HackerNews Digest

July 20, 2026

Show HN: I replaced a $120k bowling center system with $1,600 in ESP32s

None
Read full article →
The comments express strong enthusiasm for modernizing legacy bowling equipment with open‑source hardware and software, highlighting the technical appeal, nostalgic value, and community‑building potential of such retrofits. Many contributors share related experiences, suggest documentation, videos, and broader sharing of schematics, while also noting practical concerns about reliability, maintenance workload, and cost‑effectiveness for typical alley owners. Overall, the discussion favors open‑source solutions as a way to revitalize underused recreational spaces, though it urges careful engineering and realistic business planning.
Read all comments →

Claude Fable produced a counterexample to the Jacobian Conjecture

A polynomial map \(F:\mathbb{C}^3\to\mathbb{C}^3\) is presented: \[ F(x,y,z)=\big((1+xy)^3z+y^2(1+xy)(4+3xy),\; y+3x(1+xy)^2z+3xy^2(4+3xy),\; 2x-3x^2y-x^3z\big). \] The Jacobian determinant of \(F\) is constant \(-2\). The map sends the three distinct points \((0,0,-\tfrac14)\), \((1,-\tfrac32,\tfrac{13}2)\), and \((-1,\tfrac32,\tfrac{13}2)\) to the identical image \((-\,\tfrac14,0,0)\). This provides a counterexample to the Jacobian conjecture in dimension three, showing that a polynomial map with non‑zero constant Jacobian determinant need not be invertible. The author attributes the insight to collaborators but the mathematical content is limited to the explicit map, its constant Jacobian, and the three pre‑images of a single point.
Read full article →
The comments reflect a mix of skepticism and cautious interest regarding the AI‑generated counterexample to the Jacobian conjecture. Many note the conjecture’s long history of failed proofs and question whether the result represents genuine insight or a simple, high‑school‑level construction. Some appreciate the provided links and recent surveys on AI‑driven mathematical discoveries, while others dismiss the achievement as trivial or premature, expressing a desire for more substantive AI contributions before granting credibility. Overall, the thread balances doubt about the claim’s significance with acknowledgement of ongoing AI research in mathematics.
Read all comments →

Claude Code uses Bun written in Rust now

Claude Code v2.1.181 (released June 17) and later versions use a Rust‑ported version of Bun, yielding roughly a 10 % startup speed improvement on Linux. Inspection of the installed binary with `strings` shows “Bun v1.4.0 (macOS arm64)”, a version newer than the latest public GitHub release (v1.3.14). The Rust canary build can be installed via `bun upgrade --canary`. Further analysis with `strings` extracts 563 Rust source file paths (e.g., `src/runtime/bake/dev_server/mod.rs`, `src/bundler/bundle_v2.rs`), confirming Rust code inclusion. A workaround using `BUN_OPTIONS="--preload=/tmp/bun-version.ts"` displays the embedded Bun version (`1.4.0`). The version bump to 1.4.0 appears in a May 17 commit to `package.json`, but it remains untagged outside the canary channel. This indicates Claude Code’s production deployment runs the Rust‑based Bun across many devices.
Read full article →
The comments show a mixed reaction to the AI‑driven rewrite of Bun from Zig to Rust. Critics emphasize poor communication, trust erosion, lingering bugs, performance and stability concerns, and worries about open‑source governance and supply‑chain security. Supporters point to modest speed gains, the value of deterministic compiler checks, and the demonstration that large‑scale code translation can now be achieved quickly, suggesting broader potential for AI‑assisted development. Overall, opinion splits between skepticism about quality and safety and optimism about the technology’s capabilities.
Read all comments →

What I learned selling 2,500 MIDI recorders: Hardware is not so hard

The author launched Jamcorder, a simple, single‑screw, single‑PCB MIDI recorder, selling 2,500 units. Contrary to expectations, hardware development proved smoother than software; roughly 200 K lines of firmware, app, and tooling required three years, while the hardware faced no major production or component‑sourcing crises (aside from a brief tariff concern). Simplicity was deliberate: low‑battery detection, ambient‑light sensing, power button, and USB‑C were omitted, and the injection‑molded enclosure used generous draft without slides. The author emphasizes that hardware difficulty scales with complexity and volume, noting that a ten‑fold increase or entry into highly competitive markets would be far harder. Practical advice for medium‑scale hardware launches includes: keep the BOM simple and avoid single‑source parts; minimize assembly and calibration; use Chinese assembly partners (Alibaba); target ≥70 % gross margin; maintain a lean organization; implement anti‑counterfeit measures; conduct in‑house final QA and hold inventory locally; request samples before each run; document manufacturing steps with photos; and use compact, value‑dense packaging.
Read full article →
Comments overall praise the project’s simplicity, reliability and user experience, while emphasizing that hardware difficulty varies with product complexity, supply‑chain logistics, cash‑flow constraints, certification and anti‑counterfeit measures. Many note that keeping designs minimal reduces risk and cost, yet acknowledge that scaling, component sourcing, tooling and regulatory testing become significant hurdles for larger runs. Technical critiques focus on component choices and durability, and numerous questions seek deeper insight into manufacturing, marketing, and protection strategies, reflecting a balanced view of both success and remaining challenges.
Read all comments →

The Zen of Parallel Programming

The passage draws an analogy between parallel programming and human cognition. It notes that modern computational power enables large‑scale tasks such as genome decoding, medical imaging, climate modeling, protein folding, drug discovery, energy research, and big‑data analysis. However, merely adding processors does not increase useful output; a problem must be partitioned, and its parts must communicate, synchronize, and share work to avoid idle or overloaded processors and resource contention. The author parallels this with the human mind, arguing that intelligence, emotion, memory, and creativity become ineffective when they fail to coordinate, leading to anxiety, exhaustion, and burnout. Zen concepts are invoked: “wholehearted activity” likened to a fire that fully consumes experience, and honest communication described as synchronization that aligns thoughts, emotions, and actions. The core claim is that the primary limitation—both in computing and in personal life—is not insufficient power but the division of existing power into uncoordinated components.
Read full article →
The comment reflects agreement with Brooks’s observation that communication overhead grows faster than linearly and that simply adding personnel yields diminishing returns. It expresses concern that the team is consistently overextended and that expanding headcount would only partially alleviate the issue. Emphasis is placed on the importance of synchronization, honesty, and mutual openness, acknowledging personal contribution to communication gaps. Additional analogies invoke divide‑and‑conquer principles and a classical notion of coherence, linking effective coordination to both technical and philosophical concepts.
Read all comments →

Minecraft: Java Edition now uses SDL3

Minecraft 26.3 Snapshot 4 introduces several core updates: - **Platform integration:** Window management, input, and platform handling have moved from GLFW to SDL 3. Keyboard input now uses SDL scancodes and keycodes; raw mouse input is removed, and relative mouse mode is always active. - **UI adjustments:** Borderless fullscreen becomes the default, can toggle without restart, and exclusive fullscreen is dropped on macOS. Minimum window size is 320×240, and Wayland is preferred on Linux. Debug overlay gains a separate GUI scale, player‑speed readout, and refresh‑rate display. - **Gameplay additions:** Spectator mode can interact with portals. New item components include `minecraft:cooking_fuel`, `minecraft:brewing_fuel`, `minecraft:sign_text_front/back`, `minecraft:waxed`, `minecraft:cushion/color`, `minecraft:villager_food`, and `minecraft:mob_visibility`. Armadillos no longer roll up in liquids. - **Technical changes:** Data pack version → 111.0, resource pack version → 92.0. Loot‑table fields now accept registry elements, tags, and inline values; numerous advancement, trigger, and slot‑source schemas were tightened. Signs no longer auto‑execute click events; a boolean `allow_op_features` restores previous behavior. - **World‑gen & environment:** Added `minecraft:gameplay/natural_mob_spawns` and `minecraft:gameplay/creature_world_gen_spawn_probability` attributes. Known issues include crashes in exclusive fullscreen on Windows (multi‑monitor) and Wayland.
Read full article →
The comments show overall interest in Minecraft’s move from GLFW to SDL3, with many noting SDL3’s improved performance, latency reductions, and modern GPU API support compared to the aging GLFW/SDL2 stack. Users share personal experiences converting games, report occasional bugs such as exclusive‑fullscreen crashes on Windows and Wayland, and request documentation on the rationale behind the switch. Additional discussion includes advice for setting up a family server, curiosity about Python integration, and observations that the change may aid modding without fundamentally altering cross‑platform capabilities.
Read all comments →

Orion Browser by Kagi

Orion Browser, developed by Kagi, operates on a user‑funded model with no advertising or third‑party agreements. Revenue is generated through optional subscription plans or a one‑time lifetime license. The browser is available across major operating systems—including macOS (Apple), Linux, Windows, iOS, and iPadOS—and can be downloaded from the App Store. Visual assets referenced include the Orion logo, platform icons, and promotional graphics for various devices and media outlets (e.g., ZDNet, Mac Observer, OMG UBUNTU). The site also lists social and community icons (Discord, Mastodon, YouTube, X) and mentions additional features such as lock, gear, puzzle piece, and arrow icons, implying security, settings, extensions, and navigation elements. The overall presentation emphasizes self‑sustainability, cross‑platform availability, and a focus on user‑owned browsing experience.
Read full article →
Comments show a mixed but generally appreciative view of Orion. Users value its built‑in ad‑blocking, vertical tabs, ability to run Firefox extensions on iOS, zero‑telemetry stance and smooth integration with Apple Pay, noting it feels less problematic than mainstream browsers. Frequent criticisms target the closed‑source nature, absent “Search for…” context menu, broken settings pages, performance stalls on older iOS devices, UI bugs, limited site compatibility, missing native features such as Apple Private Relay, 2FA autofill, and a built‑in password manager. Several request a Windows build and source‑code release, while a few remain willing to pay if issues are resolved.
Read all comments →

A new Intel Itanium (IA-64) emulator that boots Windows

A version 0.1 of an Intel Itanium (IA‑64) emulator has been released by Yufeng Gao with assistance from gdwnldsKSC. The emulator can boot the Itanium editions of Windows Server 2003 and Windows XP 64‑bit, though performance is very low (approximately 486‑class speed on a Ryzen 5000 series). At present it does not start OpenVMS, HP‑UX, Linux, or BSD guests. The code is currently closed‑source, but the developer indicated it will be published on GitHub under TheBrokenPipe/ski after cleanup. An update notes a QEMU fork that reportedly runs Windows Server 2008 on IA‑64 and is claimed to be generated with LLM assistance. No commercial IA‑64 emulators are known. The release is primarily of interest to preservation and hobbyist communities seeking to explore legacy Itanium software, especially Windows, while broader OS support remains unavailable.
Read full article →
The passage reflects a nostalgic and humorous tone about using an outdated IA‑64 Windows box for testing WebSphere, noting the need for workarounds and the limited performance of the platform. It mentions the rarity and quirks of Itanium hardware, including power consumption, noise, and slow emulation, while expressing mild criticism of the architecture’s market relevance and its strategic role in competitive industry dynamics. The overall view mixes appreciation for past experiences with skepticism toward Itanium’s legacy.
Read all comments →

Moonshine: Lets you stream games from your PC to any device running Moonlight

Moonshine is a Rust‑based, head‑less streaming server enabling Moonlight clients to stream games from a Linux PC. Each game runs in an isolated compositor, allowing the host to remain usable and eliminating the need for a physical monitor or HDMI dummy plug. It leverages GPU‑accelerated video encoding via Vulkan, supporting H.264, H.265, and experimental AV1 (the latter requires NVIDIA driver ≥ 595.44.3.0). Features include true 10‑bit HDR, full mouse/keyboard/gamepad input (including motion, touchpad, haptics), and low‑latency Opus‑encoded stereo or 5.1/7.1 audio. Moonshine requires systemd, a Vulkan‑capable GPU (NVIDIA RTX, AMD RDNA2+, Intel Arc), and Moonlight v6.0.0+. Installation is typically through the AUR; building needs clang, cmake, rust, Vulkan headers, Wayland, and related libraries. Configuration is via a TOML file where applications and optional pre/post commands are defined; scanners auto‑detect Steam games and desktop .desktop entries. Moonshine is Linux‑only, runs without an active desktop session, and is not intended for direct internet exposure—VPNs are recommended for remote access. Compared to Sunshine, Moonshine offers fewer platforms and features but provides isolated sessions on headless Linux hosts.
Read full article →
Comments show curiosity about Moonshine’s isolated streaming sessions, with several users asking how the separate environment functions and how it compares to existing solutions like Sunshine, Apollo, and Artemis. There is general interest in trying the newer Rust‑based server and appreciation for its headless‑machine friendliness, but concerns are raised about its modern GPU requirements and lack of support on certain Linux distributions. Some participants note past difficulties with FOSS streaming tools’ usability and express uncertainty about the practical differences between Moonshine and other forks.
Read all comments →

Biggest Probabilistic Computer Turns Noise into Answers

The article reports the creation of the largest probabilistic computer to date, featuring one million probabilistic bits (p‑bits) implemented on 18 field‑programmable gate arrays (FPGAs). P‑bits differ from classical bits (fixed 0/1) and qubits (superposition) by stochastically flipping between 0 and 1 with a tunable probability; when many operate in a correlated manner they can solve stochastic optimization problems (e.g., shortest‑path, QUBO, Ising). The new system performs over a trillion p‑bit flips per second and operates without global lockstep synchronization. Researchers identified a predictable design rule governing the minimum inter‑chip data‑exchange rate required for the array to function as a single machine; below this threshold speed‑accuracy trade‑offs emerge. The findings suggest that arbitrarily large probabilistic computers can be built by networking many chips, regardless of hardware platform. Future work will explore specialized, energy‑efficient p‑bit hardware such as magnetic tunnel junctions and MRAM‑based stochastic memory integrated with CMOS.
Read full article →
The discussion centers on digital probabilistic computing, noting a cloud‑based processor that claims substantial speed gains for certain workloads and describing p‑bits as toggling between binary states with adjustable probabilities. The commentator likens this behavior to an ADC lacking a voltage input and questions whether it constitutes a new analog computing approach, while also speculating that the architecture might excel at brute‑force password or cryptographic key cracking. Additional curiosity is expressed about the simulation environments used by mathematical modelers, alongside light‑hearted remarks.
Read all comments →