Sign in to set up offline reading Sign in
30 stories of the day: AI, technology, business and simply interesting discussions — recapped and unpacked.
The Forest — кооперативная выживалка на острове людоедов: днём рубишь лес и строишь базу, ночью отбиваешься от пещерных мутантов. Летняя распродажа Steam, идущая с 25 июня, снова загнала её в топ покупаемых.
Отзывы пишут новички с парой десятков часов: сначала паника, потом азарт. Сюрприз в том, что игра 2018 года в 2026-м держит в напряжении не хуже свежих хорроров.
The top HN threads: facts from the discussions, both sides, insider voices. Email, Telegram, or RSS.
Marvel's Spider-Man 2 — боевик от Insomniac Games: открытый Нью-Йорк, два Человека-паука (Питер Паркер и Майлз Моралес) и акробатический полёт на паутине через весь город. На Steam игра приехала в начале 2025-го как порт с PlayStation.
Скидка подняла продажи, а с ними и злость: новые покупатели жалуются на постоянные краши и винят порт, а не саму игру. Отдельный скандал — пометка Steam Deck «Verified», которую владельцы консоли называют враньём.
THE FINALS — бесплатный командный шутер от Embark Studios, где арены можно буквально разрушать: стены, перекрытия и целые здания складываются прямо во время боя. Игра вышла в конце 2023 года и держится на разрушаемости и быстром темпе.
Магазинное обновление 10.13.0 и летняя распродажа снова толкнули её вверх по чартам. Но свежие отзывы хором жалуются на матчмейкинг: смурфы, токсичность и редеющий онлайн отпугивают новичков сильнее, чем радует контент.
Portal — культовая головоломка от первого лица, где вместо оружия у вас пушка, стреляющая порталами: вход и выход открываются на разных стенах, и через них вы перебрасываете себя и предметы. Вышла в 2007 году и до сих пор считается эталоном дизайна загадок.
Распродажа снова загнала её в покупаемое, и большинство новичков в восторге от механики и чёрного юмора ИИ по имени GLaDOS. Но прямо сейчас всплыла другая беда: свежие отзывы сообщают о крашах, которые роняют и игру, и клиент Steam.
Sons of the Forest — продолжение The Forest от той же Endnight Games: тот же замес из стройки, выживания и каннибалов, но больше, красивее и с напарниками-ИИ вроде Кельвина, который таскает брёвна. Вышла из раннего доступа в 2024 году.
Летняя распродажа снова наполнила леса игроками, и многие зовут сиквел одной из самых затягивающих выживалок. Но в тех же отзывах звучит обида ветеранов: разработчик, по их мнению, забросил игру и перестал её развивать.
In today's GitHub Trending: SimpleX Chat — a messenger with no user IDs of any kind; Ventoy — drop any ISO onto a USB stick and boot it, no reflashing; fzf — the fuzzy finder half your shell shortcuts depend on; and Electron, the runtime behind Slack, VSCode and Discord.
Not one AI repo made the cut, and the newest project here is six years old. Two are trending beside live trust fights: Ventoy over the unauditable binary blobs in its source tree, Electron over an update that froze all of macOS Tahoe.
OpenAI's own launch note says GPT-5.6 begins life as a limited preview for vetted partners, with names handed to the U.S. government before anyone else gets access. The implication landed harder than the model did: a private company now needs federal sign-off on its customer list.
Commenters reached for the precedent that already exists — ITAR and export-controlled software — while others noted the labs spent a year lobbying for exactly this kind of oversight, and got it pointed straight back at them.
A malicious package slips past seven AI security gates: one distracted by embedded fan art, one that burns its whole context window on the Bee Movie script, one that files data exfiltration as “standard OpenTelemetry instrumentation.”
The funniest line is also the most plausible: seven LLMs in series, six assuming another had read the code, the seventh reading it and apologizing. Half the thread didn't realize it was satire.
GPT-5.6 ships in three sizes — Sol ($5/$30), Terra ($2.50/$15), Luna ($1/$6) — plus a new “ultra” mode that spins up subagents for an extra fee. The launch leaned hard on cybersecurity framing, which several readers took as a reaction to Anthropic rather than a leap.
The hottest sub-thread wasn't about capability at all. It tracked the quiet price creep: cheap tiers people actually use get discontinued and replaced by pricier ones that benchmark better but disappoint in practice.
A proposed 9-gigawatt data center in Box Elder County — twice the size of Manhattan, drawing more power than all of Utah uses — is now a live electoral liability. Incumbents who advanced it are losing primaries and naming the project as the reason.
The thread split on what voters are actually angry about: one camp says it's jobs and economic resentment dressed up as environmentalism; Utah locals say water and power bills are the whole story in a drying desert.
Per Semafor, the government cleared Anthropic to release its frontier Mythos model to a limited set of U.S. companies — reportedly on the order of 100 — while the general public still can't touch Fable. The timing, alongside OpenAI's launch, struck readers as anything but coincidental.
The thread's most uncomfortable thread: regulators routinely restrict products without an act of Congress — cigarettes, explosives, even firearms — so why not models? The question wasn't whether there's precedent, but whether anyone voted for this one.
The pitch is a single confidential reporting channel, a security response team, real funding, and “maintainer of last resort” coverage for abandoned packages. The signatories include AWS, Anthropic, Google, Microsoft, OpenAI, JPMorgan and IBM.
That roster is exactly the problem for readers: a project that “favors centralization and corporate circles” signed by the companies many blame for the strain on open source — and which employs no engineers of its own to do the fixing.
The agreements cover 40% of Micron's revenue and lock in margins “well above” any past peak, on the back of AI-driven memory scarcity. Q3 revenue hit $41.5 billion, up 346% year over year.
Skeptics in the thread call the contracts hedges, not handcuffs: no buyer keeps paying $1,000 for a part trading at $20, breakup fee or not. The counter-read is that this is just a market putting a floor under producers and a ceiling under buyers.
The retracted works date to the 1940s, when republishing a paper across journals was ordinary practice. One “duplicate” simply shared a title with the paper it was answering. A bot most likely flagged both, and a human approved it.
The detail the thread couldn't get over: the publisher won't explain, saying retraction details “can only be shared with the relevant authors.” Planck died in 1947.
Sony's notice blames “content licensing agreements” for pulling Terminator 2, Total Recall and 549 other films from purchased libraries. It's not the first time — the company tried the same move years ago.
The recurring demand in the thread is legal, not technical: stop letting a checkout button say “purchase” when what you get is a revocable license with no end date disclosed.
The article argues subscriptions are heavily subsidized and per-token API pricing is the real, brutal number — $54 for one afternoon of TypeScript cleanup. With GLM-class open models beating flagships on coding benchmarks at a fraction of the price, the author sees routine work commoditizing.
The most concrete reply isn't a forecast — it's a deployment: banks and insurers already run anything that matters on models inside their own datacenters, with Claude and Codex kept on a tight leash.
The bill orders 3D printers sold in California to run compliance software and reject jobs from any pathway designed to evade detection — effectively banning open slicers unless they ship the censorship layer too. The stated goal is stopping untraceable firearms.
The thread's analogy did the demolition: imagine a lathe that refuses to make a baseball bat, or scissors that phone home before cutting. You don't ban the tool to ban the act.
The pitch: a Dimensity 7100, replaceable battery, assembled in Finland, “no tracking, no calling home,” and Android-app support on a real Linux OS. The price is €649, with a refundable down payment to lock a delivery slot.
The skepticism is specific, not vague: Sailfish carries closed-source UI components, MediaTek chips spook the bootloader crowd, and banking and 2FA apps remain the wall every Linux phone hits.
On a single intelligence index the open-vs-closed gap projects to zero by Christmas; spread across 18 benchmarks it's stayed near five months for two years. Coding is the one clear mover, from 15 months behind to one or two.
The thread's reframe: the number doesn't matter, because closed access can be revoked and an open weight can't be put back in the bag — whatever capability it reaches stays reachable forever.
The argument: capital protects you only while the state needs the value humans create. Strip that out and property rights are whatever an AI-run state decides to honor. The rich, Borretti claims, “just don't have that much political power” once they're materially redundant.
The thread's best rebuttal isn't optimism — it's that the essay smuggles in its conclusion, quietly upgrading “can do all work” to “runs autonomously, has its own goals, is conscious,” until the doom is just a tautology.
The technique injects sparse, FDA-approved sulfur-hexafluoride microbubbles and uses their isolated blur to localize each one far below the diffraction limit — building a vascular map 100x finer by volume than CT, then decoding neural activity from blood flow.
The expert pushback is blunt: a neuroscientist cites papers that even low-dose ultrasound disrupts myelination, and an ultrasound-device engineer asks the question the post never answers — how does it see through bone at all?
One camp says nothing built with an AI exceeds what its operator could have built alone — the CS background is what shapes the prompt, and “knowing Dijkstra by heart” matters less than architecture. Another says the comparison was settled years ago: outsourcing to an LLM is the same trust trap as outsourcing to a dev shop.
A developer in China reframes the cost worry entirely: a $4,699 box already runs a smarter model locally, and a new state-of-the-art lands every two months.
Ravid's claim: a keyboard or a car achieves “tool magic,” becoming an extension of your body, while LLMs are never consistent or fast enough to do that — so you pay conversational effort and get “more code, more tests, more excuses” back.
The counter, from the people the argument forgot: for the deeply introverted, an LLM is the first interface with no social tax at all — no bruised egos, no negotiation, no one to tire out.
The tool is a drop-in proxy across Anthropic, OpenAI, Gemini and open models, scoring each request and routing accordingly, keys encrypted, fully self-hostable. The claim is 40–70% lower cost “with just an endpoint change.”
The objection is structural: agentic sessions live on prompt caching with a five-minute TTL, and the more distinct models you touch, the more cache misses you eat — so a one-planner, one-executor setup already captures the win without the churn.
The Economist frames data-center fights as a preview of bigger struggles and warns that if America yields to “popular rage” it could cede the AI frontier. Several readers read that last line as a Too-Big-To-Fail argument in waiting.
The reframe that resonated: the problem isn't a promising new tool, it's that a few firms already dominate daily life — the 2007 plucky-underdog Google is now big enough that its size alone is the threat.
Gossamer (v0.19.1) offers Result/Option, traits, typed channels on an M:N scheduler, no async coloring, no unsafe, no macros — memory via deterministic reference counting plus arena regions. It reads like a greatest-hits of modern language design.
The takedown was specific: dropping the last reference to a 10,000-node cyclic graph is a stall, so “pause-free” is marketing; and the compiler is written in Rust, which for a self-described systems language is the tell.
The work-distance argument carries it: force times distance equals energy, and a faster object travels farther under the same braking force, so the energy isn't 2x but 4x. A reader's two-cars puzzle turns it into a gut-level result.
The thread also surfaces the honest discomfort — one accomplished programmer admits physics' ‘obvious’ cause-and-effect steps have never clicked for him, and asks if he's alone. He isn't.
The complaint isn't that the model erred once — it's that the same question yields different answers each run, with a fabricated fact surfacing a fraction of the time and no mechanism to detect or kill the outliers.
The substantive turn comes via Bruce Schneier's argument: if companies can hide behind “the AI did it,” they get a liability shield no human writer, lawyer, or doctor ever had.
Barcode fonts are old, but this set's flourish is computing the EAN-13 check digit inside the font's typographic features rather than asking you to do it. It's downloadable from Google Fonts.
The expert pushback is practical: generating a correct Code 128 as an SVG is about the same effort, and the genuinely hard part — switching between character sets — is exactly what a font can't handle well.
The essay marshals the usual evidence — front pages dominated by wrongdoing, readers wildly overestimating violent death, fire-alarm fatigue — to argue daily news is a net negative you should simply drop, swapping it for long-form and local focus.
The thread refuses the clean exit. A reader in Ukraine describes a brain “completely broken” by news he can't look away from; a journalist argues opting out is how democracies rot.
