Hacker NewsTuesday, October 13

Many fear sites will gradually require attested passkeys, locking out Linux, GrapheneOS, and open-source managers. Japanese brokerages already restrict authenticators. Critics counter that Apple consumer devices don't support attestation, making mass enforcement unlikely.

Many argue the article undersells phishing/MITM resistance and the fact that server breaches leak only public keys, while TOTP is a phishable shared secret. Others prefer passwords and TOTP for simplicity, flexibility, and use on any device.

Users dislike device or vendor lock-in, forced vendor-app passkeys, and dark-pattern opt-ins. Some say Bitwarden, KeePassXC and 1Password solve backup. Others note export limits, 1Password's userVerification changes, and the new FIDO credential-exchange spec.

Users ask why browsers expose audio, WebGL and core counts without permission, and some blame Chrome's ad-driven owner. Others cite permission fatigue and user confusion, and note that CSS and HTTP headers enable fingerprinting even without JS.

Commenters debate randomizing values versus blending in as a common Chrome/iPhone setup, citing tools like JShelter. Some say fingerprints decay quickly with updates. Others argue that logins, Google's policy shift and CreepJS show tracking persists.

Many stress that EU law requires consent for tracking, not banners for cookies. Banners listing hundreds of "partners" are seen as malicious compliance, though critics note EU sites use them too. Weak enforcement and the ePrivacy/GDPR overlap add confusion.

Commenters doubt an elite class of lab insiders is far ahead, calling the post vagueposting or fundraising hype. Defenders say the gap is about access to massive agent swarms, not skill; debate continues over whether 90% vs 93% reliability compounds meaningfully.

Some say office workers underestimate automation after shallow ChatGPT trials. Others counter that tasks, not jobs, get automated, that losing tacit expertise is costly, and that non-coders see too many confident errors and lack products beyond chatbots to benefit.

Many praise BAR as the best modern RTS for being open source, polished, and able to run 150-player battles. Debated points include harsh 8v8 communities, float-based desyncs across platforms, over-microing AI, and whether the new publisher deal threatens its open-source status.

A developer shares a SupCom engine rewrite with Linux/Mac support and 64,000-unit fights, aiming to fix the original's single-threaded slowdowns. Others recommend FAF, LOUD, and Zero-K, and say Planetary Annihilation failed to capture TA/SupCom's magic.

Many disliked the project's AI-made site, citing samey card layouts, numbered sections, punchy repetitive copy, and a laggy demo. Others argued the design looks fine and that bland templated sites predate AI, as in the Bootstrap era.

Deloitte's claim that 20-year-old records are inaccessible drew cynicism. Others defended retention limits (7–10 years) as standard liability and privacy practice, noting HN usually criticizes data hoarding; critics said money handlers must keep records.

Commenters mocked integrity training as liability cover while alleging padded billing and weak work. Some said firms like Deloitte are hired to rubber-stamp decisions and shift blame, and that all Big Four compete on dishonesty, with Enron and Siemens cited as parallels.

Users described Cyprus IP-licensing tax schemes and noted Deloitte's franchise-like structure isolates local liability. Many doubted consequences, citing Panama Papers, and urged jail time or EU sanctions; one argued the article misrepresents the legal auditor role.

Users fear firms will dodge the rule with annual plans billed monthly plus hefty early-cancel fees. Some defend Adobe since "annual, paid monthly" is labeled. Others call it deceptive and fault the "cancellation fee" framing. An open-source "ArtCraft" alternative was dismissed as vibe-coded.

NYC is seen as eager to legislate but weak on enforcement. Commenters cite unenforced salary-transparency rules and small fines, and suggest bounties for complainants. A lawyer asked where the "as easy as signup" line lies. Others say it's clear: same medium, no hoops.

Praise for Mayor Mamdani drew pushback that NY state, California and other states already have click-to-cancel laws, and the NYC rule reuses state text. Critics call it social-media hype. Supporters cite his other wins. The Biden FTC's scrapped federal rule and an EU Digital Fairness Act were noted.

Pure math is framed as a shared "codebase" of understanding. AI reaches "lighthouse" problems with opaque proofs that need human digestion and may sterilize fields. Others counter that AI introduces new methods, like the sub-O(n log n) result, and will get better at explaining.

Some read the slides as public grieving over lost purpose and prestige, comparing it to past automation and post-scarcity hopes. Others call this uncharitable psychologizing, note that Tao is pro-AI, and say both AI boosters and skeptics sound quasi-religious.

Some argue that having more frontier labs, ideally from different countries, limits power concentration if recursive self-improvement arrives. Others doubt RSI is inevitable, citing verification problems, and note that a few billionaires still hold enormous power either way.

Hands-on tests (image-to-HTML, dev work, trip planning) put Muse Wonder between Haiku and Sonnet, below Opus 5.5, with weak taste and judgment. Commenters note the benchmarks are for Wonder Full and that $4/$20 pricing is Opus-level. Some ask whether models are becoming commodities.

A broken unsubscribe flow and Meta's history prompt deep distrust about handing it code. One user reported seeing strangers' profiles at login but later traced it to the Wipr 2 content blocker. Others criticize Muse ads that reveal usage to Instagram mutuals and unfulfilled open-weights promises.

Many dismissed Google's "high standards" statement as PR spin for breaking a legal obligation. Commenters doubted fines would hurt Google, expected subcontractors to take the blame, and cited PM Orpo's support for data centers. Others called it a procedural misstep that will eventually be approved.

Some argued the cleared land was a low-biodiversity monoculture timber plantation, not true forest. Finns countered that these native-species forests host wildlife, that some rare native forest was also cut, and that illegal is illegal. One Finn called the early logging a calculated bet.

Some linked lenient treatment to Finland's struggling economy: high unemployment, Nokia's fall, paper industry decline and lost Russian trade. Others noted data centers create few lasting jobs and suggested requiring them to use old industrial brownfields instead.

AI workslop takes no effort to produce but heavy effort to debunk. Commenters suggest holding the human committer accountable and pushing back firmly. Others note managers reward speed, AI revisions rewrite entire documents, and hidden technical debt piles up in codebases.

Some say porting a human-designed system suits AI because the thinking is already done. Others note the article's Ruby-to-Rust port still failed, since language semantics force real design choices. It's also unclear which codebase should get new features afterward.

Many insist understanding the code remains essential. One compares this to old demands to inspect compiler assembly, but others reject the analogy because LLMs are nondeterministic. Some describe workplaces that discourage hand-written code and critical pushback against AI output.