The best Hacker News stories from Show from the past day
Latest posts:
Show HN: Spam classifier in Go using Naive Bayes
Show HN: Spam classifier in Go using Naive Bayes
Show HN: An easy-to-use online curve fitting tool
This is a powerful online curve fitting tool that supports fitting dozens of commonly used functions and implicit functions. It features a clean interface and simple operation. If you need to perform curve fitting but don't want to learn professional software like Matlab or Origin, you can try this tool.
Show HN: Chirp – Local Windows dictation with ParakeetV3 no executable required
I’ve been working in fairly locked‑down Windows environments where I’m allowed to run Python, but not install or launch new `.exe` files. In addition the built-in windows dictations are blocked (the only good one isn't local anyway). At the same time, I really wanted accurate, fast dictation without sending audio to a cloud service, and without needing a GPU. Most speech‑to‑text setups I tried either required special launchers, GPU access, or were awkward to run day‑to‑day.<p>To scratch that itch, I built Chirp, a Windows dictation app that runs fully locally, uses NVIDIA’s ParakeetV3 model, and is managed end‑to‑end with `uv`. If you can run Python on your machine, you should be able to run Chirp—no additional executables required.<p>Under the hood, Chirp uses the Parakeet TDT 0.6B v3 ONNX bundle. ParakeetV3 has accuracy in the same ballpark as Whisper‑large‑v3 (multilingual WER ~4.9 vs ~5.0 in the open ASR leaderboard), but it’s much faster and happy on CPU.<p>The flow is:
- One‑time setup that downloads and prepares the ONNX model:
- `uv run python -m chirp.setup`
- A long‑running CLI process:
- `uv run python -m chirp.main`
- A global hotkey that starts/stops recording and injects text into the active window.<p>A few details that might be interesting technically:<p>- <i>Local‑only STT:</i> Everything runs on your machine using ONNX Runtime; by default it uses CPU providers, with optional GPU providers if your environment allows.<p>- <i>Config‑driven behavior:</i> A `config.toml` file controls the global hotkey, model choice, quantization (`int8` option), language, ONNX providers, and threading. There’s also a simple `[word_overrides]` map so you can fix tokens that the model consistently mishears.<p>- <i>Post‑processing pipeline:</i> After recognition, there’s an optional “style guide” step where you can specify prompts like “sentence case” or “prepend: >>” for the final text.<p>- <i>No clipboard gymnastics required on Windows:</i> The app types directly into the focused window; there are options for clipboard‑based pasting and cleanup behavior for platforms where that makes more sense.<p>- <i>Audio feedback:</i> Start/stop sounds (configurable) let you know when the mic is actually recording.<p>So far I’ve mainly tested this on my own Windows machines with English dictation and CPU‑only setups. There are probably plenty of rough edges (different keyboard layouts, language settings, corporate IT policies, etc.), and I’d love feedback from people who:<p>- Work in restricted corporate environments and need local dictation.
- Have experience with Parakeet/Whisper or ONNX Runtime and see obvious ways to improve performance or robustness.
- Want specific features (e.g., better multi‑language support, more advanced post‑processing, or integrations with their editor/IDE).<p>Repo is here:
`<a href="https://github.com/Whamp/chirp" rel="nofollow">https://github.com/Whamp/chirp</a>`<p>If you try it, I’d be very interested in:<p>- CPU usage and latency on your hardware,
- How well it behaves with your keyboard layout and applications,
- Any weird failure cases or usability annoyances you run into.<p>Happy to answer questions and dig into technical details in the comments.
I made a better DOM morphing algorithm
At least I think it’s better, but also I could also be missing something obvious.
I made a better DOM morphing algorithm
At least I think it’s better, but also I could also be missing something obvious.
Show HN: Unflip – a puzzle game about XOR patterns of squares
Show HN: Unflip – a puzzle game about XOR patterns of squares
Show HN: Unflip – a puzzle game about XOR patterns of squares
Show HN: Dumbass Business Ideas
Discover hilariously terrible business ideas that probably shouldn't exist. Get inspired by the worst startup concepts, share them with friends, and submit your own dumbass ideas!
Show HN: Dumbass Business Ideas
Discover hilariously terrible business ideas that probably shouldn't exist. Get inspired by the worst startup concepts, share them with friends, and submit your own dumbass ideas!
Show HN: Pegma, an open-source version of the classic Peg solitaire
Discover Pegma, the free and open-source version of the classic Peg solitaire game! Pegma offers a clean, minimal design and smooth gameplay across multiple platforms.<p>Key features:<p>Fully open-source code available on GitHub, inviting community contributions and transparency<p>Custom-designed font created by the developer to enhance the game’s unique style<p>Cross-platform support: play on iOS, Android<p>Lightweight, intuitive interface that stays true to the timeless puzzle mechanics<p>Try it now:<p>Website: <a href="https://pegma.vercel.app" rel="nofollow">https://pegma.vercel.app</a><p>GitHub: <a href="https://github.com/khlebobul/pegma" rel="nofollow">https://github.com/khlebobul/pegma</a><p>App Store: <a href="https://apps.apple.com/ru/app/pegma-peg-solitaire/id6754343848">https://apps.apple.com/ru/app/pegma-peg-solitaire/id67543438...</a><p>Google Play: <a href="https://play.google.com/store/apps/details?id=com.khlebobul.pegma">https://play.google.com/store/apps/details?id=com.khlebobul....</a><p>If you appreciate open-source projects and classic brain teasers, Pegma is definitely worth checking out!
Show HN: European tech news in 6 languages
I built a multilingual news aggregator for European tech in 6 languages
(EN, DE, FR, ES, IT, NL), filterable by audience (consumers, businesses, government) to help users discover relevant European alternatives.<p>A few interesting technical challenges:<p>PATTERN-BASED IMAGE GENERATION
Generic AI images all look the same ("holographic businessman with laptop").
I solved this with a JSON database of 60+ concrete visual patterns—context-aware selection prevents repetition. Funding news → coins/contracts, security → locks/shields. Much more visual diversity. Still needs tweaking though.<p>GRADUAL SITEMAP GROWTH FOR NEW SITES
Daily news pipeline generates lots of content. Challenge: Google indexes new sites slowly—exposing all 2K+ URLs at once resulted in 1-3% indexing rate. Solution: Dynamic rolling window in sitemap—only expose top 50 most recent articles (300 URLs), gradually increasing to 100 → 150 → 200 over 6 months as site builds authority. Result: Indexing rate improved to 10%.<p>AUTOMATED TRANSLATION PIPELINE
RSS aggregation → AI summaries and translation with context-aware prompts → human review queue.
Still tweaking quality per language pair.<p>STACK: Next.js 15 with Turbopack, PostgreSQL, deployed via Sliplane@Hetzner which works pretty ok.<p>STATUS: 1+ month post launch, 80 pages indexed, 100+ European apps featured (Mollie, Wire, Ecosia, LanguageTool, etc.)<p>FEEDBACK WELCOME ON:
• European news sources I'm missing?
• Better approaches to multilingual SEO?
• Translation quality red flags?
• How to optimize for LLM discovery (ChatGPT, Claude, etc.)?
Show HN: European tech news in 6 languages
I built a multilingual news aggregator for European tech in 6 languages
(EN, DE, FR, ES, IT, NL), filterable by audience (consumers, businesses, government) to help users discover relevant European alternatives.<p>A few interesting technical challenges:<p>PATTERN-BASED IMAGE GENERATION
Generic AI images all look the same ("holographic businessman with laptop").
I solved this with a JSON database of 60+ concrete visual patterns—context-aware selection prevents repetition. Funding news → coins/contracts, security → locks/shields. Much more visual diversity. Still needs tweaking though.<p>GRADUAL SITEMAP GROWTH FOR NEW SITES
Daily news pipeline generates lots of content. Challenge: Google indexes new sites slowly—exposing all 2K+ URLs at once resulted in 1-3% indexing rate. Solution: Dynamic rolling window in sitemap—only expose top 50 most recent articles (300 URLs), gradually increasing to 100 → 150 → 200 over 6 months as site builds authority. Result: Indexing rate improved to 10%.<p>AUTOMATED TRANSLATION PIPELINE
RSS aggregation → AI summaries and translation with context-aware prompts → human review queue.
Still tweaking quality per language pair.<p>STACK: Next.js 15 with Turbopack, PostgreSQL, deployed via Sliplane@Hetzner which works pretty ok.<p>STATUS: 1+ month post launch, 80 pages indexed, 100+ European apps featured (Mollie, Wire, Ecosia, LanguageTool, etc.)<p>FEEDBACK WELCOME ON:
• European news sources I'm missing?
• Better approaches to multilingual SEO?
• Translation quality red flags?
• How to optimize for LLM discovery (ChatGPT, Claude, etc.)?
Show HN: Encore – Type-safe back end framework that generates infra from code
Show HN: Encore – Type-safe back end framework that generates infra from code
Show HN: Encore – Type-safe back end framework that generates infra from code
Show HN: Tiny Diffusion – A character-level text diffusion model from scratch
This is a character-level language diffusion model for text generation.<p>The model is a modified version of Nanochat's GPT implementation and is trained on Tiny Shakespeare!<p>It is only 10.7 million parameters, so you can try it out locally.
Show HN: Tiny Diffusion – A character-level text diffusion model from scratch
This is a character-level language diffusion model for text generation.<p>The model is a modified version of Nanochat's GPT implementation and is trained on Tiny Shakespeare!<p>It is only 10.7 million parameters, so you can try it out locally.
Show HN: Tiny Diffusion – A character-level text diffusion model from scratch
This is a character-level language diffusion model for text generation.<p>The model is a modified version of Nanochat's GPT implementation and is trained on Tiny Shakespeare!<p>It is only 10.7 million parameters, so you can try it out locally.