🛠️ Hacker News Tools

[Other] Serverless backend hosting without idle costs – open-source

[Database] Show HN: Pinecone Explorer – Desktop GUI for the Pinecone vector database <a href="https:&#x2F;&#x2F;github.com&#x2F;stepandel&#x2F;pinecone-explorer" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;stepandel&#x2F;pinecone-explorer</a>

[Other] Show HN: Mystral Native – Run JavaScript games natively with WebGPU (no browser) Hi HN, I&#x27;ve been building Mystral Native — a lightweight native runtime that lets you write games in JavaScript&#x2F;TypeScript using standard Web APIs (WebGPU, Canvas 2D, Web Audio, fetch) and run them as standalone desktop apps. Think &quot;Electron for games&quot; but without Chromium. Or a JS runtime like Node, Deno, or Bun but optimized for WebGPU (and bundling a window &#x2F; event system using SDL3).<p>Why: I originally started by starting a new game engine in WebGPU, and I loved the iteration loop of writing Typescript &amp; instantly seeing the changes in the browser with hot reloading. After getting something working and shipping a demo, I realized that shipping a whole browser doesn&#x27;t really work if I also want the same codebase to work on mobile. Sure, I could use a webview, but that&#x27;s not always a good or consistent experience for users - there are nuances with Safari on iOS supporting WebGPU, but not the same features that Chrome does on desktop. What I really wanted was a WebGPU runtime that is consistent &amp; works on any platform. I was inspired by deno&#x27;s --unsafe-webgpu flag, but I realized that deno probably wouldn&#x27;t be a good fit long term because it doesn&#x27;t support iOS or Android &amp; doesn&#x27;t bundle a window &#x2F; event system (they have &quot;bring your own window&quot;, but that means writing a lot of custom code for events, dealing with windowing, not to mention more specific things like implementing a WebAudio shim, etc.). So that got me down the path of building a native runtime specifically for games &amp; that&#x27;s Mystral Native.<p>So now with Mystral Native, I can have the same developer experience (write JS, use shaders in WGSL, call requestAnimationFrame) but get a real native binary I can ship to players on any platform without requiring a webview or a browser. No 200MB Chromium runtime, no CEF overhead, just the game code and a ~25MB runtime.<p>What it does: - Full WebGPU via Dawn (Chrome&#x27;s implementation) or wgpu-native (Rust) - Native window &amp; events via SDL3 - Canvas 2D support (Skia), Web Audio (SDL3), fetch (file&#x2F;http&#x2F;https) - V8 for JS (same engine as Chrome&#x2F;Node), also supports QuickJS and JSC - ES modules, TypeScript via SWC - Compile to single binary (think &quot;pkg&quot;): `mystral compile game.js --include assets -o my-game` - macOS .app bundles with code signing, Linux&#x2F;Windows standalone executables - Embedding API for iOS and Android (JSC&#x2F;QuickJS + wgpu-native)<p>It&#x27;s early alpha — the core rendering path works well &amp; I&#x27;ve tested on Mac, Linux (Ubuntu 24.04), and Windows 11, and some custom builds for iOS &amp; Android to validate that they can work, but there&#x27;s plenty to improve. Would love to get some feedback and see where it can go!<p>MIT licensed.<p>Repo: <a href="https:&#x2F;&#x2F;github.com&#x2F;mystralengine&#x2F;mystralnative" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;mystralengine&#x2F;mystralnative</a><p>Docs: <a href="https:&#x2F;&#x2F;mystralengine.github.io&#x2F;mystralnative&#x2F;" rel="nofollow">https:&#x2F;&#x2F;mystralengine.github.io&#x2F;mystralnative&#x2F;</a>

[Other] Nannou – A creative coding framework for Rust

[Other] Show HN: I built a CSV parser to try Go 1.26's new SIMD package Hey HN,<p>A CSV parser using Go 1.26&#x27;s experimental simd&#x2F;archsimd package.<p>I wanted to see what the new SIMD API looks like in practice. CSV parsing is mostly &quot;find these bytes in a buffer&quot;—load 64 bytes, compare, get a bitmask of positions. The interesting part was handling chunk boundaries correctly (quotes and line endings can split across chunks).<p>- Drop-in replacement for encoding&#x2F;csv - ~20% faster for unquoted data on AVX-512 - Quoted data is slower (still optimizing) - Scalar fallback for non-AVX-512<p>Requires GOEXPERIMENT=simd.<p><a href="https:&#x2F;&#x2F;github.com&#x2F;nnnkkk7&#x2F;go-simdcsv" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;nnnkkk7&#x2F;go-simdcsv</a><p>Feedback on edge cases or the SIMD implementation welcome.

[Database] Show HN: ShapedQL – A SQL engine for multi-stage ranking and RAG Hi HN,<p>I’m Tullie, founder of Shaped. Previously, I was a researcher at Meta AI, worked on ranking for Instagram Reels, and was a contributor to PyTorch Lightning.<p>We built ShapedQL because we noticed that while retrieval (finding 1,000 items) has been commoditized by vector DBs, ranking (finding the best 10 items) is still an infrastructure problem.<p>To build a decent for you feed or a RAG system with long-term memory, you usually have to put together a vector DB (Pinecone&#x2F;Milvus), a feature store (Redis), an inference service, and thousands of lines of Python to handle business logic and reranking.<p>We built an engine that consolidates this into a single SQL dialect. It compiles declarative queries into high-performance, multi-stage ranking pipelines.<p>HOW IT WORKS:<p>Instead of just SELECT <i>, ShapedQL operates in four stages native to recommendation systems:<p>RETRIEVE: Fetch candidates via Hybrid Search (Keywords + Vectors) or Collaborative Filtering. FILTER: Apply hard constraints (e.g., &quot;inventory &gt; 0&quot;). SCORE: Rank results using real-time models (e.g., p(click) or p(relevance)). REORDER: Apply diversity logic so your Agent&#x2F;User doesn’t see 10 nearly identical results.<p>THE SYNTAX: Here is what a RAG query looks like. This replaces about 500 lines of standard Python&#x2F;LangChain code:<p>SELECT item_id, description, price<p>FROM<p><pre><code> -- Retrieval: Hybrid search across multiple indexes search_flights(&quot;$param.user_prompt&quot;, &quot;$param.context&quot;), search_hotels(&quot;$param.user_prompt&quot;, &quot;$param.context&quot;) </code></pre> WHERE<p><pre><code> -- Filtering: Hard business constraints price &lt;= &quot;$param.budget&quot; AND is_available(&quot;$param.dates&quot;) </code></pre> ORDER BY<p><pre><code> -- Scoring: Real-time reranking (Personalization + Relevance) 0.5 * preference_score(user, item) + 0.3 * relevance_score(item, &quot;$param.user_prompt&quot;) </code></pre> LIMIT 20<p>If you don’t like SQL, you can also use our Python and Typescript SDKs. I’d love to know what you think of the syntax and the abstraction layer!</i>

[Other] Show HN: Build Web Automations via Demonstration Hey HN,<p>We’ve been building browser agents for a while. In production, we kept converging on the same pattern: deterministic scripts for the happy path, agents only for edge cases. So we built Demonstrate Mode.<p>The idea is simple: You perform your workflow once in a remote browser. Notte records the interactions and generates deterministic automation code.<p>How it works: - Record clicks, inputs, navigations in a cloud browser - Compile them into deterministic code (no LLM at runtime) - Run and deploy on managed browser infrastructure<p>Closest analog is Playwright codegen but: - Infrastructure is handled (remote browsers, proxies, auth state) - Code runs in a deployable runtime with logs, retries, and optional agent fallback<p>Agents are great for prototyping and dynamic steps, but for production we usually want versioned code and predictable cost&#x2F;behavior. Happy to dive into implementation details in the comments.<p>Demo: <a href="https:&#x2F;&#x2F;www.loom.com&#x2F;share&#x2F;f83cb83ecd5e48188dd9741724cde49a" rel="nofollow">https:&#x2F;&#x2F;www.loom.com&#x2F;share&#x2F;f83cb83ecd5e48188dd9741724cde49a</a><p>-- Andrea &amp; Lucas, Notte Founders

[Other] I made my own Git

[Other] Show HN: Externalized Properties, a modern Java configuration library Externalized Properties is powerful configuration library which supports resolution of properties from external sources such as files, databases, git repositories, and any custom sources

[Database] Quack-Cluster: A Serverless Distributed SQL Query Engine with DuckDB and Ray

[Other] ChatGPT Containers can now run bash, pip/npm install packages and download files

[Other] Porting 100k lines from TypeScript to Rust using Claude Code in a month

[Other] Show HN: Deploy backends without the hassle. An Open source alternative

[Other] Show HN: A small programming language where everything is a value This is a hobby project of mine that I started a few years ago to learn about programming language implementation. It was created 95% without AI, although a few recent commits include code from Gemini CLI.<p>I started out following Crafting Interpreters, but gradually branched off that until I had almost nothing left in common.<p>Tech stack: Rust, Cranelift (JIT compilation), LALRPOP (parser).

[CLI Tool] Show HN: Bucket – Encrypted file sharing for people who live in the terminal I built this because I am perpetually drowning in file transfer hell.<p>As a systems engineer, I am constantly moving massive files. Whether it&#x27;s getting new releases of custom ISOs to clients, downloading entire filesystems to spin up clones, or uploading massive support packages to L3.<p>I always have to get something to somebody. Usually, it&#x27;s something large that lives on a headless server. I can&#x27;t tell you how many times I&#x27;ve modified my &#x27;clone&#x27; script to tar most of a filesystem over ssh (&quot;hello --exclude, my old friend...&quot;) just to get 10GB that someone will need for two days.<p>Tedious is an understatement.<p>And the truth is, whatever I am sending, it is ephemeral. I download it, I upload it, I delete it. They download it only to delete soon after.<p>I kept wondering - why isn&#x27;t there a simpler tool that works where I spend 90% of my time -- the terminal? I wanted something that I could install quickly on whatever server I was currently connected to, and run a simple command to push to the cloud.<p>Enter Bucket – a secure, encrypted file sharing platform that works the way I do. I upload from the CLI, share with a link and unique secret, done. No browser required (unless you want it). The people you&#x27;re sharing with can simply click the link you share to download, no need for them to make an account (unless they want it).<p>What makes it different:<p>- CLI-first: Built for terminal workflows first, web UI second<p>- Encryption: End-to-end AES-256 encryption. We never see your files or your secrets. The secret generated during a push is required for the pull, keeping your data private even from us. If you lose the secret, you can&#x27;t access the file.<p>- Simple scalability: Free tier for trying it out, plans up to 250GB for serious use, enterprise coming soon.<p>- File retention: Bucket is designed for ephemeral storage. Upload things that people need <i>now</i>. Free tier keeps files for 3 days (email notifications are sent to you before they are deleted, in case you need to upload again)<p>Colleagues I have introduced this to tell me they use it every day. I know that I do.<p>Free tier gives you 3 GB to try it out - enough for most one-off transfers.<p>Here&#x27;s the workflow:<p><pre><code> $ bucket push api.bin Verifying upload... Upload complete! bID: bkdf1e848ea23e bURL: api.bucketlabs.org&#x2F;d&#x2F;bkdf1e848ea23e Secret: 9b1e15167403a88cadb7d0f4d2890856 Expires: 2026-01-28T03:15:18.277988Z </code></pre> Now you have a link for your browser-dwellers to simply click, paste the Secret and download. (go ahead, try on the link above to download the readme.md)<p>... but if the person you&#x27;re sharing with has bucket installed they simply run:<p><pre><code> $ bucket pull bkdf1e848ea23e Enter secret: ⠇ Downloading... Downloaded: api.bin </code></pre> Check your file:<p><pre><code> $ bucket list ID Filename Size Expires ---------------------------------------------------------------------- bkdf1e848ea23e api.bin 204.2 MB 2026-01-28T03:15:18.277988Z $ bucket account Account Info ------------ Subscription: free Used: 204.2 MB Quota: 3.00 GB To increase storage limits, visit: https:&#x2F;&#x2F;bucketlabs.org&#x2F;auth </code></pre> Everyone you needed to pull have the file? Go ahead and delete:<p><pre><code> $ bucket del bkdf1e848ea23e Deleted: bkdf1e848ea23e </code></pre> Built in Go. Single static binary, works anywhere. Download at bucketlabs.org

[DevOps] Show HN: Netfence – Like Envoy for eBPF Filters To power the firewalling for our agents so that they couldn&#x27;t contact arbitrary services, I build netfence. It&#x27;s like Envoy but for eBPF filters.<p>It allows you to define different DNS-based rules that are resolved in a local daemon to IPs, then pushed to the eBPF filter to allow traffic. By doing it this way, we can still allow DNS-defined rules, but prevent contacting random IPs.<p>There&#x27;s also no network performance penalty, since it&#x27;s just DNS lookups and eBPF filters referencing memory.<p>It also means you don&#x27;t have to tamper with the base image, which the agent could potentially manipulate to remove rules (unless you prevent root maybe).<p>It automatically manages the lifecycle of eBPF filters on cgroups and interfaces, so it works well for both containers and micro VMs (like Firecracker).<p>You implement a control plane, just like Envoy xDS, which you can manage the rules of each cgroup&#x2F;interface. You can even manage DNS through the control plane to dynamically resolve records (which is helpful as a normal DNS server doesn&#x27;t know which interface&#x2F;cgroup a request might be coming from).<p>We specifically use this to allow our agents to only contact S3, pip, apt, and npm.

[Other] Xmake: A cross-platform build utility based on Lua

[Other] Show HN: Sightline – Shodan-style search for real-world infra using OSM Data Hi HN,<p>I built *Sightline*, a Shodan-style search engine for *physical-world infrastructure*.<p>Shodan makes it easy to explore exposed internet services. Sightline applies the same idea to the real world, using OpenStreetMap as the data source.<p>You can search things like:<p>* “telecom towers in karnataka” * “power plants near mumbai” * “data centers in paris france”<p>or use structured queries:<p>* `type:telecom operator:airtel region:karnataka` * `type:data_center operator:google`<p>Sightline:<p>* uses Overpass API for querying OSM features * uses Nominatim for resolving countries, regions, and cities * avoids hardcoded geography * uses deterministic, rule-based parsing (no AI inference)<p>Repo: <a href="https:&#x2F;&#x2F;github.com&#x2F;ni5arga&#x2F;sightline" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;ni5arga&#x2F;sightline</a> Try it out: <a href="https:&#x2F;&#x2F;sightline-maps.vercel.app" rel="nofollow">https:&#x2F;&#x2F;sightline-maps.vercel.app</a>

[CLI Tool] Show HN: VM-curator – a TUI alternative to libvirt and virt-manager I&#x27;ve long wanted to harness QEMU&#x2F;KVM for my desktop virtual machines, but I&#x27;m befuddled by virt-manager&#x27;s lack of support for working NVIDIA 3D acceleration, dogmatic embrace of ugly XML, and the puzzling UI decision of having to click what seems like 15 buttons to attach an ISO to a VM image. When I further learned that NVIDIA&#x27;s broken 3D acceleration is the fault of libvirt as opposed to QEMU&#x27;s virtio driver, I had an idea...<p>Behold, vm-curator! A fast and friendly VM management TUI written in Rust. You can create, configure, organize, and manage VMs directly with QEMU. No libvert. No XML. No wonky UI&#x27;s. Just the right level of friendliness, customization, and speed to be really really useful.<p>The best part? 3D para-virtualization works with NVIDIA cards (via virtio-vga-gl!) No jumping through hoops to get GPU passthrough working!<p>(Disclaimer: This works great with other guest Linux VMs, but is not suitable for Windows gaming. If you want to game on Windows within a VM, passthrough is a must. vm-curator will have fast and friendly support soon.)<p>Looking for contributors (especially to help with the ascii art,) and donations are welcome. (Claude was a big help, but this was not a vibe-coded affair. We pair-programmed approx. 10,000 lines of code here. It was a great way to learn Rust, actually!)

[Other] Building a High-Performance Rotating Bloom Filter in Java