Apple at 50 (apple.com)
Apple’s “50 Years of Thinking Different” page marks the company’s 50-year milestone, looking back at its history while emphasizing its ongoing focus on building tools and experiences that enrich people’s lives. The text is largely promotional, with no specific new product or policy details beyond celebrating Apple’s past and future.
Windows 95 defenses against installers that overwrite a file with an older one (devblogs.microsoft.com)
Raymond Chen explains how 16-bit Windows guidance expected installers to compare file versions and only overwrite system components with newer builds. Because many installers ignored that and overwrote newer Windows 95 files with older Windows 3.1 ones, Windows 95 added a repair mechanism: it stored backup copies of commonly overwritten files in SYSBCKUP and, after an installer ran, restored the higher-version copy. Chen notes that earlier approaches that simply blocked overwrites often caused installers to fail or fall back to reboot-and-retry tactics, so the post-install cleanup strategy worked best.
"Special 301" Comments on Nintendo Game Piracy in Asia and Latin America (1994) (archive.gamehistory.org)
The provided archive entry (“Special 301” Comments on Nintendo Game Piracy in Asia and Latin America, 1994) appears to document policy-related commentary from that period, but the fetched text contains no substantive details to summarize.
How-to guide: Commissioning a Sensor Physics R&D Lab (gist.github.com)
The article is a practical, experience-based checklist for commissioning a physics and electronics R&D lab for sensor development, emphasizing that suitable, purpose-built space and infrastructure must be secured before buying equipment. It outlines three lab zones (main lab, mechanical workshop, and PCB/PCBA area), lists required services such as power with UPS, managed networking, gas/vacuum, coolant, and controlled air quality, and covers key safety provisions. The guide then recommends core instrumentation categories (bench power supplies, DMMs/oscilloscopes, inspection tools, low-level measurement gear like lock-in amplifiers and SMUs, plus initial DAQ options) and stresses organization, labeling, and documentation.
ZomboCom stolen by a hacker, sold, now replaced with AI-generated makeover (old.reddit.com)
A site called ZomboCom was reportedly stolen by a hacker, listed for sale, and later replaced with an AI-generated redesign, according to discussion on the linked post.
Jax's true calling: Ray-Marching renderers on WebGL (benoit.paris)
The article shows how to build a simple ray-marching (sphere tracing) renderer in Python using JAX, running in the browser via WebGL. It explains using signed distance functions for geometry, vectorizing pixel computations with vmap, and leveraging JAX’s differentiability to derive surface normals for lighting. The author also notes ideas for future experiments such as WebGPU backends and related mathematical extensions.
SpaceX files to go public (nytimes.com)
SpaceX has filed paperwork to go public, according to the report, setting up a potential IPO that would further expand the company’s access to capital. The story links the move to Elon Musk’s role and discusses what the filing could mean for SpaceX’s future and investors.
Scientists crack a 20-year nuclear mystery behind the creation of gold (sciencedaily.com)
A team of nuclear physicists used experiments at CERN’s ISOLDE facility to clarify key steps in the r-process, the rapid neutron-capture pathway that forges heavy elements like gold. They report the first measurements of neutron energies tied to beta-delayed emission of two neutrons, the observation of a long-predicted single-particle neutron state in tin-133, and evidence that the population of the relevant nuclear state doesn’t follow expected statistical patterns. The findings are intended to improve models of the extreme stellar events that produce heavy elements and refine understanding of exotic, short-lived nuclei.
Show HN: Flight-Viz – 10K flights on a 3D globe in 3.5MB of Rust+WASM (flight-viz.com)
Flight-Viz is a Rust + WebAssembly project that renders about 10,000 flights on a 3D globe in real time while keeping the delivered code around 3.5MB.
IPv6 address, as a sentence you can remember (sentence2ipv6.tib3rius.com)
The article introduces “sentence2IPv6,” a tool that converts an IPv6 address into a memorable sentence and then decodes the sentence back into the original IPv6 address.
Show HN: Dull – Instagram Without Reels, YouTube Without Shorts (iOS) (getdull.app)
Dull is an iOS browser that loads major social apps (Instagram, YouTube, Facebook, X) while stripping out engagement features like Reels/Shorts, explore feeds, and ads. It adds on-device controls such as daily time limits, scheduled quiet hours, optional grayscale, usage tracking, and friction gates (e.g., a short challenge) before opening apps, without using accounts or sending data off the device. The app is offered with a 3-day free trial and then low-cost subscription or a lifetime option.
The revenge of the data scientist (hamel.dev) AI
The post argues that much of “LLM harnessing” and evaluation is still traditional data science, despite claims that the field is declining or that engineering teams can rely on APIs and generic tooling. It highlights common eval pitfalls—such as using generic metrics, unverified LLM judges, weak experimental design, low-quality data/labels, and over-automation—and explains how data scientists would approach each with trace analysis, error breakdowns, proper validation, and domain-expert labeling.
Set the Line Before It's Crossed (nomagicpill.substack.com)
The piece argues that people and groups should define clear “soft,” “firm,” and “hard” lines for unacceptable behavior and the specific actions that follow, rather than deciding at the moment something happens. It warns that failing to set lines leads to normalization of deviance—excusing violations, waiting for repeat offenses, or moving the threshold—and offers a step-by-step method for establishing criteria, consequences, and accountability. It also provides example line/action pairings across areas like government overreach, relationships, workplace conduct, and health.
Obfuscation is not security – AI can deobfuscate any minified JavaScript code (afterpack.dev) AI
The AfterPack blog argues the “Claude Code source leak” didn’t expose hidden code: Claude Code’s CLI JavaScript was already publicly accessible on npm, with only a source map accidentally revealing additional internal comments and file structure. It also contends the bundled code is minified rather than truly obfuscated, and that AI/AST parsing can extract large amounts of prompts, tool descriptions, and configuration strings directly from the minified bundle. Anthropic says the issue was a packaging mistake and not a security breach, noting similar source map exposure occurred before.
A new C++ back end for ocamlc (github.com)
The ocaml/ocaml pull request proposes a new C++ backend for ocamlc, replacing the older non-incremental C++ generation used by the runtime/FFI. It demonstrates compiling a purely functional OCaml prime-sieve program into readable C++ templates via ocamlc -incr-c, and discusses how C++ template depth and interpreter choice affect build/run behavior. The author notes that larger computations can require special flags and improved algorithms, and suggests the approach could potentially be extended to other languages in future work.
Fast and Gorgeous Erosion Filter (blog.runevision.com)
The post describes a fast, GPU-friendly shader technique for generating realistic-looking mountain erosion—branching gullies and ridges—without simulating water drops. It builds on earlier ShaderToy “eroded terrain noise” implementations, explaining the underlying octave-based stripe method, how it avoids chaotic artifacts by using per-cell stripe pivots and blending, and how the author refines peak/valley behavior with two slope-dependent strategies (a “frequency” approach and a “fade” approach). The article also provides implementation-focused details intended to complement a companion video.
Show HN: Git bayesect – Bayesian Git bisection for non-deterministic bugs (github.com) AI
Git bayesect is a Python tool that applies Bayesian inference to automate “git bisect” for flaky or non-deterministic failures, estimating which commit most likely introduced a change in failure likelihood. It uses a greedy entropy-minimization strategy and a Beta-Bernoulli approach to handle unknown failure rates, with commands to record pass/fail observations and select the most probable culprit commit. The README also includes examples and a demo that simulates a test whose failure probability shifts over a repo’s history.
Signing data structures the wrong way (blog.foks.pub)
The article argues that common serialization/signing approaches can suffer from “domain separation” failures, where two different message types serialize identically and a signature for one can be misapplied to another. It proposes “domain separators in the IDL” as part of the Snowpack/FOKS approach: random, immutable type tags are embedded into the schema and are included in the bytes that are signed, encrypted, or MAC’d, while the tags themselves are not serialized. The post also describes how Snowpack aims to produce canonical encodings for safer verification and forwards/backwards-compatible upgrades.
Artemis II lifts off: four astronauts begin 10-day lunar mission (theguardian.com)
NASA’s Artemis II crewed rocket lifted off from Cape Canaveral with four astronauts aboard for a 10-day mission that will fly past the moon but not land. After a brief hold for issues related to launch systems and a battery sensor reading, Orion reached Earth orbit and mission controllers began assessing early flight data. The cislunar test flight is designed to evaluate the spacecraft for future Artemis missions, including carrying the first woman and first person of color to travel beyond Earth orbit.
Swappa.com for GrapheneOS compatible devices – Stay Away (discuss.grapheneos.org)
A GrapheneOS user warns that Swappa listings marked “unlocked” did not allow OEM/bootloader unlocking on three purchased Pixel phones. They say Swappa staff stated the “unlocked” category requires factory-unlocked Google Edition devices, but multiple sellers still delivered phones that were not bootloader unlockable. The poster argues the marketplace cannot be trusted for GrapheneOS compatibility because moderation of listings appears inconsistent, advising others to seek different sources.
DRAM pricing is killing the hobbyist SBC market (jeffgeerling.com)
Jeff Geerling argues that sustained high DRAM (especially LPDDR) prices are pushing hobbyist single-board computer (SBC) options out of reach, citing Raspberry Pi’s recent memory-related price increases for 4GB+ models. He says DRAM now makes up much of the board cost, reducing new product releases and steering hobbyists toward older SBCs or microcontrollers, with smaller vendors at greater risk.
Safe ways to do things in bash (2023) (github.com)
The article explains Shellharden’s methodology for writing bash “safely,” arguing that correct bash scripting is achievable through a finite set of rewrites. It focuses first on the core rule to always quote variable expansions and command substitutions to prevent word splitting and pathname globbing. It also covers related style/syntax guidance such as preferring $(cmd) over backticks, using braces for certain substitution cases (like numbered positional parameters), and notes that scripts often require more than simply removing vulnerabilities—they may need to be rewritten differently.
Butterfly-collecting: The history of an insult (2017) (lughat.blogspot.com)
The article traces how the metaphor “butterfly-collecting” (and earlier “stamp-collecting”) was used to disparage descriptive, data-gathering work in linguistics and other fields. It reviews the quote’s shifting attributions—from an alleged Rutherford remark, to anthropology and rhetoric—and shows how it later became associated with Chomskyan critiques of descriptive approaches. The author argues that the insult’s spread reflects broader 20th-century disciplinary tensions about what counts as “real” science.
Prompt Engineering for Humans (michaelheap.com) AI
The article argues that “prompt engineering” is essentially the same as good management: providing clear context, constraints, success criteria, and validation so people (and AI) don’t have to guess. Using an example with an agent building a Trello CLI feature, the author shows that vague instructions produced a technically correct but incomplete result, while more specific context led to an immediately usable command. The piece concludes that at scale, ambiguity is costly and managers must design requirements carefully rather than simply assign tasks.
We intercepted the White House app's network traffic (atomic.computer)
Atomic Computer reports a live network capture (MITM) of the official White House iOS app (v47.0.4) and finds that most requests go to third parties rather than whitehouse.gov. The app sends device and session metadata—including language, timezone, IP address, and a persistent OneSignal identifier—to OneSignal via decrypted HTTPS traffic, and it loads multiple Elfsight-controlled domains using a two-stage script loader to inject widget code. The researchers also observe Google DoubleClick ad tracking within YouTube embeds and note that these behaviors were not disclosed in the app’s privacy manifest. The post describes logging activity across multiple tabs without modifying traffic and redacts personal identifiers in the published examples.