← 2026-06-14
Theme: Auto
Font: Serif
About
Kobo e-readers reject valid EPUB files due to Adobe's outdated CSS parser frozen since 2013, despite passing standard validation.
Commenters fault Adobe for non-spec-compliant CSS parsing; suggest .kepub.epub workaround and targeting older EPUB versions for portability.
787 pts · 263 comments
Becoming a billionaire requires exponential growth and market size, not wrongdoing; the math is achievable through user satisfaction.
Commenters debate whether startup-driven wealth requires worker exploitation and whether creative destruction's costs are adequately addressed.
673 pts · 1748 comments
Kage mirrors websites into offline-browsable folders or binaries with JavaScript stripped out.
SingleFile offers similar functionality with single-HTML output; users requested cookies, markdown export, smaller binaries, and SPA support.
631 pts · 122 comments
Curl will suspend vulnerability reports throughout July 2026 to give maintainers a planned break.
Strong approval for maintainer vacation; skeptics note attackers won't wait regardless.
621 pts · 251 comments
Tech leaders have abandoned nerd values for attention-seeking self-promotion, liquidating decades of trust.
Commenters distinguish real nerd culture from VC-driven grift; actual nerds still thrive on non-mainstream platforms.
615 pts · 392 comments
Windows 11 users report growing frustration with Microsoft account requirements expanding across more OS features.
Users feel locked into Windows despite account requirements; Linux alternatives face software compatibility and setup friction barriers.
487 pts · 350 comments
About one-third of Americans actively use AI, one-third occasionally, one-third never, contradicting claims that everyone uses it.
Engineers describe widespread pressure to add AI unnecessarily, degrading system performance; LLM value is better spent generating code for deterministic tools.
484 pts · 519 comments
Developer indexed 668 GB of GoPro cycling videos using M1 Max and local ML models for searchable clip extraction.
DaVinci Resolve 21 offers competing built-in indexing; M1 Max's unified memory architecture enables the performance.
391 pts · 102 comments
Rio de Janeiro's Rio-3.5-Open-397B model is a 60/40 weighted merge of Nex-N2 Pro and Qwen3.5-397B, not an original training effort.
Rio updated attribution after publication but claims accidental upload; no corrected model released yet.
380 pts · 201 comments
Free browser-based tool converts SQL CREATE TABLE statements into interactive entity-relationship diagrams with no data upload.
Commenters praise mobile usability and code quality; creator discusses performance tradeoffs between JavaScript and WASM.
353 pts · 73 comments
Anthropic releases a Swift package integrating Claude into Apple's Foundation Models framework for iOS 27 and later.
Commenters split on whether Apple designed this for developer convenience or to facilitate monetization and lock-in when Apple's own models mature.
339 pts · 153 comments
Jane Street shifts from skepticism to investment in formal methods, driven by AI-generated code verification needs.
Skepticism centers on whether AI-written proofs genuinely reduce verification burden or merely shift complexity; defenders note proofs self-verify where code doesn't.
322 pts · 99 comments
Linux kernel 7.1 released. Linus Torvalds traveling during merge window with potential latency.
Commenters treat 7.1 as unremarkable; discussion shifts to Debian stable kernel timelines and local kernel compilation practices.
306 pts · 116 comments
Article surveys lesser-known built-in Emacs features like dictionary tooltips, wildcard file operations, and comparison commands.
Dired's keybinding UX frustrates power users; vanilla Emacs stability versus framework-based instability divides adopters.
291 pts · 93 comments
Monthly thread collecting projects developers are building, ranging from search engines to health trackers.
Practical feedback on tooling and launches; Uruky's developer disclosed 150 active accounts and upcoming crypto payment support.
281 pts · 1007 comments
LLM context windows degrade sharply around 100k tokens regardless of advertised size; treat available context as a budget.
Commenters report vastly different degradation points: some see problems at 60k, others push past 500k-800k without issues; likely model and configuration dependent.
263 pts · 187 comments
Hand-crank and pedal-powered local AI system that runs models on-device without cloud infrastructure or data transmission.
Technical feasibility confirmed; humans generate 120-160W sustainably, making local LLM inference viable but current cloud models impractical.
232 pts · 94 comments
A 2014 talk tracing JavaScript's history from 1995 to a speculative 2035, examining the language's flaws and industry impact.
Commenters reject the death premise, viewing JavaScript as a persistent compilation substrate rather than a dying language.
231 pts · 130 comments
New study links Atlantic's 'cold blob' to weakening ocean currents that could reach a tipping point.
Commenters stress AMOC collapse's severity is underestimated; agricultural collapse in Europe could follow within decades.
228 pts · 358 comments
Game where you play as a ghost hunting AI-controlled Pac-Man in a maze.
Heavy criticism of control lag and input latency; some confirm game is winnable by strategy.
213 pts · 80 comments
HN Flash, daily summary of top 20 stories on Hacker News