HackerNews Digest

OpenCode is an open‑source AI coding agent that has attracted a substantial community on GitHub, accumulating more than 120 000 stars, around 800 contributors, and over 10 000 commits. The project is actively maintained and sees widespread adoption, with over 5 million developers reporting monthly use. Its popularity reflects broad trust in the tool’s capabilities for automating or assisting software development tasks, supported by a large, collaborative contributor base and a significant volume of ongoing code contributions.

The openui‑lang parser was originally written in Rust, compiled to WebAssembly, and called from JavaScript. Each call incurred a fixed overhead: copying the input string into WASM memory, parsing in Rust, serialising the result with `serde_json::to_string`, copying the JSON back, and deserialising it with `JSON.parse`. This boundary cost dominated latency, not the Rust parsing itself. Replacing JSON with `serde‑wasm‑bindgen` (direct `JsValue` return) was ~30 % slower because constructing a JS object requires many fine‑grained conversions across the runtime boundary. The team rewrote the six‑stage pipeline (autocloser, lexer, splitter, parser, resolver, mapper) in TypeScript, eliminating the WASM boundary. Single‑call benchmarks showed 2.2–4.6 × faster parses and 2.6–3.3 × lower total streaming cost. Further gains came from an incremental streaming parser that caches completed statement ASTs, reducing cumulative work from O(N²) to O(N). Key takeaways: WASM excels for compute‑heavy, low‑interop tasks (e.g., image/video processing, cryptography). For frequent small‑input parsing where data must cross the boundary, JS implementations with algorithmic improvements are more efficient. Profiling actual bottlenecks before choosing a language is essential.

Ghostling is a minimal terminal emulator demo built on the libghostty C API, implemented in a single C file and using Raylib for windowing and 2D rendering. It showcases libghostty‑vt, a zero‑dependency library that parses VT sequences, manages terminal state (cursor, styles, reflow, scrollback), and provides renderer state without any drawing code; the host supplies rendering. Features inherited from Ghostty include resize with text reflow, 24‑bit and 256‑color support, bold/italic/inverse styles, Unicode grapheme handling, keyboard input with modifiers, Kitty keyboard protocol, mouse tracking/reporting (X10, normal, button, any‑event, SGR, URxvt, UTF8), scroll‑wheel handling, draggable scrollbar, and focus reporting. Planned exposures are Kitty Graphics Protocol, OSC clipboard, and OSC title. libghostty does not provide GUI elements such as tabs, splits, session management, or configuration UI; these must be implemented by consumers. Build requires CMake ≥ 3.19, a C compiler, and Zig 0.15; Raylib is fetched via CMake. Debug builds are slow; release builds use `-DCMAKE_BUILD_TYPE=Release`. libghostty‑vt’s C API is language‑agnostic, allowing bindings for Rust or other languages, and it is renderer‑agnostic, supporting alternatives to Raylib, including WASM.

None

The GitHub repository “arnoldrobbins/LinuxByExample-2e” contains the source code accompanying *Linux Application Development By Example – The Fundamental APIs* by Arnold Robbins. The work is copyrighted 2004‑2026 by Pearson Education (ISBN‑13 978‑0‑13‑532552‑0; ISBN‑10 0‑13‑532552‑8). The repository’s **Documents** folder stores reference material, including the author’s code license and an **errata.txt** file that is updated with any discovered errors. Subfolders correspond to each chapter and hold the example programs discussed in the book. Users can report book issues via GitHub issues, and the repository’s latest commit was on 10 Oct 2025 17:07 IDT. An image with alt text “@arnoldrobbins” is also present.

The article lists Japanese chop‑stick etiquette violations (kiraibashi) in alphabetical order, describing each prohibited action. Highlights include: Agebashi – raising chopsticks above the mouth; Araibashi – cleaning them in soup or drinks; Awasebashi (or hashiwatashi) – passing food between chopsticks, a serious taboo linked to funeral rites; Ukebashi – holding out a bowl while still gripping chopsticks; Utsuribashi – repeatedly using the same side‑dish chopsticks without alternating with rice; Urabashi – picking up food then returning it; Ogamibashi – holding chopsticks with both hands while saying *itadakimasu*; Oshikomibashi – pushing food deep into the mouth; Kaeshibashi – reversing chopsticks when serving to avoid contaminating food; Tatebashi – standing chopsticks upright in rice, a grave funeral symbol; Chigaibashi – using mismatched‑material chopsticks; Hashibashi – laying chopsticks across a dish as a “finished” signal (instead of a rest). Additional entries detail actions such as dropping, licking, stabbing, pointing, stirring, tapping, shaking, or using chopsticks as a toothpick, each considered impolite or taboo in Japanese dining culture.

The author expands a previous post on coinciding religious dates by adding Ash Wednesday and Chinese New Year, using calendar algorithms from Reingold & Dershowitz’s *Calendrical Calculations*. The original Common Lisp source (calendar.l) fails to load in SBCL/CLISP due to package/export declarations; removing the duplicated `(in-package "CC4")` and export list allows the file to load as `calendar.lisp`. A CLISP script (`ramadan‑lent`) loads this library, defines weekday and month name vectors, a `gregorian-date-string` formatter, and iterates 500 Islamic years (±250 years from the current year). For each year it computes Ramadan 1 (fixed‑date) and the corresponding Ash Wednesday (`easter(gyear) - 46`). When the dates match it prints the Gregorian date. Output shows nine matches between 1799 and 2253, with typical intervals of 98 years and occasional 31‑ or 67‑year gaps. A second script adds a check for Chinese New Year using `chinese-new-year-on-or-before`; matches occur in 1799, 2026, 2124, and 2222, confirming a triple coincidence after a 227‑year gap and predicting the next after 98 years. The author plans further exploration of the calendrical library.

Attention Residuals (AttnRes) replace standard Transformer residuals with depth‑wise softmax attention, allowing each layer l to compute its hidden state as hₗ = ∑ᵢ₌₀^{l‑1} α_{i→l} vᵢ, where α are derived from a learned pseudo‑query wₗ ∈ ℝᵈ. This mitigates PreNorm dilution, keeping output magnitudes bounded and distributing gradient norms more uniformly across layers. - **Full AttnRes**: attends over all previous layer outputs; memory cost O(L d). - **Block AttnRes**: partitions L layers into N blocks, uses standard residuals within blocks and inter‑block attention over block representations, reducing memory to O(N d). With ≈8 blocks it recovers most Full AttnRes gains with marginal overhead. Empirical results (Kimi Linear 48B/3B, 1.4 T tokens) show consistent loss reductions; Block AttnRes matches a baseline trained with 1.25× more compute. Downstream gains include +7.5 points on GPQA‑Diamond (multi‑step reasoning) and +3.1 on HumanEval (code generation). The method is a drop‑in replacement for residual connections, requiring only a pseudo‑query projection and RMSNorm per layer.

The term “molly guard” originates from early computing hardware and refers to a small plastic safety cover that must be removed before operating a significant button, preventing accidental activation. The name allegedly derives from an incident where an engineer’s daughter, Molly, pressed a critical red button twice during a datacenter visit. Physical examples appear in aircraft combat movies, recessed buttons, protective ridges around keys, and even SIM‑card ejection holes. Software analogues include confirmation dialogs and required modifier keys (e.g., Ctrl+Alt+Del) that act as guards. A contrasting concept, “reverse molly guards,” describes buttons that activate automatically after a period of inactivity, viewed as considerate design. The author notes the frustration for programmers when a system remains idle overnight, awaiting a non‑essential response.

The TransAvia AirTruk originated from Luigi Pellarini’s unconventional agricultural designs. In the mid‑1950s New Zealand needed a low‑cost top‑dressing aircraft, prompting the PL‑7 (a steel‑tank biplane with twin tail booms, tricycle gear, and a 400 hp Armstrong Siddeley Cheetah radial). After a 1958 hangar fire, Pellarini redesigned it as the PL‑11 (monoplane, 600 hp Pratt & Whitney radial, cockpit atop the tank, twin booms) using parts from surplus North American Harvards; first flight occurred in 1960 but bureaucratic delays halted production. The subsequent PL‑12, renamed “AirTruk,” was a sesquiplane with a 285 hp Continental flat‑six, retaining the twin‑boom layout and allowing a small cabin for crew. Formed in 1965, TransAvia produced 120 AirTruks (1967‑1988), employed for fertilizer spreading, occasional air‑ambulance or forward‑air‑control proposals, and a brief appearance in *Mad Max Beyond Thunderdome*. Competition from U.S. models and lack of government support ended production. Today only a few examples remain operational, chiefly as airshow aircraft and hobby flyers in Australia and the United States.