Published on July 20, 2026 at 06:00 CEST (UTC+2)
Show HN: I replaced a $120k bowling center system with $1,600 in ESP32s (1747 points by section33)
Show HN: I replaced a $120k bowling center system with $1,600 in ESP32s
An SRE and his family bought an abandoned 8-lane bowling center in a small town. The existing scoring system, installed in 2008, cost six figures and handles ball speed, pin detection via cameras, foul detection, and pinsetting. The author replaced it with a system built on ESP32 microcontrollers for about $1,600, demonstrating that modern low-cost hardware can replicate expensive proprietary systems. The story highlights the gap between legacy industrial pricing and modern embedded solutions.
Claude Code uses Bun written in Rust now (421 points by tosh)
Claude Code uses Bun written in Rust now
According to Simon Willison, Claude Code v2.1.181 (released June 17, 2026) uses the Rust port of Bun, replacing the original JavaScript runtime. Evidence includes a version number (Bun v1.4.0) higher than the official release, and strings revealing Rust source files in the binary. The transition was largely invisible to users, with only a 10% startup speed improvement on Linux. This reflects a trend of rewriting performance-critical tools in Rust for efficiency.
The Zen of Parallel Programming (130 points by edgar_ortega)
The Zen of Parallel Programming
This reflective essay draws parallels between parallel computing and human cognition. It argues that adding more processors (or mental faculties) doesn't automatically increase productivity; problems must be divided, coordinated, and synchronized. The author extends the metaphor to inner conflicts—where memories, emotions, and logic can become deadlocked. The piece suggests that effective coordination, whether in hardware or within the self, is the key to unlocking true potential.
What I learned selling 2,500 MIDI recorders: Hardware is not so hard (431 points by chipweinberger)
What I learned selling 2,500 MIDI recorders: Hardware is not so hard
Chip Weinberger launched Jamcorder, a piano recording device, and sold 2,500 units. Coming from a software background, he expected hardware to be the most difficult part, but found the process surprisingly smooth. He hand-assembled the first 500 units without any major issues, contradicting the common belief that hardware is inherently hard. The article emphasizes that with modern manufacturing and careful design, hardware startups can succeed with fewer obstacles than expected.
Minecraft: Java Edition now uses SDL3 (286 points by ObviouslyFlamer)
Minecraft: Java Edition now uses SDL3
The article announces that Minecraft: Java Edition has switched to SDL3 (Simple DirectMedia Layer 3) for its input handling and window management. This update improves cross-platform compatibility and performance, particularly for Linux and macOS users. The move reflects the ongoing modernization of game engines to leverage better abstraction layers and community-driven libraries.
A new Intel Itanium (IA-64) emulator that boots Windows (61 points by jandeboevrie)
A new Intel Itanium (IA-64) emulator that boots Windows
Yufeng Gao released version 0.1 of an Intel Itanium (IA-64) emulator that successfully boots Windows Server 2003 and Windows XP 64-bit. The emulator runs slowly (486-level performance on a Ryzen 5000) but represents a breakthrough in emulating the otherwise niche Itanium architecture. It does not yet support OpenVMS, HP-UX, or Linux, but the progress is significant for retro-computing and preservation efforts.
Claude Fable produced a counterexample to the Jacobian Conjecture (41 points by loubbrad)
Claude Fable produced a counterexample to the Jacobian Conjecture
A mathematician named levent (@alpoge) posted that Claude (an AI assistant nicknamed "Fable") found a counterexample to the Jacobian Conjecture—a long-standing open problem in algebraic geometry. The counterexample is a specific polynomial map from ℂ³ to ℂ³ with Jacobian determinant -2 that is not invertible but satisfies the condition of constant non-zero Jacobian. If verified, this would disprove the conjecture, marking a major AI-assisted mathematical discovery.
Orion Browser by Kagi (88 points by sebjones)
Orion Browser by Kagi
Kagi (the search engine company) introduces Orion, a WebKit-based browser focused on privacy and customization. It blocks zero-telemetry, supports Safari, Chrome, and Firefox extensions (curated list of 20 guaranteed), and integrates with Kagi services. With 4 million users, it targets users who want native performance without data collection. The browser is available for macOS, iOS, iPadOS, and Linux.
Biggest Probabilistic Computer Turns Noise into Answers (46 points by rbanffy)
Biggest Probabilistic Computer Turns Noise into Answers
IEEE Spectrum reports on the largest probabilistic computer to date, using 1 million p-bits (probabilistic bits). Unlike classical bits, p-bits fluctuate randomly and leverage noise to solve optimization problems efficiently. The computer is built from existing CMOS technology and is particularly suited for problems like factoring and combinatorial optimization. This marks a step toward alternative computing paradigms that exploit randomness rather than fight it.
Who Is America's Homer? (47 points by samclemens)
Who Is America's Homer?
This symposium in Plough magazine asks various authors and poets to nominate a poet who captures the American spirit, akin to Homer for Greece, Shakespeare for England, or Dante for Italy. Contributors discuss candidates like Walt Whitman, Emily Dickinson, and others. The piece explores the cultural need for a national epic poet and the challenges of defining a unified American identity through literature.
Rust adoption in AI/ML tooling is accelerating
Claude Code’s switch from Bun (JavaScript) to a Rust port of Bun, along with the prevalence of Rust in high-performance infrastructure, indicates a clear trend. Rust’s memory safety and performance make it attractive for AI/ML runtime environments, especially for latency-sensitive tasks like code generation. Expect more AI tools to rewrite components in Rust, improving speed and reliability without sacrificing developer experience.
AI-assisted mathematical discovery is becoming real
Claude Fable’s apparent counterexample to the Jacobian Conjecture—if confirmed—would be a landmark: an AI system solving a decades-old open problem. This demonstrates that large language models (LLMs) can not only manipulate symbolic mathematics but also generate novel, non-trivial insights. The implication is that AI will increasingly act as a collaborator in theoretical research, exploring combinatorial spaces too vast for humans.
Probabilistic computing as a niche AI accelerator
The 1-million p-bit probabilistic computer shows that noise-based computation is scaling up. While not a general-purpose replacement for GPUs, it excels at optimization problems (e.g., factoring, graph problems) that underpin some AI workloads. This trend suggests hybrid architectures where probabilistic co-processors handle specific tasks (e.g., sampling, Bayesian inference) more efficiently than deterministic digital logic.
Replacing expensive legacy systems with commodity hardware + AI
The bowling center example (ESP32 replacing a $120k system) mirrors a broader trend: AI and embedded systems can disrupt traditional industrial automation. With cheap microcontrollers and open-source ML for tasks like pin detection (object detection), small businesses can bypass costly vendor lock-in. This will accelerate democratization of automation in sectors like retail, manufacturing, and sports.
Hardware startups are becoming easier, aided by AI tools
The Jamcorder story challenges the “hardware is hard” mantra. With modern PCB design, 3D printing, and supply chain access, hardware development is more approachable. AI-assisted design tools (e.g., generative PCB layout, automated testing) further lower barriers. Expect an explosion of niche hardware products powered by AI, similar to the software startup boom.
Emulation and preservation of legacy architectures support AI training data diversity
The Itanium emulator that boots Windows demonstrates interest in preserving non-x86 systems. For AI/ML, emulation enables generation of synthetic training data for rare or exotic hardware (e.g., for security testing or binary analysis). It also fuels retrocomputing communities that create unique datasets, which could be used to train models on legacy code or system behaviors.
Privacy-first browsing is evolving alongside AI data collection
Kagi’s Orion browser positions itself as anti-telemetry and anti-AI-data-collection, directly countering the trend of browsers harvesting user data for model training. This highlights a growing tension: as AI requires vast amounts of web data, privacy-focused browsers and search engines (like Kagi) create a parallel ecosystem. Implications include the need for AI companies to respect opt-out signals and for users to have more control over their data used in training.
Analysis generated by deepseek-reasoner