Jun 21Monday, June 22, 2026 · all days
1.Deno Desktop(docs.deno.com)
593 points by GeneralMaximus 7 hours ago | 217 comments | permalink
tl;dr: Deno 2.9 introduces `deno desktop`, a command that bundles a Deno/TypeScript project (including Next.js, Astro, SvelteKit, etc.) into a self-contained cross-platform desktop binary with the Deno runtime and a webview. It defaults to the OS's native webview for smaller binaries (with optional bundled Chromium), uses in-process bindings instead of IPC, supports cross-compilation from a single machine, and includes built-in binary-diff auto-updates. The feature is currently only available in the canary build and APIs may change before stabilization.
HN Discussion:
  • Excited about Deno Desktop as a compelling Tauri/Electron alternative with TypeScript-first approach
  • Skeptical of framing web tech as a UI toolkit since it doesn't deliver native look and feel
  • Questions and curiosity about specific technical details like permissions, CEF versioning, and hidden windows
  • ~Concerned that Deno is becoming a bundled integration toolkit rather than an agnostic runtime
  • Prefers truly native UIs or non-browser-based alternatives like WASM bundling
2.Help I accidentally a wigglegram(lmao.center)
331 points by gregsadetsky 2 days ago | 71 comments | permalink
tl;dr: A photographer realized their habit of taking many near-identical shots from slightly different angles had accidentally produced years of raw material for "wigglegrams" (stereoscopic GIF loops). They wrote a script using perceptual hashing and Hamming distance to scan their iCloud library, find clusters of similar images, and auto-stitch them into wigglegrams. The script is on GitHub and works with Mac iCloud libraries or any image directory.
HN Discussion:
  • Sharing related tools or libraries for stereoscopic/pseudo-3D visualization
  • Appreciation for the website design, dithering aesthetics, and hand-written (non-AI) code
  • The wigglegram effect actually reduces depth illusion and looks like disjointed animation
  • Mentions of similar motion-capturing alternatives like phone Top Shot features or short videos
  • Interest in using the script to handle large personal photo libraries with duplicates
3.Did my old job only exist because of fraud?(david.newgas.net)
646 points by advisedwang 15 hours ago | 278 comments | permalink
tl;dr: A software engineer reflects on his early-career job at GenieDB, a UK startup acquired by Stuart Frost's VC fund, which later faced SEC fraud charges for charging portfolio companies excessive fees. Digging into court records, the author found evidence suggesting GenieDB may have been kept alive primarily to siphon investor money through fees, though the core technology concept was legitimate. He grapples with the realization that the fraud indirectly shaped his entire life, including his move to the US.
HN Discussion:
  • Sharing similar personal experiences of working in companies involved in fraud or waste
  • Corporate waste and squandered budgets are widespread and akin to investor fraud
  • The author shouldn't feel shaken since they did honest work and benefited
  • ~Many jobs are essentially pointless or ethically questionable, not just fraud-related ones
  • Providing supporting reference material about Stuart Frost's fraud
4.Apertus – Open Foundation Model for Sovereign AI(apertvs.ai)
446 points by T-A 15 hours ago | 145 comments | permalink
tl;dr: Apertus is a fully open foundation model from the Swiss AI Initiative (EPFL, ETH Zurich, CSCS), releasing weights, training data, code, and methods under reproducible terms. Available in 8B and 70B parameter sizes, it's trained on 1000+ languages and designed for EU AI Act compliance, including opt-out respect, PII removal, and memorization prevention. Swisscom is a strategic partner.
HN Discussion:
  • Other fully open LLMs already exist and may outperform Apertus
  • Project moves slowly like a committee and won't produce competitive models
  • Model quality is poor based on testing (strawberry test, language hallucinations, copyright claims)
  • Real value is the team's experience gained for future iterations
  • Open foundation models threaten commercial AI labs and represent the future
5.There is minimal downside to switching to open models(marble.onl)
275 points by amarble 16 hours ago | 227 comments | permalink
tl;dr: The author argues that switching from proprietary LLMs (Claude, GPT) to open-weight models now carries only a minor productivity penalty, similar to how the Linux vs. Windows gap has narrowed over time. Triggered by Claude's new ID verification requirement, he's moving to open models—which trail leaders by only a few months and have decent tooling—while acknowledging tradeoffs around privacy (third-party hosts like OpenRouter feel less trustworthy) and the cost/complexity of self-hosting.
HN Discussion:
  • Open models a few months behind is fine since older proprietary models were already useful
  • Open models aren't capable enough yet and self-hosting economics don't work out
  • Proprietary models from Anthropic/OpenAI remain significantly better in real-world use
  • ~Switching is feasible but harness/tooling compatibility creates non-trivial friction
  • Exploring practical alternatives like local collaboratives or third-party routers for open model access
6.Everything is logarithms(alexkritchevsky.com)
245 points by E-Reverance 15 hours ago | 50 comments | permalink
tl;dr: The author argues that logarithms should be thought of as "baseless" objects analogous to coordinate-free geometric vectors, where writing log_b(x) is just choosing units (like "bits" or "nats") much like projecting a vector onto a basis. From this lens, many seemingly unrelated math operations—p-adic valuations, orders of vanishing in complex analysis, vector space dimension, and even translation operators in differential geometry—are all essentially logarithms or logarithm-like projections. The piece is speculative numerology aimed at uncovering a unifying coordinate-free framework that the author suspects underlies these scattered notations.
HN Discussion:
  • Baseless logarithms are formally torsors, supporting the article's coordinate-free framing
  • ~Logarithms as a single physical quantity with base-as-unit is correct, but 'baseless' terminology is misguided
  • The essay is sloppy and needs rigorous type/domain specifications and Lie theory grounding
  • Notational play without demonstrating a novel mathematical result limits the article's value
  • Appreciation for the power and elegance of logarithms as a computational/conceptual tool
7.Identity verification on Claude(support.claude.com)
802 points by bathory 1 day ago | 669 comments | permalink
tl;dr: Anthropic is rolling out identity verification for certain Claude use cases, requiring government-issued photo IDs processed through third-party partner Persona Identities. The verification applies to specific capabilities and compliance checks, with Anthropic stating the data is used solely for identity confirmation. Failed verifications can be retried, and users whose accounts are banned post-verification can appeal through a form on claude.ai.
HN Discussion:
  • ID verification harms non-US users and accelerates competition from international LLMs
  • This policy isn't new and has existed for months, contextualizing the panic
  • Concerns about AI neutrality and silent blocking based on usage patterns
  • Persona as verification provider is problematic given past incidents like Discord's
  • Cancelling Claude due to surveillance concerns and government pressure
8.JSON-LD explained for personal websites(hawksley.dev)
234 points by ethanhawksley 18 hours ago | 75 comments | permalink
tl;dr: A practical guide to implementing JSON-LD structured data on personal websites using Schema.org vocabulary, with copy-paste examples for the key node types: WebSite, WebPage, Person, ProfilePage, SoftwareApplication, BreadcrumbList, CollectionPage, Blog, and BlogPosting. The author explains how nodes link via `@id` references, which properties matter most for SEO and LLM crawlers (notably `sameAs` for identity disambiguation), and recommends that even static sites add at minimum WebSite, ProfilePage, and Person to their root page.
HN Discussion:
  • JSON-LD's SEO benefits are obsolete as LLMs now scrape and replace original content in search results
  • ~Better to consult search engine documentation directly since only specific JSON-LD subsets are supported
  • OpenGraph is more widely supported than JSON-LD for link previews, questioning the article's utility
  • Re-expressing semantics in JSON script tags is redundant when semantic HTML already exists
  • Helpful supplementary context, recommending Schema.org structured data and tooling resources
9.Prefer duplication over the wrong abstraction (2016)(sandimetz.com)
503 points by rafaepta 20 hours ago | 319 comments | permalink
tl;dr: Sandi Metz argues that premature or incorrect abstractions are more costly than duplicated code, because subsequent developers tend to preserve the abstraction by piling on parameters and conditionals, eventually producing incomprehensible code. Her recommended fix is to fight the sunk cost fallacy: inline the abstraction back into each caller, strip out the unused branches, and let the resulting duplication reveal a better abstraction. The fastest way forward, when an abstraction is wrong, is backward.
HN Discussion:
  • ~Single source of truth must override duplication preference when divergence causes bugs
  • Agrees that under-engineered code is easier to work with than over-engineered code
  • Abstractions are actually underused and enable solving bigger problems
  • Personal experience confirms DRY-driven abstractions create unmaintainable messes requiring duplication to escape
  • ~Functional programming or alternative paradigms largely sidestep these abstraction problems
10.Danish privacy activist Lars Andersen raided by police(twitter.com)
297 points by I_am_tiberius 8 hours ago | 250 comments | permalink
tl;dr: Danish privacy activist and ex-cop Lars Andersen was arrested by masked police who broke down his door after he obliquely published PM Mette Frederiksen's social security and phone numbers, alongside criticism of her encryption-ban and mass-surveillance proposals. According to Andersen, officers immediately cut power to his router and seized his Nest cameras (which had local storage) to prevent the arrest from being recorded, and refused to state the charges against him.
HN Discussion:
  • ~Andersen is a problematic figure who oversteps but raises some valid points about government hypocrisy
  • Andersen is not a credible privacy activist, but a right-wing extremist or drug-addled idiot
  • Using Google Nest cameras undermines his credibility as a privacy advocate
  • Police tactics (masked, cutting power, seizing cameras) were dangerous and concerning
  • Technical advice on making surveillance setups more resilient (UPS, hidden cameras)
11.Beyond All Reason (Free Total Annihilation Inspired RTS)(beyondallreason.info)
482 points by mosiuerbarso 1 day ago | 289 comments | permalink
tl;dr: Beyond All Reason is a free, open-source RTS inspired by Total Annihilation, featuring real-time simulation of every unit, projectile, and explosion, plus terrain deformation and ballistic physics. Gameplay emphasizes resource scaling, terrain-driven strategy, and large-scale battles across 10+ unit classes including Experimental units. Maps play distinctly due to terrain affecting radar, unit effectiveness, and tactics like nuclear strikes that physically reshape the battlefield.
HN Discussion:
  • ~Game is technically impressive and well-made but the community is toxic and unwelcoming to new players
  • Nostalgic appreciation for Total Annihilation and its legacy in the RTS genre
  • Recommends related TA-inspired alternatives like Zero-K and FAF for better experiences
  • ~BAR lacks the atmospheric single-player feel of original TA, preferring PvE over PvP clickfest
  • Concerns about governance drama with admins selling out to a publisher despite open-source roots
12.Google Hits 50% IPv6(blog.apnic.net)
405 points by barqawiz 1 day ago | 430 comments | permalink
tl;dr: Google's measurements show IPv6 adoption has reached 50% of users accessing its services, though APNIC Labs' independently weighted measurements (using World Bank data to adjust for uneven ad sampling across economies) put the figure closer to 42%, with the two datasets effectively bracketing actual global capability. Adoption remains uneven, with newer entrants like Reliance Jio going IPv6-first while incumbents maximize IPv4 investments via NAT/CGNAT, and interoperability is handled at higher layers (TCP/UDP/QUIC) rather than at the IP layer itself.
HN Discussion:
  • ISPs are dragging their feet on IPv6 deployment despite years of promises
  • ~Full IPv6 adoption is blocked by major services and infrastructure still lacking support
  • Economic incentives favor keeping IPv4 via NAT/CGNAT and selling static IPs
  • ~IPv6 growth is slow and linear, comparable to other lengthy infrastructure transitions
  • Consumer routers and weekend usage patterns reveal IPv6 gaps in home and corporate networks
13.Tell HN: Happy Fathers Day
308 points by consumer451 20 hours ago | 51 comments | permalink
tl;dr: A Hacker News user shares a Father's Day tribute to his uncle, who filled the dad role in Soviet-era Poland. Memorable hacks included launching Estes-style rocket cars at a quarry and rigging a stolen Milicja (military police) siren onto his banana bike, wired up to actually work thanks to his uncle's EE skills.
HN Discussion:
  • Fatherhood is transformative and the most joyful, rewarding experience of life
  • First-time fathers sharing emotional reflections on early days with their child
  • Sharing memories of own fathers' hands-on skills and influence, echoing the article's tribute
  • Practical unsolicited advice for new fathers, like buying arch support
  • Sharing related external content (corporate Father's Day message) that resonated
14.The brain was not designed for this much bad news(sciencedaily.com)
428 points by colinprince 1 day ago | 319 comments | permalink
tl;dr: News avoidance is at record highs (69% in Canada, 40% globally) because the brain's evolved negativity bias—designed to detect local threats—is now being overwhelmed by a constant global feed of bad news, with 17% of US adults showing severe "Problematic News Consumption." The fix isn't avoidance but deliberate consumption: time-boxed sessions, long-form over social media bursts, avoiding rage bait, and closing the gap between awareness and actionable agency.
HN Discussion:
  • ~People have unrealistic expectations and should accept that things go wrong rather than overreact
  • Attention-grabbing media exploits the brain's threat-detection wiring
  • The internet has shifted from extending the mind to creating noise and negativity
  • Books like Factfulness reveal the world is improving despite our negativity bias
  • Personal coping strategies like local-only news or black-and-white viewing reduce emotional impact
15.GPT-5.5 hallucinates 3x more than MIT-licensed GLM-5.2(arrowtsx.dev)
565 points by oshrimpton 2 days ago | 288 comments | permalink
tl;dr: Open-weight GLM-5.2 (753B params) scores within 4 points of GPT-5.5 on the AA Intelligence Index while hallucinating far less (28% vs 86%), suggesting that scaling parameters and training data has plateaued and often worsens uncertainty calibration. The author argues models should be evaluated on a trilemma of raw capability, hallucination rate, and compute efficiency rather than size alone, citing a coding test where DeepSeek V4 Pro burned 10x more reasoning tokens to produce a confidently wrong answer that GLM-5.2 solved in 12 seconds.
HN Discussion:
  • Skepticism that bigger models cause more hallucination, contradicting recent trends
  • ~Hallucination rate metrics are conditional and shouldn't be attributed solely to model size
  • RLVR could easily train models to say 'I don't know' on uncertain answers
  • Raw accuracy matters more than abstention; verifiable answers preferred over 'I don't know'
  • Suspicion that the article is self-promotional, cherry-picking rate over overall accuracy
16.15-minute at-home Lyme disease tick test(bostonglobe.com)
244 points by bookofjoe 4 days ago | 154 comments | permalink
tl;dr: LymeAlert, a $40 at-home test launching in August, lets users grind up to five ticks in a container and detect Lyme bacteria via chemically treated paper in 15 minutes, helping avoid unnecessary precautionary antibiotics. Created by a PA and MIT Sloan grad, the product also includes an app that anonymously maps infected tick locations, combining the data with NASA satellite and animal migration info for AI-driven spread predictions. Critics note it won't detect other tick-borne pathogens like Alpha-gal, though a multi-pathogen version is reportedly planned for next year.
HN Discussion:
  • Skepticism about the test's usefulness since negative ticks don't rule out infection from unfound ticks
  • Personal anecdotes about suffering from possible tick-borne illness or alpha-gal concerns
  • Pointing out existing alternatives like prophylactic doxycycline or cheaper European tests
  • Highlighting the test's limitation in not detecting other tick-borne pathogens like alpha-gal
  • Questioning the scientific validity and effectiveness of the reactive strip technology
17.Loupe – A iOS app that raises awareness about what native apps can see(github.com)
530 points by Cider9986 2 days ago | 230 comments | permalink
tl;dr: Loupe is an open-source iOS/iPadOS app from Mysk that demonstrates the device fingerprinting surface exposed to third-party apps, displaying raw values from public APIs grouped into three tiers: passive (no prompt), permission-gated, and advanced side-channel techniques like `canOpenURL` probing and Keychain persistence. All data stays local unless explicitly exported. The app was written almost entirely with AI coding tools and serves partly as promotion for Mysk's privacy-focused browser, Psylo.
HN Discussion:
  • Shock and concern at the amount of fingerprintable data exposed to apps without permission
  • Apple should provide finer-grained controls and better disclosure about app data access
  • ~Correction/clarification that iOS already restricts some access like installed app enumeration
  • Apps should simply not collect extra data in the first place as the real solution
  • Sharing similar awareness tools or demos for other platforms
18.Developers don't understand CORS (2019)(fosterelli.co)
381 points by toilet 1 day ago | 260 comments | permalink
tl;dr: Zoom's recent vulnerability stemmed from a hack to bypass CORS: their localhost webserver encoded response data in image dimensions to avoid AJAX restrictions, inadvertently exposing the API to every website. A proper fix would have been setting `Access-Control-Allow-Origin: https://zoom.us` on the localhost server. The author argues this reflects a broader pattern of developers misunderstanding CORS and working around it insecurely rather than configuring it correctly.
HN Discussion:
  • The article itself misunderstands CORS, conflating it with SOP and misrepresenting what Access-Control-Allow-Origin actually does
  • The comment section proves the article's point that developers don't understand CORS
  • CORS is genuinely confusing because of poor error messages, changing standards, and infrequent exposure for many developers
  • ~Understanding CORS requires first understanding the Same Origin Policy and the underlying threat model
  • Recommending educational resources like MDN to help developers actually learn CORS properly
19.Show HN: TownSquare, a tiny presence layer for websites(townsquare.cauenapier.com)
260 points by cauenapier 2 days ago | 147 comments | permalink
tl;dr: TownSquare is a drop-in `<script>` tag that adds a real-time presence layer to any website, letting visitors see each other as avatars, move around, chat, and interact (jump, high-five) without accounts or signups. It pitches itself as a way to make static web pages feel "inhabited," with a network of participating sites shown on a shared map.
HN Discussion:
  • ~Live demo is overrun with offensive spam, undermining the concept's appeal
  • ~Moderation is the critical challenge, potentially solvable via AI/LLM classification
  • ~Client-side moderation with user-controlled filters would be a better approach
  • Similar prior art exists (Third Voice, Cursor Party, p2p chat extensions)
  • ~Cute concept but hesitant to actually adopt it on own site
20.Slow breathing modulates brain function and risk behavior(cell.com)
392 points by croes 1 day ago | 113 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • Personal experience confirms slow breathing's calming and confidence-boosting effects
  • Surprised by counterintuitive finding that parasympathetic activation increases risk-taking
  • Curious about extending breathing benefits to subconscious patterns and monitoring tools
  • Skeptical due to conflicting claims from other breathing research
  • ~Cautions that suppressing fear via breathing isn't always desirable