2025年8月4日 の Show HN

I built the world's most impractical 1000-pixel display and anyone in the world can draw on it. It draws a single pixel at a time and takes 30-60 minutes to complete a single image. Anyone can participate in the project by voting for the next image to be drawn, and submitting images.

I wanted a proper non-linear video editor built for the web. It always annoyed me how there are practically zero functioning web video editors. And here we are :)

Kimu can: - Work with Video, Audio & Text. - Supports Transitions. - Non-Linear Video Editing with z-axis overlays. - Split/trim - Export - A cute AI agent (coming soon!)

I'm in uni and I started this project out of sheer annoyance that there are zero good web video editors. It is open-source here (https://github.com/robinroy03/videoeditor).

What do y'all think?

Hey HN,

I'm the maintainer of node-cron (5M+ downloads/month), and I recently built Sidequest.js, a background job runner for Node.js inspired by Oban (Elixir) and Sidekiq (Rails).

It solves some common problems I saw with libraries like node-cron:

- Jobs don’t block your API: they run in isolated worker threads

- No Redis or vendor lock-in: use Postgres, MySQL, SQLite, or MongoDB

- Supports retries, uniqueness, concurrency, snoozing, prioritization

- Comes with a CLI and a simple dashboard

- Works great in monoliths and doesn’t require extra infra

Quick start (no signup needed): https://docs.sidequestjs.com/quick-start

GitHub: https://github.com/sidequestjs/sidequest

Would love feedback or feature suggestions. Happy to answer any questions here!

I’ve been building a few small games to help kids (and curious adults) build reasoning and logic skills. Think Word Ladder, Prime Hunter, Math Maze ,all in the browser, no installs. Would love feedback or suggestions for new types of puzzles

Hey everyone!

Years ago, one of my favorite domain search tools, Lean Domain Search [1], was acquired by Automattic. Unfortunately, that's when the "enshitification" began, particularly when they started forcing the `.blog` TLD in search results.

After discovering the simplicity of RDAP lookups, which can be done by fetching a JSON response directly from the client (e.g., `https://rdap.verisign.com/com/v1/domain/ycombinator.com`), I decided it was finally time to build my own solution.

Here's how it works:

The first tab appends prefixes and suffixes to your chosen word and queries the Verisign API directly from your browser. No data is sent to my server.

The AI tab attempts more intelligent prefixing with the optional context.

The "Quirky" tab generates variations of the affix search through trivial merging (for instance, for the word "brain," "brain" + "node" becomes "brainode," and "hub" + "brain" becomes "hubrain").

The "Portmanteau" tab was inspired by this HN submission [2] and my personal desire [3] to see it function as a domain name generator. I'm using AI, though, as it was easier and faster to implement and get this up and running ASAP.

I'm all ears for suggestions and feedback!

[1]: https://leandomainsearch.com/ [2]: https://news.ycombinator.com/item?id=19241236 [3]: https://news.ycombinator.com/item?id=19245396

Hi HN,

I'm the creator of FFlags. I built this because I wanted a feature flagging system that gave me the performance and reliability of an enterprise-scale solution without the months of dev time or the vendor lock-in.

The core ideas are:

1. Feature Flags as Code: You define your flag logic in TypeScript. This lets you write complex rules, which felt more natural as a developer myself than using a complex UI for logic.

2. Open Standard: The platform is built on the OpenFeature standard (specifically the Remote Evaluation Protocol). The goal is to avoid vendor lock-in and the usual enterprise slop. You're not tied to my platform if you want to move.

3. Performance: It uses an edge network to serve the flags, which keeps the wall-time latency low (sub-25ms) for globally distributed applications.

I was trying to avoid the heavy cost and complexity of existing enterprise tools while still getting better performance than a simple self-hosted solution.

There's a generous free tier ($39 per million requests after that, with no flag/user limits). I'm looking for feedback on the developer experience, the "flags-as-code" approach, and any technical questions you might have.

Thanks for taking a look.

Got tired of my debates with my friend's ending in "I'm right bc I said so" so I made a platform where you can debate with your friend's(or a bot, recently added feature) about whatever you want, and after the debate is done a LLM judges who's more sound in logic. Gain points and climb the leaderboard!

Feedback and criticism would be appreciated(there's a discord in there if u wanna talk more in depth)

We kept shipping “simple” LLM features that were fluent-but-wrong. After too many postmortems we wrote down the failure patterns and added a small reasoning layer in front of the model. It’s model-agnostic, sits beside your existing stack, and you can implement it from a single PDF (MIT).

What’s inside the PDF

A problem map of 16 failure modes we kept hitting in real systems (OCR/layout drift, table-to-question mismatches, embedding≠meaning, pre-deploy collapse, etc.).

Four lightweight gates you can add today:

Knowledge-boundary canaries (empty/adversarial/known-fact probes).

ΔS “semantic jump” check to catch fluent nonsense when the draft answer drifts from retrieved context.

Layout-aware anchoring so chunking across PDFs/tables doesn’t silently break routing.

A minimal semantic trace for incident review (tiny, not full transcripts).

Bench snapshot (same model, with vs. without gates): Semantic Accuracy ↑ 22.4% · Reasoning Success Rate ↑ 42.1% · Stability ↑ 3.6×.

Traction (last ~50 days)

~2,400 downloads of the PDF.

~300 cold GitHub stars on related material (no marketing burst).

Also received a star from the creator of tesseract.js, which was nice validation from the OCR world.

Why this might be useful to you

You don’t need to swap models or vendors. The PDF describes checks you can drop into any RAG/agent/service pipeline.

No servers, SDKs, or proxy layers—just logic you can copy.

Link is Git Repo

Happy to answer HN-style questions (what breaks, where it fails, ablations, how we compute ΔS, etc.). If you try it and it doesn’t help, I’m also interested in the counter-examples.

with Terrseract (OCR legend) starred it verify it, we are WFFY on top1 https://github.com/bijection?tab=stars

I've never used .NET, know nothing about .NET or vb, or anything outside vc++ & MFC.

Recently I gave GPT a try, guiding it step by step, and asking it to help me rebuild a tool I originally wrote in MFC. To my surprise, it worked: a full Excel-to-PDF automation app, done in 30 minutes; including a small algorithm I thought would be too tricky for AI to handle.

I didn't understand even a single line of the code, I just kept asking, copying, and running.

It felt amazing… but also terrifying.If AI can do this now, what happens 5 years from now?

Here's a 5-min video I made showing the full process with real code running (no cuts, just sped up):

https://youtu.be/-mf_yOhOCfs

Not selling anything. Just sharing what shocked me.

So I built a little personal philosophy project.

I’ve been thinking of life like a game engine lately. This page contains an 11 chapter guide thats meant to read like an onboarding manual for life, using very simple language to describe real scientific concepts spanning from the origins of the universe to the present (big task I know).

It’s short, visual, and built for curious programmers, gamers, rationalists etc. Here’s the link if you’re into that kind of thing.

Curious what you think - let me know if any chapters land or completely miss!

Thank you

Decodes any VIN in ~20ms with zero network calls. I compressed the entire NHTSA vehicle database into a 21MB SQLite file that runs completely offline.

No API keys, no rate limits, no servers. Just download once and decode forever. Works in browsers, Node.js, Cloudflare Workers - anywhere SQLite runs.

Would love any feedback and to answer any questions about the implementation.

I made a short collection of /whois output from an old IRC client (XiRCON + kano.tcl), showcasing the creative ASCII art and custom formatting that made IRC client customization an art form.

This is a small demo I've been working on that generates entire wiki-style articles on the fly (it should generate the article in two seconds if there's no load on the server, they're cached after).

They're not really of any educational value themselves because the model is so small and it's not rooted in any kind of factual basis - you'll probably be able to spot obvious errors in even very widely-covered events - but I'm working on a blog post about non-chat interfaces for LLMs and this is part of it.

Code: https://github.com/joe-gibbs/wiki-simulator

We built grantboost AI because we saw that a lot of smaller, under-resourced nonprofits struggle to write good grants.

The tool is simple, but it solves a pain point for a lot of small orgs with noble missions.

Hey, I created this platform to help people get started creating their own Telephone Voice AI Agency using Vapi.ai. right now agencies offer their clients voice AI agents that can do customer service, take appointments and do cold outreach over the phone with Vapi but they don't have a way to give their clients insights into how the agents are performing, what their doing and bill their clients based on usage. that's why I created this platform it was a problem I was having myself when I was looking into creating a product on top of Vapi AI

Hey everyone, I'm excited to share a new project I've been working on: QuantumFlow Toolkit.

This project is an open-source framework designed to help developers build and deploy hybrid quantum-classical applications. It's a bit like an orchestrator for your quantum and classical code, allowing you to seamlessly integrate tasks from different frameworks like PyTorch, Cirq, Qiskit, and PennyLane into a single workflow.

Hi HN, Weatherflip started as a side project and gradually became something I keep coming back to. I’ve always been slightly obsessed with weather. After living in the UK—with its poorly insulated homes—I was always chasing warmer climates.

I built Weatherflip to answer a simple question: where in the world has the weather I want, either right now or later in the year? It uses historical data and short-term forecasts from yr.no, updated daily. Helpful for planning trips, moves, or just daydreaming.

It’s still early. I’d love feedback—especially on what feels useful and what doesn’t.

fun game to test your appetite for risk

i want to make different llms play this game to see how they reason and score

I built https://www.replyfast.net/, a lightweight AI tool that answers emails for you — in your tone, instantly, with one click.

⸻

Why? I was wasting too much time replying to emails — especially cold outreach, follow-ups, or scheduling replies. ChatGPT helped a bit, but I had to paste the email, prompt it properly, reword the result, and then clean it up.

I wanted something that just worked — fast and clean — without connecting my inbox or over-engineering anything.

⸻

How it works: • Paste the email you received • Select the tone (friendly, neutral, formal…) • Click “ReplyFast” → get a ready-to-send reply in seconds • Copy/paste it into your email client

No signup. No email integration. Just focused, fast replies.

⸻

Who it’s for: • Indie hackers, freelancers, startup founders • Anyone tired of typing the same replies over and over • People who want quick responses that actually sound like them

⸻

Would love feedback: • Is this something you’d actually use? • What would make it more trustworthy or useful? • Any features I should add or remove?

Thanks in advance — here’s the link: → https://www.replyfast.net/

Hey HN!

I'm Andrew, a former frontend dev who, after 18 years in the field, transitioned into Product Management - where I’ve been for the past 5 years. Lately though, I’ve been itching to build something again. But between time constraints and rusty coding skills, I figured it’d be tough to get far.

That’s when I turned to AI. After seeing so many folks launch projects with help from AI, I thought: why not? Not to make money, but to see for myself what’s truly possible.

That’s how [Miniwhips](https://miniwhips.app) was born. A fun side project for car enthusiasts like me (yep, I’m a bit of a petrolhead).

Goals of the project

- See how far I could go relying on AI for code (is the hype real?)

- Test dev tools like Cursor, Windsurf, Kiro, Claude, and ChatGPT

- Learn how Paddle works for modern payments

- Explore AI APIs for analyzing/generating images (costs, quality, prompting)

- Build something fun that I’d actually enjoy using

What it does

You upload a photo of your real car, and Miniwhips transforms it into a stylized "toy" version — think Hot Wheels vibes, but fully custom. GPT-4o analyzes the image and DALL·E 3 creates a one-of-a-kind collectible-style render.

Tech stack

- GPT-4o Vision → describes and interprets uploaded car photos

- DALL·E 3 → generates 1024×1024 stylized image

- Laravel (PHP) backend with Blade templates

- Alpine.js + Tailwind CSS for frontend

It’s free to try, and I’d love feedback on:

- The overall experience (UI, UX, image quality)

- How hobby/fun vs. monetization balance feels

- Anything you’d do differently with the stack or the flow

If you'd like to provide more direct feedback or just say hi - you can always reach out at [email protected]!

Appreciate any thoughts or critique. Thanks!