HackerNews Digest

NASA’s Artemis II crewed test flight launched from Kennedy Space Center’s Pad 39B at 6:35 p.m. EDT on a Space Launch System (SLS) rocket. The vehicle’s twin solid rocket boosters, providing ~75 % of lift‑off thrust, separated at +02:09 MET, followed by core‑stage main‑engine cutoff and separation at +08:02 MET and +08:14 MET respectively. The interim cryogenic propulsion stage (ICPS) ignited, and the Orion spacecraft “Integrity” deployed its four solar‑array wings (SAWs) about 18 minutes after launch, each wing containing 15,000 cells and spanning a total 63 ft wingspan. Subsequent perigee‑raise (PRM) and apogee‑raise (ARB) burns will lift the orbit for deep‑space operations. The mission crew—NASA astronauts Reid Wiseman, Victor Glover, Christina Koch, and CSA astronaut Jeremy Hansen—will conduct a proximity‑operations demonstration with the ICPS, test life‑support systems, and complete a ~10‑day lunar flyby, marking the first crewed deep‑space flight in over 50 years and a stepping stone toward future Moon and Mars missions.

Two recent quantum‑computing announcements significantly lower resource estimates for cryptographic attacks. A Caltech team, including John Preskill, presented fault‑tolerant protocols using high‑rate quantum error‑correcting codes that reduce overhead, especially for architectures supporting non‑local gates such as neutral‑atom arrays or trapped ions. Separately, Google disclosed a more compact implementation of Shor’s algorithm capable of breaking 256‑bit elliptic‑curve cryptography; the result was released via a cryptographic zero‑knowledge proof that confirms the circuit’s existence without revealing its structure. Combined, these advances shrink the projected hardware needed to compromise Bitcoin signatures from millions of physical qubits to roughly 25 000 qubits, compressing the timeline for viable quantum attacks. The author emphasizes that, while fundamental quantum principles remain unchanged, the new quantitative thresholds strengthen the case for immediate migration to quantum‑resistant cryptographic schemes.

The pull request titled “C++ support by stedolan” (PR #14701) is opened in the ocaml/ocaml repository on GitHub. The conversation displayed for the PR contains a single message stating “You can’t perform that action at this time.” No additional discussion, code changes, or review comments are provided in the extracted text. The page includes several images whose alt‑text labels list GitHub usernames: @stedolan, @redianthus, @dra27, @avsm, @sergezloto, @ducminhgd, and @AdelKS. No further technical details, change descriptions, or contextual information about the C++ support implementation are present in the scraped content.

Raspberry Pi announced further price hikes for all models with LPDDR4 RAM, including a 3 GB Pi 4 at US $83.75 and a 16 GB Pi 5 at US $299.99. The increases reflect a surge in DRAM costs, which now represent the majority of component expenses for most hobbyist single‑board computers (SBCs). Consequently, boards with more than 4 GB RAM have become unaffordable for many hobbyists, and new SBC releases have sharply declined, with Radxa being the only vendor maintaining a recent product cadence. Comparable mini‑PCs now exceed US $250 for 8 GB configurations, and used PCs with similar memory are also pricier. Hobbyist developers are reverting to older SBCs or microcontrollers to keep project costs below US $100. While Raspberry Pi’s leadership expects DRAM prices to eventually fall, the timing is uncertain, and smaller SBC manufacturers may struggle to survive if current pricing trends persist.

EmDash is an open‑source, MIT‑licensed CMS written in TypeScript and built on the Astro framework. It runs serverlessly on platforms such as Cloudflare Workers, scaling to zero when idle and billing only for CPU time. Plugins execute in isolated Dynamic Workers and must declare required capabilities (e.g., read:content, email:send) in a manifest, providing a strict, OAuth‑like permission model that eliminates the 96 % of WordPress‑related security issues caused by unrestricted PHP plugins. Because plugins are sandboxed and license‑agnostic, developers can choose any license and distribute code without a central marketplace, reducing lock‑in. EmDash includes native x402 support for pay‑per‑use content charging, passkey‑based authentication with role‑based access control, and AI‑native tooling (MCP server, CLI, and Agent Skills) for automated migration and site management. A WordPress exporter allows rapid import of posts, media, and custom types. The v0.1.0 preview is available on GitHub, deployable to Cloudflare or any Node.js server, with an online Playground for testing.

The post describes a GPU‑friendly erosion filter implemented as a directional noise function that can be applied as a height‑map overlay. The technique builds on Clay John’s 2018 “Eroded Terrain Noise” shader and Felix Westin’s 2023 refinement, adding intuitive parameters and two alternative ways to handle flat areas. Basic operation: compute the gradient of an underlying height field, generate stripe patterns aligned with the gradient in a cell‑based Worley layout, and blend neighboring cells so the stripes form gullies and ridges. Two methods preserve peaks and valleys: the **frequency approach** scales stripe frequency with slope (thickening stripes on flat terrain) and avoids peak bulging but creates valley bulges; the **fade approach** keeps stripe width constant and fades stripe amplitude toward a value derived from altitude, allowing crisp V‑shaped valleys when combined with a custom easing curve. Further enhancements for sharp, branching features include **stacked fading** (masking each octave with previous‑octave creases), **normalized gullies** (unit‑circle normalization of interpolated sine/cosine pairs with thresholded scaling to prevent loops), and **straight gullies** (restricting small‑scale gullies from running parallel to larger ones). The final filter yields fast, chunkable erosion with controllable detail, and the underlying directional noise (Phacelle Noise) is released separately.

None

Git _Bayesect_ is a Python tool that applies Bayesian inference to git bisect, identifying the commit where a change in event likelihood (e.g., increased test flakiness) occurs. It selects commits by greedily minimizing expected entropy and uses a Beta‑Bernoulli conjugacy to handle unknown failure rates. Users start a bisection with `git bayesect start --old ` (optionally specifying `--new`), record observations via `git bayesect pass`/`fail` (or on a specific commit), and check status with `git bayesect status`. Priors can be set per commit (`git bayesect prior --commit --weight `), derived from filenames (`priors_from_filenames`) or commit text (`priors_from_text`) using user‑provided callbacks. The tool logs commands for reproducibility, supports undoing observations, and can run automatically with `git bayesect run `. A demo repository (`scripts/generate_fake_repo.py`) creates a fake history with a flaky script, allowing users to practice locating the commit that altered the failure probability. Installation is via `uv tool install git_bayesect`.

Meta is releasing BOxCrete, a Bayesian‑optimization AI model for concrete mix design, alongside the foundational dataset used for a high‑performance mix in its Rosemount, MN data center. The open‑source model (MIT license) predicts strength, slump, and other key properties, enabling rapid exploration of formulations that incorporate domestically produced cement, which accounts for ~20‑25 % of U.S. cement consumption. Partnerships include Amrize (cement producer), the University of Illinois at Urbana‑Champaign, Mortenson, and SaaS platform Quadrel, which integrates Meta’s framework into daily mix‑design workflows. Field trials in Illinois, Minnesota, and Pennsylvania show AI‑optimized mixes achieving 43 % faster strength gain and ~10 % reduced cracking risk while using U.S. materials. Meta’s Adaptive Experimentation platform iteratively learns from historical mixes and lab results, proposes high‑potential candidates under user‑specified constraints, and refines predictions with each test, accelerating discovery without altering standard validation procedures. The initiative aims to increase reshoring of cement, lower emissions, and improve supply‑chain resilience across the American concrete industry.

The post explains that cryptographic operations (signing, encryption, MAC, hashing) require data encodings that are both canonical and domain‑separated to prevent type‑confusion attacks, where distinct message types serialize to identical byte streams (e.g., TreeRoot vs. KeyRevoke). Existing ad‑hoc mitigations (context strings, method‑name hashes) are error‑prone. Snowpack, the serialization format used by FOKS, embeds a random, immutable 64‑bit domain separator directly in the IDL definition of each struct. Compilers generate a `GetUniqueTypeID()` method; signing and verification concatenate this identifier with the serialized object, while the identifier itself is omitted from the transmitted bytes. The type system enforces that only objects with a domain separator can be processed, eliminating mismatched‑type signatures. Snowpack also provides canonical encodings via positional JSON‑like arrays and Msgpack, enforcing minimal integer sizes and avoiding dictionary key ordering issues. It supports forward/backward compatibility through nullable fields and variant handling, and tooling ensures uniqueness of separators within a project. The approach is open‑sourced for Go and TypeScript, with plans for additional languages.