2025年12月4日 的 Show HN

Hello HN! We're a team of three building a new kind of web-based markdown editor.

There are many editors out there, so one is spoiled for choice, but Kraa's approach is a little different. It's trying to be both a minimal and distraction-free experience while being feature-rich and allowing for tons of use cases.

What Kraa's good for:

- Distraction-free writing & reading (minimal UI, performant, styling logic completely separated from the editing experience)

- Quick sharing of any written text – compared to many other writing tools, your content can be easily shared just by posting a link and giving 'read' or 'edit' access (we also have password-protection)

- Real-time chat / communities – Kraa has some unique features around real-time editing and our Chat widget allows for a frictionless chat experience. No send button.

- Kraa works well on mobile (though dedicated apps are planned)

---

Demo examples (all live, no login needed):

Blog article: https://kraa.io/kraa/examples/echolibrary

Long-form story: https://kraa.io/kraa/examples/insidekick

Magazine: https://kraa.io/weeklyinspiration

Kraa is built on top of ProseMirror (and TipTap) and Svelte.

You don’t need an account to try Kraa. We’d really appreciate your thoughts and feedback!

Hi HN,

I built a tiny tool because I couldn’t find a clean, distraction-free monthly planner that:

1. shows a clean monthly task view 2. doesn’t require an account 3. doesn’t sync or store anything online 4. works offline 5. is printable 6. and keeps a minimal, distraction-free aesthetic

So I made https://printcalendar.top/ — a minimal monthly task planner.

It’s intentionally simple. No logins, no integrations, no dashboards. Just a small tool for people who want structure without clutter.

Got tired of wiring up vector stores, embedding models, and chunking logic every time I needed RAG. So I built piragi.

  from piragi import Ragi

  kb = Ragi(\["./docs", "./code/\*\*/\*.py", "https://api.example.com/docs"\])

  answer = kb.ask("How do I deploy this?")

That's the entire setup. No API keys required - runs on Ollama + sentence-transformers locally.

What it does:

  - All formats - PDF, Word, Excel, Markdown, code, URLs, images, audio

  - Auto-updates - watches sources, refreshes in background, zero query latency

  - Citations - every answer includes sources

  - Advanced retrieval - HyDE, hybrid search (BM25 + vector), cross-encoder reranking

  - Smart chunking - semantic, contextual, hierarchical strategies

  - OpenAI compatible - swap in GPT/Claude whenever you want

Quick examples:

  # Filter by metadata
  answer = kb.filter(file_type="pdf").ask("What's in the contracts?")

  #Enable advanced retrieval

    kb = Ragi("./docs", config={
     "retrieval": {
        "use_hyde": True,
        "use_hybrid_search": True,
        "use_cross_encoder": True
     }
   })


  # Use OpenAI instead  
  kb = Ragi("./docs", config={"llm": {"model": "gpt-4o-mini", "api_key": "sk-..."}})

Install:

  pip install piragi
  PyPI: https://pypi.org/project/piragi/

Would love feedback. What's missing? What would make this actually useful for your projects?

I finally got tired of XKCD #2347 living rent-free in my head, so I built Stacktower: a tool that takes any real package’s dependency graph and turns it into an actual tower of bricks. Along the way I had to wrestle some surprisingly spicy problems.

Full blog post here: https://stacktower.io

The result is half visualization tool, half love letter to the chaos of modern dependency trees. Open-source, works with PyPI, Cargo, npm, and more.

Code: https://github.com/matzehuels/stacktower

I've been tinkering with a browser-based rhythm game creation tool. The pitch is simple: AI makes the music, Essentia.js figures out the beats, and you write the game logic in JS.

Demo: https://rhythm-seodang-web.vercel.app/

The problem I wanted to solve: most rhythm game workflows are heavy. Proprietary editors, manual charting, fixed gameplay patterns, desktop-only. I wanted something where you could just... open a browser tab and start messing around.

How it works:

- Music comes from AI services (Suno/Udio). No user uploads, no copyright headaches.

- Essentia.js (WASM port) runs entirely in-browser. Beat tracking, onset detection, energy curves, segment boundaries, all client-side.

- The output is a timing-only chart. What you do with that timing is up to you.

The fun part is the minigame sandbox. Charts and gameplay are completely decoupled. You define spawn rules, input handling, rendering, all in short JS functions. Same chart can become a taiko-style drum game, a directional swipe thing, or something experimental. Preview runs instantly.

Tech: Next.js, Essentia.js, custom rhythm engine, Canvas rendering, deployed on Vercel.

Current state: playable with sample tracks, chart generation works, minigame workshop is functional. In-platform AI music generation (prompt to track to playable) is next. Still rough around the edges.

If you've worked with WebAudio or rhythm engine internals, curious what you think. Feedback welcome.

Hey everyone, I am sharing my open-source storage engine project TidesDB. I'd love to hear your feedback and thoughts.

Cheers.

I build a lot of small products myself and I realized how many indie founders want an exit that doesn’t involve platforms charging large commissions or requiring big revenue numbers. Sometimes a project makes $50/month, sometimes $500, sometimes $0 but has a great codebase, and that’s enough value for someone else.

So I built buy-startups.com, a very simple, privacy-friendly marketplace for buying and selling small online startups.

It’s not fancy. There’s no commission, no listing fee, no hidden funnel. Just a straightforward way for indie devs and bootstrappers to list a project and connect with someone who wants to pick it up.

I built Banana Pro, a simple web app for text-to-image generation and context-aware editing using Google’s official flash image API. Upload an image (JPG/PNG/WebP, up to 6MB)

Edit with text prompts or blend styles

Get consistent, high-quality results in seconds

It’s free to try (first enhancement is free). Feedback and ideas welcome!

Hey HN,

I've been working on Onetone Framework for the past few years and finally releasing it as open source (AGPL 3.0).

*What is it?*

Onetone is an ambitious full-stack development framework that includes:

- Custom C interpreter with its own scripting language (.otc files)

- 27,000+ line OpenGL 3D graphics engine with PBR materials, skeletal animation, physics, and particle systems

- PHP web framework with MVC architecture

- Python utilities and tooling

- 716,000+ lines of code across 17 programming languages

*The scripting language features:*

- Classes, inheritance, generators, async/await

- Records, enums, pattern matching

- Built-in collections (ArrayList, HashMap, HashSet, TreeMap, etc.)

- Template strings, destructuring, spread operators

- Native bindings for OpenGL, Windows API, audio, networking

*Why build this?*

I run a game localization and needed a unified toolset for:

- Visual novel engines

- Translation management tools

- Quick prototyping with native performance

Instead of gluing together multiple languages and frameworks, I built one cohesive system.

*Current status:*

- Windows-focused (uses WinAPI extensively)

- Some features still in development (generators, full async support)

- Documentation is a work in progress

GitHub: https://github.com/onetoneframework/framework

Would love feedback from the community.

*Roadmap & Vision*

My goal is to evolve Onetone's scripting language to reach Python-level usability and ecosystem richness. I want developers to be able to pick it up as easily as Python while retaining native performance.

*A note on development process*

I want to be transparent: this project was developed with significant assistance from Claude (Anthropic's LLM). The codebase is a mix of hand-written code and LLM-generated code, with me directing the architecture, debugging, and integration.

I found this workflow surprisingly effective for a project of this scale – the LLM helped with boilerplate, documentation, and exploring implementation approaches, while I focused on design decisions and fixing the subtle bugs that AI still struggles with.

Whether you see this as "cheating" or the future of development, I think it's worth discussing. The 700K+ lines wouldn't exist without this collaboration, and I'm curious how others feel about AI-assisted open source projects.

There were many errors and strange bits of code produced by the LLM, and I spent a lot of time tracking down memory leaks; in fact, there isn’t a single piece of LLM-generated code that I didn’t end up modifying. I still think "vibe coding" has a number of issues.

Side project: exploring storing and querying OpenTelemetry data with duckdb, open table formats, and cheap object storage with some rust glue code.

Yesterday, AWS made this exact sort of data architecture lot easier with new CloudWatch features: https://aws.amazon.com/blogs/aws/amazon-cloudwatch-introduce...

I had a sweet deal $5k/mo customer. Their usage had been dropping for weeks, but I was busy building features to notice. Got the cancellation email on a Tuesday morning. That's a $60k/year in revenue I could have saved with a 10-minute phone call, if I'd known they were struggling. So I built https://firstdistro.com . It watches your customers' health (usage, engagement, activity) and alerts you when someone's at risk. One script tag. Health score in minutes. Alerts to Slack, email or Attio crm. Happy to answer questions about churn prediction, customer health scoring, or how we calculate signals. https://firstdistro.com

Hi everyone, I built a small desktop markdown note app called Mdit.

I wanted something as simple and minimal as Apple Notes.

I also wanted Notion style slash commands so you can write without knowing markdown syntax, while still keeping the freedom of plain local .md files like Obsidian.

I'd appreciate any feedback.

Website: https://mdit.app

GitHub: https://github.com/hjinco/mdit

Currently supports macOS only.

Hey HN, I've been trying to correct for one of the biggest difficulties with moving from SEO -> GEO, which is the lack of public data (in the same way that for SEO we had ahrefs/semrush/etc).

To try and start fixing this, I've built a searchable database of 15k brands, across 500 different industries, with daily updates.

Each morning, the tool goes out and queries 10k different prompts, and normalises the results to all the relevant brands.

You can also see some of the most interesting tech rivalries here: https://trakkr.ai/rankings/rivalries

Would love your thoughts on anything to add :)

I've been building an AI agent framework for myself for the last year because I got tired of re-explaining context to my LLM every morning.

Google just announced "Antigravity" to solve this, which looks amazing, but it locks you into their cloud and their models (Gemini). I wanted something that:

1. Runs 100% locally on my machine. 2. Works with any model (I switch between Claude 3.5 Sonnet and GPT-4o). 3. Persists memory in simple JSON files I can edit/version control.

I call it *EchoCopi*. It's a Python-based "memory organ" + a background worker script that executes tasks while I sleep.

I'm releasing the core memory module as open source (MIT) today. I'm also finalizing a "Full Autonomy" suite (background worker + VS Code integration) that I'll release later this month.

*Core Repo:* https://github.com/SparkSupernova/EchoCopi *Full Suite:* (Coming late Dec 2025)

Happy to answer questions about the architecture or how I use it to maintain a "sentient" coding partner.

I built Sonusly. Think Hacker News, but for music.

You search for a song, create a post with a title, and people can vote and discuss.

Here’s what you can do:

You can vote on posts you like by clicking ▲, with each vote costing 1 karma. Your karma starts at 100 and you earn 5 more each day you’re active if it’s below 100. You can save songs with the bookmark icon and save post discussions you care about for later. Click “listen” to hear a song in Spotify, use comments to discuss it with others, and click “share” to copy the post’s link/image.

Post example: https://www.sonusly.com/s/21B4gaTWnTkuSh77iWEXdS/p/dyi3czjdl...

First project I’ve ever shipped. I’d love feedback

Hi HN — ProbeOps Horizon is a Playwright-based geo testing browser. It lets you test how a site behaves from different regions using real Chromium sessions routed through regional egress nodes (instead of just header spoofing).

Use cases:

- geo redirects / pricing / localization

- CDN / edge routing differences (Cloudflare/Akamai/Fastly/CloudFront)

- consent banners, payment flows, bot checks

- per-region artifacts: HAR + TTFB/waterfall, screenshots/video, console/network logs

Current: local client (per tab/context region routing).

Links: (demo)

https://youtu.be/jXAnWzLue_c

Feedback I’m looking for:

which regions matter most,

which geo-specific behaviors you need to validate,

local vs cloud mode preference.

Happy to answer technical questions.

About two weeks ago I built a small text-comparison tool as a simple front-end project. Recently I ran into another annoyance: converting CSV/Excel/JSON with tools that upload files to servers, feel slow, or impose limits. Since I prefer privacy-first tools, I built this one as well.

100% local: All parsing and conversion run in Web Workers. No uploads.

Format support: CSV ↔ Excel (.xlsx), JSON, SQL, XML, Markdown.

Smart column restoration: Fixes copied tables that collapse into a single column (enable under “More Options”).

No size limits: Only limited by your RAM.

My goal is to grow this into a small, one-stop CSV/format toolbox. It just launched, so there may be rough edges — feedback is welcome.

Tech

Next.js, Tailwind, SheetJS, Web Workers, i18next.

Looking for feedback

Try it with your odd CSVs: unusual delimiters, quoted newlines, mixed encodings, huge files, broken pasted tables. Also curious whether the column-restoration feature feels intuitive.

Thanks for checking it out!

Hello Everyone,

My name is Karan, and I'm a Frontend Developer, but I feel like I'm more of a Design Engineer because of my love for creating UIs

When I started my development journey, I fell for frontend development and stuck with it ever since

But I noticed that many of my friends hated writing CSS because creating UIs is a very tedious and time-consuming process, and you have to be pixel-perfect

But at the same time, they also wanted their project to look premium with beautiful animations and a world-class user experience

That's when I thought

"What if anyone could integrate beautiful animated components into their website regardless of their CSS skills?"

And after six months of pain and restless nights, I finally built ogBlocks to solve this problem.

It is an Animated UI Library for React that contains all the cool animations that will make it look premium and production-grade

ogBlocks has navbars, modals, buttons, feature sections, text animations, carousels, and much more.

I hope you'll love it

Best Karan

A built a small WhatsApp bridge to keep track of claude code projects as they run on my laptop. There is an experimental permission hook to allow proxying of permission requests via the WhatsApp bridge. All messages are sent via a personal channel.

AI assistance and expert human design to deliver brand identities like a high-end agency would do.

Hi HN,

we’ve been working on Wan 2.6, a multimodal video model built for real creative production workflows—not just single random clips. Wan 2.6 generates complete videos with synchronized audio, motion design, and consistent character identity across shots.

Key features

Social Media Creation

Generate TikToks, Reels, and Shorts in 9:16 with voiceover and soundtrack in one pass. Describe the scene and Wan 2.6 handles motion, framing, timing, and output formats.

Built for Marketers

Turn ideas into pitch-ready product videos: cinematic lighting, active camera work, and synced narration. All outputs include commercial usage rights.

Tools for Filmmakers

Storyboard concepts, cut short scenes, and explore character-driven ideas. The motion engine improves transitions, continuity, and detail accuracy frame-to-frame.

E-commerce at Scale

Produce large batches of product videos with consistent style. Showcase rotations, lifestyle scenes, and multiple aspect ratios from one workflow.

Why Wan 2.6 stands apart

Multimodal Core: text + image + video + audio in one model

Two Variants: 5B for efficiency, 14B for maximum capacity

Precision Lip-Sync: accurate voice-to-motion alignment

Audio-Driven Video: generate visuals from music or audio tracks

Flexible Formats: 16:9, 9:16, 1:1 (MP4, MOV, WebM)

Full Usage Rights: unrestricted commercial deployment

We’d love feedback on:

Where consistent video generation fits in your workflow, and what APIs or tools make adoption easier.

Try it here:https://www.wan26.info/?i=d1d5k

I made https://startuplaunchday.com because I was constantly bouncing between Hacker News, random newsletters, Twitter, and government portals just to answer three simple questions:

What did people launch today?

What’s actually trending in startups right now?

Are there any funding opportunities I should know about?

StartupLaunchDay pulls those into one place and updates automatically every day.

On the site, there are three main views:

Launches – a daily feed + archive of startup/product launches, so you can quickly see what other founders are shipping and dig through previous days if you need inspiration or competitors to study.

Trends – a “what people are actually searching for” view of startup topics (ai, SaaS products, developer tools, startup funding, etc.), so you can get a sense of real demand and not just vibes.

Grants – curated funding opportunities grouped by category (technology, healthcare, small business, energy, etc.), with deadlines and links to the official pages, so you don’t have to dig through government sites yourself.

On top of that, founders can also list their own startup on the site. Listings get their own SEO page, a dofollow backlink, and permanent placement for a one-time payment (no recurring fees). The idea is: if people come here specifically to discover startups and research markets, your listing should be something they can actually find later, not just a tweet that disappears in a feed.