor “pay once, no abusive IAP” games, but doubt platforms will sacrifice gacha revenue.

External Payments, Safety, and User Experience

• External billing is seen as mainly viable for big titles with “whales”; it won’t save a shrinking gacha market.
• Concerns voiced about trusting random payment processors versus app‑store refunds and protections.
• Others highlight that app stores themselves miss policy violations and host shady apps, undermining their safety narrative.

Confusion Over the De Minimis Change

• Commenters highlight two distinct shifts: ending low‑value duty-free imports and the US refusing “duty-unpaid” postal shipments at all.
• Many note that foreign posts and USPS appear unprepared, with inconsistent guidance and parcels being paused or returned.
• Several small exporters say even brokers and shipping partners “don’t know what is going on,” leading to delays and ad‑hoc workarounds.

Winners, Losers, and Privatization Concerns

• Some suspect this benefits UPS/FedEx and large importers who already have customs infrastructure.
• Others push back, saying the move is broadly destructive to commerce, not a simple “payola” to private carriers.
• There is strong criticism of emerging private duty-collection platforms (e.g., Zonos) and concern governments will become dependent on them, with fee “rackets” on top of tariffs.

Impact on Small Businesses and Niche Markets

• Non-US small sellers (especially in Canada, UK, Japan, EU) report:
  • Stopping US sales, sharply raising shipping, or switching to Delivery Duty Paid (seller pays tariffs plus broker fees).
  • Serious harm to low-ticket, high-volume businesses, handmade goods, used/retro hardware, miniature wargaming, books, and niche apparel.
• Sellers expect customer backlash over “hidden fees” even when they are clearly disclosed.

Consumer Behavior and Price Effects

• Many users regularly buy abroad: electronics from Asia, clothing and specialty goods from Europe/Canada, hobby items, and cheap general merchandise from Temu/AliExpress.
• Multiple comments note that even with tariffs, Chinese-made goods can remain far cheaper than US-made equivalents.
• Some ask whether these price changes will properly show up in inflation metrics; others say that’s unclear.

Logistics, USPS, and Collection Mechanics

• Unusual aspect: the US refusing duty-unpaid postal parcels rather than letting customs bill recipients, unlike almost all other countries.
• Debate over who should collect tariffs: USPS vs foreign posts vs brokers vs marketplaces.
• Large retailers (e.g., Walmart-scale importers) are seen as mostly insulated, while small parcel shippers (Temu, Etsy-scale, individual exporters) are hit hardest.

Political, Legal, and Macro-Economic Framing

• Many comments frame this as part of a broader, reckless tariff strategy tied to current US leadership, with predictions of recession, an “October crash,” and a bad holiday season.
• Some argue tariffs are really a regressive tax on lower-income consumers, not a serious reshoring policy.
• Others note a pending court challenge over whether the executive has authority to impose these tariffs under emergency powers, with potential future refunds if overturned.

Scams and Carrier Practices

• Users expect scams to rise as legitimate messages demanding extra import fees blend with phishing texts.
• Several share experiences of aggressive and sometimes years-late brokerage fee collection attempts by private carriers, occasionally ignored without consequences.

Bluesky’s Decentralization in Practice vs Theory

• Several commenters argue the Mississippi block “proves” Bluesky isn’t truly decentralized, since a single company can flip a switch and block a whole state.
• Others counter that the AT protocol is decentralized while Bluesky-the-company is just one large provider; users can already move to independent PDS servers and third‑party appviews.
• Skeptics respond that, in practice, most users use bsky.app and Bluesky’s infrastructure, so power is still centralized and vulnerable to pressure, similar to Twitter.

How the Mississippi Block Works and How Easy It Is to Bypass

• The “shutdown” affects mainly the official web frontend and mobile app, which on startup call a geolocation endpoint and show a block screen if a flag is set.
• This mechanism is client‑side and trivially bypassed with adblock, VPNs, or third‑party clients and alternative frontends/relays, which reportedly continue to work in Mississippi.
• Debate arises over whether this is “token/performative” compliance or a legally acceptable “best effort” under geo‑IP–based blocking.

Age Verification as a Pretext for Speech Control

• Many see the Mississippi law, and similar ones, as primarily about controlling speech platforms rather than protecting children, echoing concerns about the UK Online Safety Act.
• Commenters note that the rhetoric is “protect the children,” but the targeted services are those that influence public discourse.
• Some worry these schemes invariably create pressure to ban VPNs or heavily regulate them.

Comparisons: Mastodon, Nostr, and Other Networks

• Mastodon is cited as harder for a state to suppress because of its many independent instances in multiple jurisdictions.
• Nostr is praised as more resistant to state‑level attacks due to its relay model, though it has discovery and culture issues.
• Some feel Bluesky combines “the worst of both worlds”: central control over moderation plus partial decentralization complexity.

Effectiveness of Censorship and Social Dynamics

• Discussion extends to China and Russia’s firewalls: governments can’t reach 100% control but can effectively block most people.
• One thread argues that if only 1% can access censored ideas, they’ll be treated as cranks regardless of truth, drawing analogies to flat‑earth and religious belief to illustrate majority dynamics.

Deal Structure & Mechanics

• Government now holds ~10% of Intel common stock, bought at ~$20.47/share using:
  • ~$5.7B in CHIPS Act grants that had been awarded but not yet disbursed.
  • ~$3.2B from a “Secure Enclave” defense program.
• This replaces earlier profit‑sharing / clawback terms attached to those grants.
• Debate whether shares were newly issued (dilution) vs taken from authorized but unissued pool; either way, economic effect is ~10% dilution for existing shareholders.
• Intel confirms no board seat or formal governance rights; government gets standard shareholder voting plus warrants for an additional 5% if Intel ever loses majority control of its foundry business.

Is This Investment, Bailout, or Shakedown?

• Supporters frame it as:
  • Belatedly sensible: equity in exchange for public money instead of “free” grants.
  • A step toward a US-style sovereign wealth/development fund (Temasek, Mubadala analog).
  • Aligning taxpayer upside with industrial policy.
• Critics see:
  • Retroactive change of terms on already‑awarded grants—described as blackmail or extortion (“give us 10% or wait years in court”).
  • A precedent that government can arbitrarily rewrite deals and seize equity.
  • A de facto partial nationalization done by a single administration, not Congress.

National Security vs. Market Distortion

• Pro‑deal arguments:
  • Intel is one of the only US-based advanced logic manufacturers; x86 and onshore fabs are viewed as strategic (defense, supply resilience if Taiwan/TSMC disrupted).
  • Older nodes are still adequate for most military systems; having any domestic capacity matters more than absolute process leadership.
  • A government stake may stabilize Intel’s funding and justify further preferential policies.
• Skeptical views:
  • This props up a poorly run incumbent instead of fostering new competitors or distributed, older‑node fabs.
  • Creates moral hazard and a conflict of interest: state now has reason to tilt regulation, contracts, and export policy toward Intel and against rivals (AMD, GlobalFoundries, TI, Micron).
  • Some argue a DARPA‑style multi‑vendor fab program or forcing Apple/Nvidia‑type firms to co‑invest would be healthier than a single‑firm bet.

Rule of Law & Political Hypocrisy

• Concern that the executive is effectively overriding Congress’s intent for CHIPS (pure grants) and using appropriated funds as leverage for equity.
• Debate over whether CHIPS language (“financial assistance”) legally permits this interpretation.
• Commenters note contrast with past Republican rhetoric against “socialism” and 2008 nationalizations; many label this state capitalism or the “nationalization phase of fascism.”
• Others, including some on the left, welcome the principle of taking equity where the state puts in large strategic subsidies, but still distrust the current administration’s motives and methods.

Pandemic-Era ACA Subsidies Expiring

• Many note the core issue is a COVID-era enhancement of ACA tax credits ending, not a raw 5x price jump across the board.
• Commenters say the big increases mainly hit those earning above 4× the federal poverty level; lower-income subsidies stay intact.
• Several call the NPR example an “outlier” and “rage bait,” pointing out that final 2025 premiums aren’t published yet and that geography (e.g., West Virginia ZIPs) massively skews prices.
• Others push back that even a one-time or partial increase is catastrophic for many households, especially retirees or near-retirees.

Universal vs. US-Style Healthcare

• One camp argues universal or single-payer care would cut US costs by eliminating insurance middlemen, advertising, denial-management staff, and complexity; they cite estimates of ~15% of US spend going to insurance-related administration.
• Another camp replies that universal coverage alone doesn’t fix the underlying cost structure: cross-subsidies (e.g., private plans subsidizing Medicare), entrenched interests, and decades of half-reforms created a system that’s extremely hard to unwind.
• Some stress that simply “spending more” is not the answer; the US already spends roughly twice peer countries’ share of GDP with worse outcomes.

Quality, Access, and Wait Times

• Defenders of the US system say that with “good insurance” in a major metro, care can be “amazing” and fast, with top hospitals and drugs covered.
• Others counter with long waits for specialists even on high-tier plans and note that this good experience applies to a minority; many are uninsured or underinsured.
• Experiences from Canada and Quebec highlight serious wait times and backlogs; defenders of socialized systems reply that this is triage, not denial, and contrast it with Americans simply forgoing care or going into debt.
• Several say the US now manages to combine the worst of all rationing methods: high prices, access problems, and quality gaps.

Cost Drivers in US Healthcare

• Proposed villains include: insurance bureaucracy, private equity–owned hospital chains, drug pricing, and regulatory capture.
• Some blame high physician salaries; others point to research showing administration and overhead dwarf clinician pay.
• Commenters discuss how profit caps on insurers push profits into vertically integrated sister companies (pharmacies, clinics, PBMs).

Work, Freedom, and Health Insurance

• Multiple stories describe “job lock”: people staying in unwanted jobs, abusive relationships, or abandoning entrepreneurship and education to keep coverage.
• Some argue this is an economic and equality issue: tying healthcare to employment gives employers disproportionate power.
• Others warn that any universal scheme risks huge tax burdens and must actively control prices, not just shift costs from premiums to taxes.

Politics and ACA Sabotage Narratives

• Several believe Republicans are intentionally letting ACA supports expire (e.g., individual mandate removal, subsidy sunset) to trigger a “death spiral” while blaming Obama/Biden.
• Others note the official justification is simply that pandemic “emergency” subsidies were temporary.
• There’s debate over whether Democrats will effectively campaign on these hikes; some are pessimistic about their electoral strategy and broader media framing.

International Comparisons and Exit Talk

• Commenters compare US care with systems in Australia, Canada, Europe, Mexico, and Germany/France, debating tax rates, wait times, and private top-up insurance.
• Views range from “US with good insurance is the best in the world” to “US is by far the worst system I’ve experienced.”
• A minority seriously discuss emigration (e.g., Mexico, Europe) as a rational response to US healthcare costs.

Mississippi law vs UK rules

• Commenters highlight Bluesky’s own distinction: UK Online Safety Act targets specific features/content, with optional age checks and no tracking of minors’ status.
• Mississippi’s law is described as requiring site-wide age verification, persistent tracking of who is a child, and collection of “sensitive” IDs, seen as far more intrusive.

Bluesky’s IP block and community reaction

• Many support blocking Mississippi as the “only correct response” to an overreaching law and a way to pressure courts and lawmakers.
• Others argue this isn’t “fighting” but surrendering part of the market to stay safe legally.
• Some suggest fully exiting US jurisdictions or leaning more on the underlying protocol so alternative apps can circumvent state rules.

Child safety, porn, and social media harms

• Strong split between “parental responsibility and filters are enough” vs “that’s technically hard and many parents can’t or won’t do it.”
• Some see the real problem as addictive algorithmic feeds and social media’s correlation with teen mental health issues, more than porn itself.
• Others note obvious workarounds (VPNs, shared devices, open Wi‑Fi) will limit effectiveness, making these laws mostly “security theater.”

Motives and competence of lawmakers

• Widespread belief that legislators don’t understand internet technology or second‑order effects.
• Some speculate that think tanks and culture‑war agendas (anti‑LGBTQ, anti‑porn, broader surveillance) drive these laws under “protecting children.”
• Hypocrisy is noted: same factions talk about parental autonomy but seek broad content control.

Impacts on speech, small platforms, and power

• Many see age‑verification mandates as de facto speech regulation that burdens small and emerging platforms most, while big players and politically allied sites may be shielded.
• There’s concern this delegitimizes the state and legal system when enforcement is visibly selective.

Age verification schemes and infrastructure

• Ideas raised: centralized or app-based age-verification, government-backed digital IDs with zero-knowledge proofs, browser/HTTP headers marking “adult” content.
• Skeptics warn these create markets for “age-verified” accounts, push toward device-identity binding, and are a dream for surveillance and ad targeting.

Circumvention, enforcement, and broader trend

• VPNs, Tor, and private tunnels are seen as the only practical workarounds; some ask for low-friction, privacy-focused options.
• Questions arise about how states can fine out-of-state or foreign platforms; answer: interstate commerce and pressure on local ISPs.
• Other states (Wyoming, South Dakota) are cited as even more extreme, and there’s a sense this era is pushing toward wider internet fragmentation and control under the banner of child safety.

Perception of the article & naming

• Some initially suspected the post was AI-generated or “blogspam” due to Substack’s look and heavy bullet/bold formatting; author clarified only the thumbnail was AI, content was hand-written.
• Several jokes and side-comments about the Kafka name (writing-optimized system, “Kafkaesque” configuration, dark literary references).

When Kafka fits vs when it doesn’t

• Strong theme: Kafka is often misused as “just a queue” or generic pub/sub because of hype, resume-building, or sales pressure; this can cost a lot in money and complexity.
• Multiple commenters: if you only need a simple queue, better options include RabbitMQ, Redis, SQS, MQTT, or even named pipes/ZMQ; Kafka becomes needless overhead for small/internal apps.
• Others argue Kafka is appropriate once you have many producers, large event volumes, and multiple independent consumers that need durable, ordered, replayable streams.

Kafka’s core value: distributed log & replay

• Defenders stress that Kafka is not a traditional queue but a distributed log:
  • Very high write throughput.
  • Retention and replay for minutes to days.
  • Decoupled producers/consumers; new consumers can re-read from any offset.
  • At-least-once delivery with ordering within partitions.
• Replayability is cited as a key original motivation at LinkedIn: enabling new consumers on historical data without rebuilding pipelines.

Alternatives and comparisons

• MQTT: good for lightweight pub/sub and IoT, but weak on persistence and backpressure; often paired with Kafka/RabbitMQ rather than replacing them.
• NATS / Jetstream:
  • Praised for simplicity and flexibility, but nuanced analysis notes tradeoffs: per-stream/consumer overhead, scaling limits, 1MB message cap, more complex topologies and semantics.
• Redis Streams and cloud services (Kinesis, Pub/Sub) suggested as simpler, cheaper Kafka alternatives at non–Fortune-100 scale.
• Redpanda highlighted as Kafka-compatible with less operational/JVM overhead.

Operational & complexity concerns

• Recurrent complaints: Kafka is “Kafkaesque” to run; HA clusters are hard, especially on unreliable cloud infra. Managed services (AWS MSK) help but migrations (e.g., to KRaft) are non-trivial.
• Stream-processing systems and replay tooling (offset management, separate replay consumers, DLQs) are described as much harder than batch/S3+SQL approaches, especially for small teams.

LinkedIn scale and evolution

• Discussion of LinkedIn’s reported Kafka scale (tens of trillions of records/day, ~17 PB/day) and how that drives very different design choices.
• LinkedIn is now moving from Kafka to an internal system (Northguard) to address metadata scalability and partitioning limitations; seen as a “clean-sheet” redesign for extreme scale.

Java / JVM side-thread

• Long subthread debates whether Kafka’s Java/JVM basis makes it “bloated” or just widely adopted and well-tooled; disagreement centers on memory use, GC tradeoffs, and real-world performance vs theory.

Browser & standards status of XSLT

• Chrome is adding a flag to disable XSLT to measure breakage; some commenters initially misread this as immediate deprecation.
• Google, Mozilla, and Apple all support removing XSLT from the HTML standard; only XSLT 1.0 is referenced there.
• The XSLT spec itself is at 3.0, but browsers only ever implemented 1.0, and even non‑browser tools rarely go beyond 1.0.
• Some see this as normal standards evolution (no implementers → remove), others as large vendors unilaterally killing features.

Security, maintenance & rationale

• Pro‑removal arguments: old code (libxslt/libxml2) full of security issues, no embargoes, and an attack surface overlapping with XML parser bugs (entities, “logic bombs”, XXE).
• Counterpoint: many of these problems are solvable with better defaults and limits; XML’s bad reputation is partly historical.

Developer reactions & real‑world usage

• Some developers rely on client‑side XSLT for personal sites and niche domains (e.g., technical publishing, TEI, documentation pipelines) and feel abandoned.
• Others argue that many vocal defenders had never used XSLT before, and that widely cited examples of “real usage” are overblown or already have server‑rendered HTML equivalents.

Client vs server transforms & performance

• Concern: client‑side XSLT can cause request waterfalls (XML → includes → XSLT → assets), echoing early SPA “spinner” patterns.
• Mitigations mentioned: caching static includes, embedding styles/templates in the XSLT, or running the transform in CI/CD or server‑side (PHP, Java, CMSs).

Alternatives & related technologies

• CSS-only styling of XML can handle simple display but not templating, includes, or logic.
• HTML Imports were abandoned as redundant and poorly integrated with JS modules; HTML Modules are suggested as a future direction.
• Some wish browsers had safe external entities or a simple declarative “HTML includes” mechanism.
• Others propose JSON + template engines (a hypothetical “JSLT”) as a modern parallel.

Ergonomics, complexity & tooling

• XSLT is praised for declarative power (XPath, templates) and rich XML workflows, but criticized as verbose, unintuitive, and inconsistent across processors.
• Several note that general-purpose languages (Ruby, PHP, JS) are often easier for transforms.
• A few report mixed results using LLMs to write XSLT: good for syntax scaffolding, poor at nontrivial logic.

MoQ, WARP, and Streaming Formats

• WARP is a media format candidate for MoQ, built around CMAF/fMP4 segments, so it can share caches and remain backward-compatible with HLS/DASH during rollout.
• MoQT deliberately separates transport (fan-out, priorities, groups/tracks) from media details, allowing different streaming formats and use cases to coexist on the same relay/CDN infrastructure.

MoQ vs WebRTC and Transport Behavior

• QUIC/WebTransport can drop media on streams or datagrams, giving WebRTC-like congestion behavior.
• Current issue: common QUIC congestion controllers favor throughput over latency, unlike WebRTC’s GCC; server-side tuning can help but browser changes are also needed.
• Latency is largely under receiver control: players must build their own jitter buffers and choose when to render frames.

QUIC, NAT, and Infrastructure

• For client–server, QUIC has no special NAT traversal problem: 1 RTT to connect; main issue is environments that block UDP.
• P2P over QUIC in browsers is considered impractical today; WebRTC+TURN/ICE remains the path there.

CDN / Relay Architecture and Load Balancing

• Public MoQ relay uses GeoDNS today; anycast plus QUIC preferred-address is seen as the better long-term option but cloud anycast+UDP is limited.
• Others warn that anycast can be fragile at low QPS, while DNS steering has caching trade-offs but more control.

Performance and Goodput

• Some discuss research showing lower goodput/bitrate with HTTP/3 vs HTTP/2; consensus is that QUIC implementations aren’t yet as tuned as TCP, especially for very high bitrates and low loss.
• Hardware offload and better GSO/GRO-style optimizations are seen as key future improvements.

Browser Support and Bugs

• WebTransport works in Chrome and Firefox desktop; Safari has partial support behind a flag and is still buggy.
• Firefox has known WebTransport and HTTP/3 quirks; people share bug links and debugging tips (DoH, HTTPS DNS records, Alt-Svc behavior).

Demos, UX, and Rendering Issues

• Many report the MoQ demos as “instant” and extremely low-latency, praising “click-to-streaming time.”
• Several mobile users see fast-moving horizontal black lines when using canvas-based rendering; others see aspect-ratio issues. Suspected causes include browser rendering bugs and canvas vs <video> behavior, with some discussion of power usage and autoplay blocking.

Other Use Cases and Features

• MoQ could support low-latency multiplayer/game networking by mapping game data onto MoQT objects/tracks and using relays for fan-out.
• Some ask about multicast; responses argue CDNs already approximate multicast at L7, and true IP multicast likely only helps the very largest events or P2P.
• Multiple publisher bitrates/simulcast and receiver-driven track switching are already possible; subgroup-based and sender-side adaptation need more experimentation.

Ecosystem, Tooling, and Meta

• Third-party implementations (e.g., MediaMTX integration with WebTransport + native QUIC) show early ecosystem growth.
• Questions arise about adoption in OBS/YouTube and relationship to existing APIs like WebRTC, WebCodecs, and MSE.
• One thread branch debates HN’s “original source” rule regarding this blog vs the Cloudflare announcement; moderators ultimately restore the blog post, with recognition that it adds distinct technical content and tooling.

XML/XHTML/XSLT: Promise vs Reality

• Several comments recall an “alternate universe” where XML and XHTML won, yielding strict, beautiful, long-lived tooling.
• Many others say that universe briefly existed: XML “took hold” in the 2000s, then developers abandoned it after painful real-world experience.
• XHTML’s strictness (fatal errors on trivial mistakes) is cited as a major reason it lost to forgiving HTML, though some still view that strictness as a feature.

Why XML and XSLT Lost Ground

• Core complaints: verbose markup, awkward data model vs typical programming objects, confusing decisions about attributes vs elements, and hated namespaces.
• XML specs often became “design-by-committee tire fires” (SOAP, SAML, XML-DSIG, multiple schema systems), which soured people on XML generally.
• XSLT is widely described as powerful but hard to read, debug, and reason about; many tried it for HTML rendering and later regretted the complexity.
• JSON gained favor because it matches language data structures, is simpler to type and experiment with, and works natively with JavaScript.
• Some argue XML was unfairly blamed for bad schema and API design in an immature era; others insist devs just didn’t like using it and moved on.

Disagreements on Causes and Merits

• One camp blames Google and the search/advertising ecosystem for preferring a sloppy, JS-centric web and starving XHTML/semantic web efforts.
• Others strongly reject that, saying XML “died because it sucks” and the semantic web was flawed conceptually.
• There’s debate over streaming: some say XML/JSON both fit poorly; others note proper streaming parsers (SAX, XmlReader) work fine if used correctly.

Current and Potential Uses of XSLT

• Real-world uses mentioned: publishing pipelines (e.g., JATS), RSS display, document-heavy sites, and static site generation.
• Supporters emphasize XSLT’s role in separating content from presentation, its concision via XPath, and its longevity vs churn in JS frameworks.
• Skeptics see browser-side XSLT as niche and mostly obsolete compared with server-side templating, HTML <template>, and JS/WASM.

XSLT 3.0 on the Web Platform

• Pro-XSLT-3.0 arguments: modern features (JSON, maps/arrays, streaming, better iteration) could make in-browser XML toolchains more useful and future-proof.
• Anti arguments: browsers shouldn’t carry complex, rarely used engines; XSLT can live in server-side or WASM-based tools instead.
• Some suggest if XSLT remains, upgrading to 3.0 and deprecating 1.0 is preferable to outright removal.

What the USDA Policy Change Actually Does

• USDA will no longer use its programs to fund solar projects on “prime American farmland” or projects using panels from “foreign adversaries.”
• Support is shifted toward biofuels, biomass, natural-gas-derived hydrogen, and “American energy” on national forest land.
• Several commenters stress this is not a legal ban on renewables, but a ban on support / subsidies via USDA, which may still be critical because farmers mainly interact with USDA, not DOE.

Perceived Climate and Energy Consequences

• Many see the move as deeply shortsighted given drought, crop stress, and the need to decarbonize electricity.
• Commenters argue renewables (especially solar + batteries and offshore wind) are already the cheapest new generation and will “win anyway,” but policy can significantly delay deployment.
• Others note that energy used in production and industry is large and must also be decarbonized; full electrification is argued to reduce total energy demand.

Renewables vs Nuclear and Grid Reliability

• A long sub-thread debates whether high-renewables grids are viable without massive storage.
• One side: renewables are now the cheapest energy in history; batteries are improving; rare “Dunkelflaute” events can be handled by overbuild, flexible demand, or limited fossil/bio/synfuel backup. Nuclear is portrayed as too expensive and slow.
• The other side: examples like Germany and Spain are cited as warning signs—needing huge multi-day storage (estimates run to tens of TWh) and new gas plants; proponents argue that if past subsidies had gone to nuclear, countries could already have fully decarbonized grids at lower prices.

Motivations and Politics

• Many commenters see the policy as driven by fossil-fuel and corn/ethanol interests, culture-war hostility to “woke” solar/wind, and a broader “burn it all down” nihilism.
• Others frame it as “America first” and food-security policy: prioritizing cropland for food, curbing farmland price inflation from subsidized solar, and blocking Chinese panels.
• There is sharp disagreement over whether this genuinely protects national interests or merely entrenches legacy industries and raises emissions.

Impacts on Farmers, Land Use, and Biofuels

• Several point out that farmers often want solar and wind for extra income and reduced costs; hybrid agrivoltaic setups (shade + crops/grazing) may be especially harmed.
• Land-use arguments against solar are widely criticized: wind has tiny ground footprint; solar can coexist with ag; by contrast, 30–40% of U.S. corn/soy is tied to biofuels according to one summary, though another commenter notes USDA data show biofuels use much less than a majority of total cropland.
• Overall sentiment: the policy favors biofuels and fossil-adjacent systems over much more land-efficient PV and wind, with unclear long-term benefits for food security or farmers.

Gmail vs mailbox.org & PGP

• mailbox.org integrates PGP in the web UI (including mobile browsers), unlike Gmail, which requires external tools and doesn’t manage keys natively.
• Some warn that provider-managed PGP weakens the threat model: you’re protected from others but not from the provider itself.
• Others note that switching from Gmail for privacy is still a win even without full end‑to‑end encryption, simply by leaving Google’s data ecosystem.

Owning your domain & portability

• Very strong consensus: use your own domain so you can move providers without changing your public address.
• Benefits: painless migration (just update MX records), diversification away from any single provider, and easy aliasing per service.
• Risks discussed: forgetting to renew a domain (especially “premium” ones), high annual costs, and catastrophic consequences if someone else takes it and starts receiving your email.
• Mitigations: multi‑year renewals, stacked payment methods, calendar reminders, choosing common TLDs (.com/.net/.org) to avoid validation and human‑error issues.

Provider reliability, spam & self‑hosting

• Several report excellent reliability from large independents (Fastmail, Proton, Migadu, Zoho, Purelymail, etc.), but note that Gmail remains a high bar for deliverability and uptime.
• Self‑hosting is seen as attractive but risky: deliverability to big providers, ongoing security, hardware/ISP dependence, and reputation building. Some suggest hybrid models (self‑host inbound, third‑party for outbound).
• Many emphasize that backing up email (IMAP, local Dovecot, cloud storage) is as important as choosing a provider.

Experiences with mailbox.org

• Positives: integrated PGP, IMAP support, EU hosting/GDPR, ability to bring your own client and domain, decent price.
• Negatives/concerns raised:
  • Anti‑spoofing and outbound spam‑scanning behavior (including reports of silently dropped outbound mail).
  • Some spam filtering and 2FA rollout issues; business plans lacking proper 2FA in at least one account’s experience.
  • UI considered mediocre; some sites reportedly reject @mailbox.org addresses.
  • Address recycling after account lapse (90 days on some plans) is seen as a security risk.

Other providers & trade‑offs

• Fastmail receives repeated praise: excellent support, strong UI, good aliasing/catch‑all handling, reliability with large inboxes. Concerns: US hosting, relatively high base plan, and address reuse policy.
• Proton and Tutanota: appreciated for e2e encryption within their ecosystems, but IMAP limitations and reliance on their apps/bridge are seen as significant trade‑offs.
• Migadu, Zoho, NameCrane/CraneMail, Purelymail, disroot.org, mxroute and others are mentioned for specific niches (price, family domains, generous aliasing), usually with some quota or feature caveats.

De‑Googling & broader ecosystem

• Many describe broader “de‑Googling” journeys: moving email, calendars, storage, and photos (e.g., Immich, Nextcloud, Ente), and experimenting with alternative Android ROMs (GrapheneOS, LineageOS, Calyx).
• Others are satisfied stopping at a non‑Google mail provider, arguing that further steps (self‑hosting, custom ROMs) are unnecessary or too fragile for their threat model.

Limits of email privacy & PGP

• Multiple comments stress that email is fundamentally not private: no ubiquitous e2e, metadata leakage, and most correspondents still sit on Google/Microsoft.
• PGP is seen as powerful but niche: few people use it, discovery and onboarding are awkward, and usability vs security remains a core tension.
• Some propose using email only for “boring” communication and moving truly sensitive conversations to modern encrypted messengers instead.

Perceived impact and real-world usage

• One side argues XSLT-in-browser is barely used; cited examples all have prominent HTML equivalents, so removal won’t meaningfully harm users.
• Others counter that certain government/regulatory and personal sites rely on XML+XSLT specifically to make machine-targeted data human-readable at the same URL (e.g., laws, feeds, IoT/admin UIs).
• Some worry especially about “silent” breakage where XML pages suddenly become “robot barf” again.

Metrics, telemetry, and “how much is enough?”

• Pro-removal commenters lean on Chrome usage stats showing very low XML/XSLT page loads compared with text or PDF.
• Critics say these metrics are biased (enterprise users, technical users, telemetry-off users) and too crude (page loads ≠ importance). They propose survey-like methods and weighting by user groups, but are challenged for lacking counter-data.

Security, maintenance cost, and standards simplification

• XSLT engines are old, C-based, and have accumulated serious, long-lived vulnerabilities; keeping them is seen as ongoing security and maintenance “tax” for a tiny audience.
• Others note the same vendors are happy to add complex, risky APIs (WebUSB, WebSerial, etc.), so “security/complexity” feels selectively applied.
• Some argue browsers should have modernized to newer XSLT versions instead of letting implementations rot, then declaring them too painful to maintain.

Alternatives: server-side, polyfills, and extensions

• Suggested workarounds: server-side transforms, JS/WASM polyfills, moving functionality to extensions, or just serving HTML instead of XML.
• Pushback: automatic application of xml-stylesheet on bare XML URLs can’t be replicated purely with a client-side polyfill unless browsers add new hooks.
• A WASM-based polyfill and a PR to allow script tags in XML are mentioned as partial mitigations.

XSLT’s unique value: format reuse and declarative web

• Supporters highlight XML+XSLT as a clean separation of data and presentation, with one canonical XML source serving both machines and humans.
• Use cases cited: RSS/Atom feeds with explanations, admin tools, low-power/IoT devices offloading rendering to the browser.
• There’s broader frustration that browsers lack good, standardized, low-JS ways to build interactive, data-driven pages (XForms/XPath-like ideas).

Governance, process, and trust

• Some see this as routine: deprecations can take a decade; removal from spec != immediate engine removal; precedents include applet and mutation events.
• Others see inconsistency with the “never break the web” ethos and suspect commercial or anti-competitive motives from dominant browser vendors.

Scope of the Decision and Standards Process

• Multiple vendors (Chrome, Firefox, WebKit) are described as wanting to remove XSLT; it is not a solely Google-initiated move, though a Google-linked issue triggered the public reaction.
• Participants note that WHATWG issues are primarily coordination tools for spec editors, not community referendums, which helped fuel misunderstandings.
• Some stress that standards cannot force implementations: “the web is what engines ship and sites use.” If all engines drop a feature, the spec becomes moot.
• Others argue that decisions by a small set of vendors are effectively unilateral for “tech that affects billions,” and that there should be a clearer, formal public process and venue for feedback.

Resources, Power, and Responsibility

• One side: browser teams have limited budgets; maintaining a niche, complex, security-sensitive feature is a poor use of resources. Companies are entitled to allocate their money as they like.
• Other side: “we don’t have resources” is viewed as a political choice, especially for a company with vast overall wealth. With monopoly-like power comes a moral duty to preserve web stability.
• There’s a split between “Play Nice” (persuade vendors respectfully) vs “Fight the Power” (public shaming and resistance are ethically justified when powerful actors shape the web).

Backwards Compatibility vs Cleanup

• Some see removing underused features as essential hygiene, citing security risk and maintenance cost of XML/XSLT, with prior precedents (FTP, unload, mutation events, Flash, cookies).
• Others counter that XSLT is currently implemented and used; removing a baseline browser feature that can entirely break pages is qualitatively different and should have an extremely high bar.
• Several note that timeline matters: people running sites, factories, and legacy enterprise systems need years of warning and viable migration paths.

Technical Merits and Alternatives

• Supporters of XSLT value client-side XML→HTML templating, separation of data/presentation, and zero-build, low-JS workflows; some still use it for hobby sites, reporting, or glue systems.
• Critics call XSLT and XML “impossible to implement correctly,” attack-surface heavy, and rarely seen on today’s web outside niche/legacy cases.
• Proposed mitigations include WebAssembly polyfills (criticized as large and impractical unless bundled), proxies, or replacing XSLT with CSS/JS and new HTML features (e.g., includes/partials).
• There’s broader debate about whether new declarative features (HTML includes, partial updates) should become native, versus “just use a few lines of JS.”

Government subsidies, equity, and state ownership

• Strong disagreement over whether CHIPS Act money should come with ownership stakes.
• One camp: grants are to “buy” a domestic chip industry; equity or bailouts distort markets and risk conflicts of interest, corruption, and regulatory capture.
• Another camp: giving “stupidly large” sums to profitable firms without equity is a taxpayer giveaway; some support government taking shares or stronger conditions.
• Ohio’s constitutional ban on state equity/credit in private firms is cited as a model to avoid financial risk and corruption, though workarounds via “development corporations” exist.

Industrial policy, national security, and self‑sufficiency

• One side argues chips are strategically like agriculture: you must be able to meet minimal domestic needs in a crisis, even if production is uncompetitive without subsidies.
• Critics say this logic can justify protectionism in “everything,” reduces aggregate output, and often creates crony, inefficient industries; comparative advantage and global supply chains matter.
• Debate over how risky it really is to rely on allies like Canada/Mexico, with some claiming war is essentially implausible in 100 years and others calling that hubristic.
• Russia’s ability to produce basic missiles under sanctions is used to argue that even legacy chips suffice for many military needs and emergency ramp‑up is feasible.

TI’s role, process nodes, and what “matters”

• Several comments clarify that TI’s core business is analog and mixed‑signal parts; calculators and MSP430s are a tiny share of revenue, though probably high margin.
• Skepticism that TI will or can compete with TSMC/Samsung at leading nodes; these Sherman fabs target 45–130 nm, which some see as underwhelming but others note are the “boring” chips in everything from cars to appliances.

Sites, water, and environment

• Concern about siting fabs in water‑stressed Texas/Arizona; Sherman’s draw on Lake Texoma is highlighted, with claims the fab nearly doubles city water use.
• Counterpoints: Texas is large, eastern reservoirs like Texoma are often full, and fab water is heavily recycled, though not perfectly and with potential wastewater issues.

Politics of CHIPS/IRA and regional siting

• Discussion on why investment clusters in red states: claims of faster permitting, weaker regulation, cheaper land and labor, and business‑friendly politics versus claims that it’s primarily political allocation of federal funds.
• Some note backlash in parts of the Midwest to foreign (especially Chinese/Korean) investors despite promised jobs, seeing this as driven by xenophobic or conspiratorial politics.

Free market vs managed economy and labels

• Many see CHIPS and forced deals (e.g., mooted Intel–TSMC tie‑ups) as “Soviet‑style” or fascistic state–corporate collusion; others counter that the U.S. has long been a mixed, heavily regulated economy, not pure capitalism.
• Disputes over the correctness of calling this “socialism” vs “authoritarian capitalism” vs “fascism,” with some emphasizing suppression of labor and corporate favoritism over ideology.

Terminology, calculators, and transparency

• Multiple nitpicks about “fabric” vs “factory/fab” in the title.
• Side thread on whether graphing calculators are still mandatory in U.S. schools and the evolution of TI calculators.
• Some frustration that major industrial policy decisions, tariffs, and corporate aid appear to be made with little transparent debate or congressional constraint.

NYC as a Testbed: Hard or Overhyped?

• Many see downtown Manhattan (irregular grid, canyons, multi-level roads) as a uniquely hard environment; others argue LA/SF are comparably complex, just different.
• Distinctive NYC traits repeatedly cited: dense and assertive drivers, constant rule-bending (blocking the box, “NY lefts”), heavy jaywalking, sidewalk overflow, chaotic bikes/scooters.
• Some think claims of NYC exceptionalism are exaggerated; others with experience in both coasts insist NYC/NJ driver psychology is qualitatively harsher and more adversarial.

Weather and Winter Driving

• Snow and ice are widely viewed as the real new challenge vs. prior Waymo cities.
• Debate over whether AVs can truly learn localized winter behavior (migrating lanes, drifts, black ice, whiteouts) versus just deciding not to operate in marginal conditions.
• Thread notes prior Waymo testing in Buffalo/Michigan but questions commercial viability of full-winter robustness.

Pedestrians, Game Theory, and “Assertiveness”

• Concern that if robots are strictly lawful, NYC pedestrians will endlessly step in front of them; cars need to “signal willingness” to move or they’ll be stuck.
• Reports from SF/LA that Waymos are already becoming more humanlike and assertive (timing yellows, inching at crosswalks, taking gaps at 4-way stops) but sometimes freeze or create awkward standoffs.
• Some fear humans will “bully” AVs—cutting them off, refusing to let them merge—unless they become more aggressive.

Safety, Enforcement, and Law

• Many riders say Waymos feel consistently safer and more predictable than average human drivers, though occasionally weird (odd routes, strange stopping points, rare dangerous edge cases).
• Broader concern: correlated software failures vs. uncorrelated human errors.
• Long subthread on decayed traffic enforcement, red-light cameras, and whether automated evidence from AV sensors should be used for mass ticketing; opinions split between safety gains and dystopian surveillance.
• One line of criticism: AV-caused crashes may be treated differently in law, effectively putting owners in a special category.

Transit, Taxis, and Economics

• Mixed views on urban impact: some see Waymo as “modern transit” that complements or substitutes for weak bus/rail; others insist real solution is walkability, bikes, and transit, not more cars.
• Concerns in NYC about medallions, congestion pricing, and whether Waymo will worsen traffic or must buy into capped fleets.
• Multiple riders strongly prefer Waymo over Uber (no tipping drama, no bad drivers, consistent experience), but some feel guilty about displacing human drivers.

Long-Term Urban Effects

• Hopes: calmer traffic, fewer human-caused deaths, eventual restriction of human driving to niche/recreational contexts, rural coverage where taxis are scarce.
• Fears: vandalism, political backlash if early deployments snarl traffic, corporate resistance to street pedestrianization/“low-traffic neighborhoods,” and further entrenchment of car-dependence in a city that could be bike- and transit-first.

Known phenomenon: tree roots and pipes

• Many commenters note that trees invading water and sewer pipes is very old news; the novelty is the study’s quantified comparison: street trees show less water stress than park trees, traced to city water leakage.
• Multiple anecdotes of roots clogging residential sewage and sprinkler lines (cast iron, clay, and PVC), sometimes causing very expensive repairs and even flooding.

Lead isotopes and tracing water sources

• Clarification that the study’s finding relies on different lead isotope “signatures”:
  • Lead in street trees matches local lead water pipes (old, geologically distinct ore).
  • Lead in park soils/trees matches atmospheric lead from gasoline additives, whose ore sources were more centralized and globally traded.
• Several comments elaborate how ore source, uranium/thorium content, and industrial supply chains lead to different isotope ratios that can be used as tracers.

Health and contamination concerns

• One camp argues leaky pressurized water mains mostly leak outward, so intrusion of contaminants is rare except during depressurization events (power outages, shutoffs), which is why boil-water advisories are sometimes issued.
• Others counter that unannounced shutoffs are common, global infrastructure is often intermittently pressurized, and contamination via rivers downstream of sewage discharges can be serious even without pipe intrusion.

Scale of leakage and whether to fix it

• Montreal’s reported 500 million litres/day loss prompts shock and comparisons:
  • Self-reported losses elsewhere: ~20–30% typical, up to ~40–50% in some cities.
• Debate on fixing leaks:
  • Pro-fix: leaks waste treated water, shorten pipe life by attracting roots, and indicate underinvestment and “technical debt.”
  • Skeptical/nuanced: in water-rich systems returning to the same basin, ecological benefits to trees and cooling might outweigh costs; moreover, digging up roads and coordinating utilities is enormously expensive and politically difficult.

Urban hydrology and design choices

• Several commenters argue the real design failure is routing rainwater rapidly into storm drains instead of deliberately watering street trees.
• Others respond that “it’s cheaper not to do it,” while some push back that short-term cost focus and fragmented incentives, not absolute cost, drive these decisions.