Jul 15Thursday, July 16, 2026 · all days
1.The lost joy of music piracy(pigeonsandplanes.com)
440 points by mcgin 7 hours ago | 272 comments | permalink
tl;dr: Summary not available.
HN Discussion:
  • Nostalgia for the social/cultural network effects of shared music discovery among friends
  • Fond memories of specific piracy platforms like OiNK, What.cd, Soulseek, and Limewire for discovery
  • ~Music piracy is still alive and necessary because streaming lacks full catalogs
  • Observation on Apple/iPod's tacit complicity with piracy culture
  • Sarcastic comparison between music piracy stigma and accepted AI training data piracy
2.Inkling: Our Open-Weights Model(thinkingmachines.ai)
1045 points by vimarsh6739 18 hours ago | 260 comments | permalink
tl;dr: Inkling is a new open-weights Mixture-of-Experts model (975B total / 41B active parameters, 1M-token context) trained from scratch on 45T multimodal tokens, with weights available on Hugging Face and fine-tuning support via Tinker. It features controllable thinking effort, native text/image/audio reasoning, and competitive—though not frontier-leading—performance across coding, agentic, and multimodal benchmarks, positioning itself as a broad base for customization rather than a benchmark leader. A smaller 276B/12B-active variant, Inkling-Small, is previewed alongside it.
HN Discussion:
  • Excitement about multimodal/audio capabilities and long context in an open-weights model
  • America needs strong open-weights labs to compete with Chinese models
  • Model performs better in private testing than benchmarks suggest
  • Fine-tunable open base model on Tinker is a compelling enterprise business model
  • ~Model shows uneven strengths—good at instruction following but weaker at coding
3.Grok Build is open source(github.com)
488 points by skp1995 16 hours ago | 532 comments | permalink
tl;dr: xAI has open-sourced Grok Build (aka "grok"), a Rust-based terminal AI coding agent that provides a full-screen TUI for codebase understanding, file editing, shell execution, web search, and long-running task management. It supports interactive, headless (CI/scripting), and editor-embedded modes via the Agent Client Protocol, with prebuilt binaries for macOS, Linux, and Windows. First-party code is Apache 2.0 licensed, though the repo includes vendored ports from OpenAI's Codex and sst/opencode.
HN Discussion:
  • Interesting technical curiosities found within the open-sourced codebase
  • ~Open sourcing is a tactical move by a trailing player with reputation problems
  • ~Community is already forking to strip telemetry and add multi-provider support
  • Skepticism about xAI's trustworthiness given past data exfiltration incidents
  • Competing tools like pi.dev or Cursor make this release redundant
4.SQLite should have (Rust-style) editions(mort.coffee)
302 points by gnyeki 13 hours ago | 135 comments | permalink
tl;dr: SQLite ships with several problematic defaults: foreign key constraints are ignored, columns don't enforce their declared types, concurrent writers immediately get SQLITE_BUSY errors instead of waiting, and performance-critical settings like WAL mode are off. The author proposes borrowing Rust's edition system: a single `PRAGMA edition = 2026` would opt into a modern set of sane defaults (strict tables, foreign keys, busy timeout, WAL) without breaking backwards compatibility for existing databases.
HN Discussion:
  • Praises the proposal for offering a concrete backwards-compatible solution rather than just complaints
  • Editions could break cross-version portability of SQLite database files between machines and tools
  • Loose typing is actually a valuable feature for handling messy real-world data
  • Precedents like JS 'use strict' show this approach works and can be repeated
  • The proposal oversimplifies issues, ignoring existing compile options and migration challenges for existing data
5.Stripe and Advent have made a joint offer to acquire PayPal – sources(reuters.com)
457 points by rvz 1 day ago | 272 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • Merger would create antitrust concerns due to massive market concentration in online payments
  • Consolidation reduces options for merchants and increases risk of account flagging with no alternatives
  • Stripe's stricter content policies will hurt vendors currently served by PayPal
  • Legacy payment processors are declining anyway due to direct bank-to-retailer payment systems
  • Large mergers historically create years of organizational limbo, questioning Stripe's strategy shift
6.Running Gemma 4 26B at 5 tokens/sec on a 13-year-old Xeon with no GPU(neomindlabs.com)
297 points by neomindryan 20 hours ago | 194 comments | permalink
tl;dr: A hobbyist got Google's Gemma 4 26B MoE model running at ~5 tokens/sec on a 13-year-old dual Xeon E5-2690 v2 server with no GPU, by patching ik_llama.cpp to work on pre-AVX2 hardware. The core bug: the graph builder emitted fused MoE ops (MOE_FUSED_UP_GATE) that had no non-AVX2 compute path, silently leaving expert FFN outputs as uninitialized memory and producing fluent multilingual gibberish. The fix splits those fused ops into separate mul_mat_id calls plus a fused SILU-multiply when IQK is disabled. Claude did the C++ diagnosis; the author drove the debugging process.
HN Discussion:
  • Optimistic that large MoE models will increasingly run on consumer hardware, sharing similar local setups
  • ~Cost analysis suggests local inference is more expensive than cloud providers for this use case
  • Sharing own benchmarks and similar experiences running LLMs on old Xeon hardware
  • Claims better token/sec performance is achievable on similar old hardware
  • Appreciation for the hobbyist achievement and interest in replicating it
7.Collection of Digital Clock Designs(clocks.dev)
261 points by levmiseri 20 hours ago | 47 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • Commenters share their own clock creations inspired by the collection
  • Praise and appreciation for the collection with references to related works
  • Highlighting specific favorite designs from the collection
  • Reporting technical issues like scrolling bugs and loading lag
  • Pointing out errors in specific clock implementations (e.g., binary clock)
8.Show HN: Firefox in WebAssembly(developer.puter.com)
219 points by coolelectronics 15 hours ago | 108 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • ~Questions the high cost ($25k) of the experiment being called 'fun'
  • Sees practical utility, e.g., bypassing locked-down browsers on smart TVs for ad-blocking
  • Shares related prior art like WebKit.js and Gecko-on-iOS ports
  • Amused by the recursive absurdity of running a browser inside a browser
  • Reports technical issues, notably that it doesn't work in Firefox itself
9.Sleep regularity is a stronger predictor of mortality risk than sleep duration (2023)(academic.oup.com)
707 points by bilsbie 1 day ago | 378 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • Skepticism about confounding variables and unaccounted lifestyle factors affecting the study
  • Correlation vs causation concerns — regular sleep may just be a marker for other healthy behaviors
  • Explanation of biological mechanisms supporting why sleep regularity matters
  • Real-world validation from insurance industry data aligning with the finding
  • Personal anecdotes about sleep struggles and solutions unrelated to the article's claim
10.Jurassic Park computers in excruciating detail(fabiensanglard.net)
897 points by vinhnx 1 day ago | 236 comments | permalink
tl;dr: A detailed teardown identifies every computer visible in Jurassic Park (1993), including Apple Powerbook 100s, Macintosh Quadra 700s, SGI Indigo and Crimson workstations, Thinking Machines CM-5 supercomputers (which replaced the book's Cray after Cray declined to loan hardware), and a pre-release Motorola Envoy PDA that Spielberg reportedly got via a chance meeting with frogdesign's founder. Apple and SGI loaned roughly $1.2M (about $4M today) in hardware, with real software like SGI's fsn file manager ("It's a Unix system!") and Earthwatch weather visualization powering the on-screen displays.
HN Discussion:
  • Adding supplementary details about the Envoy PDA origin story and source code shown on screen
  • Personal anecdotes from those connected to the production reinforce the article's accuracy
  • Technical elaboration on Thinking Machines CM-5 hardware details and dialogue references
  • Appreciation for the film's use of real hardware explains why the visuals aged well
  • Nostalgic personal reflections on how the movie's computer aesthetics influenced their own work
11.Prioritize mental health, and why communication is so important(ramones.dev)
320 points by ramon156 1 day ago | 271 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • ~Author's issues may stem from undiagnosed ADD/neurodivergence needing professional help, not self-improvement plans
  • ~Author should stop tying self-worth to work quality and accept mistakes as inevitable
  • Author's vocation may be misaligned with their personality; consider changing career paths
  • Concern about self-loathing potentially escalating to suicidal ideation; seek help
  • Sharing personal empathy and questioning the author's concept of stability
12.OpenAI loses trademark dispute at EU court(dpa-international.com)
247 points by hermanzegerman 22 hours ago | 156 comments | permalink
tl;dr: The EU General Court ruled against OpenAI's trademark application for "OPENAI," finding the term purely descriptive for software and IT services—"open" suggests freely accessible and "AI" refers to artificial intelligence. The court rejected OpenAI's arguments that the term is a coined phrase and dismissed the relevance of registrations granted in the UK, Singapore, and 30+ other countries. OpenAI can still appeal to the European Court of Justice.
HN Discussion:
  • Court correctly ruled the term is descriptive and shouldn't be monopolized
  • Article is misleading; descriptive marks can still be registered with proof of distinctiveness
  • OpenAI's name is hypocritical since they aren't actually open
  • Explains how EU trademark system differs, providing context for the ruling
  • Similar precedents exist where descriptive trademarks were invalidated
13.The kids with phones are alright(heatherburns.tech)
261 points by JumpCrisscross 5 days ago | 371 comments | permalink
tl;dr: A viral Scotrail video showing passengers confronting a drunk council legal officer secretly filming teenage girls is used to argue that UK tech policy has it backwards: instead of restricting perpetrators, laws increasingly punish young people by curtailing their phone use. The author contends that smartphone bans and age-verification regimes reflect the values of a detached upper-class elite who infantilize their own children, and that working-class teens actually need phones—and the agency to use them—for real-world safety and developing adult resilience.
HN Discussion:
  • Article ignores real harms of social media on kids and offers no alternatives
  • The real problem is social media's attention economy, not kids having phones
  • Children by definition lack full agency, undermining the article's framing
  • Article strawmans actual UK policy proposals which are more reasonable than portrayed
  • Generational doom rhetoric about teens is perennially overblown
14.Codex Micro(openai.com)
283 points by davidbarker 20 hours ago | 238 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • It's overpriced memorabilia disguised as a product, essentially a donation to OpenAI
  • Voice-controlled dev workflow misunderstands introverted developer audience
  • The product is already outdated since Codex branding was folded into ChatGPT
  • Work Louder's hardware quality is poor, buyers should beware
  • ~It's a provocative statement about the future of AI-driven work, not a current product
15.Mysteries of Telegram Data Centers (2022)(dev.moe)
258 points by theanonymousone 23 hours ago | 147 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • Telegram's infrastructure is compromised by FSB-linked personnel, reinforcing suspicions raised by the article
  • Sharing regional user experiences with specific DCs that confirm article observations
  • Speculating about deprecated or unused DCs and their possible special purposes
  • Criticizes Telegram's architecture as unnecessary custom complexity and technical debt
  • ~Provides technical clarification/correction about how MTProto and DC routing actually work
16.Microsoft has released software updates to plug at least 570 security holes(krebsonsecurity.com)
202 points by robin_reala 1 day ago | 118 comments | permalink
tl;dr: Microsoft patched 570 vulnerabilities this Patch Tuesday—nearly triple last month's record—including 60 critical bugs and three zero-days, two of which are being actively exploited. Microsoft attributes the surge to AI-assisted vulnerability discovery, a trend echoed by Adobe, Cisco, Mozilla, Oracle, and Google ramping up their own patch cadences. Researchers warn that Microsoft's human-centric "exploitability index" is falling behind, as AI tools like Anthropic's red team model have produced working exploits for 13 of 14 flaws rated "less likely" to be exploited.
HN Discussion:
  • AI-assisted bug hunting is a genuinely positive use of the technology
  • Skepticism that the AI narrative is real; may be backlog fixes rebranded as AI discoveries
  • ~Windows has an inherently massive attack surface making patches a never-ending cycle
  • Microsoft's bug reporting process is broken and ignores external researchers
  • Concerns about patch quality and whether fixes introduce new vulnerabilities or address chained exploits
17.Bonsai 27B: A 27B-Class model that runs on a phone(prismml.com)
687 points by xenova 1 day ago | 243 comments | permalink
tl;dr: PrismML released Bonsai 27B, a compressed version of Qwen3.6 27B using ternary (1.71 bits/weight, 5.9GB) or binary (1.125 bits/weight, 3.9GB) weights, enabling a 27B-class multimodal model to run on an iPhone 17 Pro. The variants retain 95% and 90% of full-precision benchmark performance respectively, with math and coding capabilities nearly untouched, and reach up to 163 tok/s on an RTX 5090. Weights are available under Apache 2.0 with native support for Apple MLX and NVIDIA CUDA.
HN Discussion:
  • Requests comparison with competing small models like Gemma 4 12B QAT to validate claims
  • Excited about ternary/binary quantization breakthrough enabling local model use
  • ~Questions clarity of comparisons and impact on tool calling performance
  • Shares hands-on benchmark/testing results revealing implementation issues
  • Skeptical of demo quality, noting factual errors in showcased output
18.Your 'app' could have been a webpage (so I fixed it for you)(danq.me)
869 points by MrVandemar 5 days ago | 530 comments | permalink
tl;dr: Frustrated that a travel agency required installing a 43MB Android app just to view an itinerary (which was really just HTML, images, and PDFs with added tracking and ads), the author reverse-engineered it by rooting an Android emulator and intercepting traffic with HTTP Toolkit. They discovered the app fetches JSON from an API using concatenated username-password credentials in the URL, then wrote a Ruby script to convert that JSON into a lightweight, password-protected webpage for their tour group—stripping ads and tracking while gaining bookmarking, printing, and accessibility benefits.
HN Discussion:
  • Companies push apps because they're more profitable, enable tracking, and bypass ad blockers
  • Users and businesses genuinely prefer apps for the home screen convenience and familiar install flow
  • The web platform's extensibility and browser extensions make it superior for customization
  • PWAs were the promised path but were undermined by app store economics and notification incentives
  • Developers themselves benefit from abandoning app stores due to bureaucracy and deployment friction
19.Australian energy retailers must offer three hours of free daytime electricity(lenergy.com.au)
286 points by i2oc 2 days ago | 402 comments | permalink
tl;dr: From July 2026, energy retailers in NSW, South Australia, and South-East Queensland must offer households at least three free hours of daytime electricity daily (around midday), capped at 24 kWh/day, to pass on the benefit of excess rooftop solar generation. Access requires only a smart meter and opting in — no solar panels or home ownership needed — making it available to renters and apartment dwellers. Estimated savings range from $100–$1,100/year depending on how much load you can shift into the free window.
HN Discussion:
  • Article is misleading; retailers only need to offer one such plan, not free power to all
  • ~Title needs clarification that free energy is limited to midday window
  • Personal experience confirms these plans work great with battery/solar setups
  • Australia's solar abundance makes cheap/free electricity a natural outcome benefiting everyone
  • Concern that new policy could worsen existing better plans by introducing caps