2025년 9월 4일의 Show HN

Hello HN,

I made this simple notepad for writing down whatever comes to mind. The goal is to force you to not get hung up on what you’ve already written. There’s no formatting, you can’t edit what you’ve typed, you can’t select text or move your caret around. There’s no backspace, delete, undo/redo, or paste. You can’t even see any words except the one you’re currently typing, and when you’re done with it, it fades away.

The text is all still there, just invisible. You can save everything you’ve written as a .txt if you want, or just empty your thoughts and let them disappear forever.

The idea was to encourage myself to do more stream-of-consciousness writing. I tend to edit while I write. This forces me to just keep moving forward, and I’ve found it very helpful so far. I was going to build this into my other writing tool, Drift, but I didn’t think it fit with the idea of the project, so I just made it its own thing. I’m calling it Vapor for now.

Enjoy, and let me know what you think!

Hi HN, I’m Felipe, a 9-5 developer aiming to become a full-time indie hacker. I struggled with flashcard apps that focus on decorating cards instead of learning concepts, so I built FlipCards to solve that problem.

What it does: FlipCards lets you:

- Create cards for any concept (math, language, coding, etc.) - Add multiple variations per card – the more variations, the better your understanding - Study smarter – the algorithm randomly selects variations so you can’t just memorize Q→A patterns

Why it’s different:

- Most apps use reverse cards (Q→A, A→Q). FlipCards uses variations to reinforce concepts in multiple contexts. - Powered by the SM2 spaced repetition algorithm, scientifically proven for long-term retention.

Pricing:

- Free – 1 deck, 3 cards, 1 study session - Annual – $20/year for unlimited decks and cards - Lifetime – $50 one-time for unlimited everything

I built FlipCards because I kept getting stuck decorating cards in other apps. Now I can create as many variations as I want, and the algorithm mixes them for me.

I’d love feedback from the HN community:

- Does this approach to flashcards make sense? - Would you use an app like this for learning?

Try it here: https://flipcardsapp.vercel.app

Thanks, Felipe

I made rm-safely, a simple shell wrapper that moves files to trash instead of permanently deleting them. It prevents accidental deletions from autocomplete mishaps or hasty rm -rf commands.

Should work as a drop-in replacement for rm but safer.

Would appreciate any feedback!

Hi HN,

I’m Thijs, new to the community and excited (and a bit nervous) to share what I’ve been working on: it's called Prototyper.

The motivation: I was curious how much more "taste" you could get out of an LLM if you built the entire infra yourself: tool calling, code execution, rendering—instead of layering on top of existing stacks. Over the past year I built a custom compiler, runtime, and design engine from scratch to see if this could make LLM-driven design genuinely better.

A few details:

- Own compiler + code runtime → no shadcn, no third-party UI kit, no external execution layer.

- Instant feedback (no compile/refresh lag).

- Deterministic design controls that integrate with llm-generated code.

Because we control the full stack, we can try things that I haven’t seen elsewhere, like inference tightly coupled with live code execution, or new approaches to integrating design controls with llm generated code.

It’s still early, and I’d love feedback from the HN community: is this approach actually better than what’s out there today? Everyone gets a free week to try it.

Happy to dive into the tech in the comments! :)

—Thijs

Hi HN! This is my first game — something I’ve always wanted to do.

It’s a small browser game built with Phaser 3, React, and the phaser template (https://github.com/phaserjs/template-react-ts/).

I made it in 2 days (like 8 hours in total real time) using gemini-cli. About 90% of the code was generated with AI, but I learned a lot by making fine tweaks.

It only works on PC since it’s a typical WASD + R (reload) shooter.

I’d love feedback on: - Gameplay (is it fun, too hard?) - Ideas for new features

Thanks in advance!

ps: I used cubes as a prototype, but now I kind of like them. Should I keep them or implement proper sprites?

I used to use https://doc.new when I needed to write quick scratchpad notes, but it takes like two seconds for Google Docs to be interactable, and ends up polluting my Drive with a bunch of "Untitled Docs".

Lately I've used a bookmarklet that opens a fullpage contenteditable div which is instantaneous and worked for my needs. But I wanted persistence when I accidentally close the tab, and data-urls can't use localstorage, so I spun up quicknote.zip.

It loads in the blink of an eye, works offline, and stores each day to localstorage. That's all it does, take it or leave it.

I hacked this together in an hour after seeing a request on Twitter. It went viral with 200k+ views in 2 days. Got live feedback from hundreds of people, added analytics, disclaimers, and open-sourced it. Curious what HN thinks about the idea and execution.

I’ve been building Invocly over the past few weeks — a tool that transforms entire documents into natural-sounding audio. Instead of copy-pasting text, you can just upload or share a document and listen to it on the go.

Features

Upload documents and convert them directly into audio

Clone your own voice with short samples for a personalized experience

Clean and simple interface — minimal setup required

Fast iteration — we’re shipping improvements quickly based on user feedback

I’d love to hear feedback from the HN community, especially on what formats or integrations would be most useful.

Try it here: https://invocly.com

Tsuki is a port of Lua 5.4 to Rust via C2Rust. I started working on this because I was tried to build a Lua bindings with overhead less than mlua and found out the approach I use is unsound due to Drop implementation on Rust is not guarantee to run. I can't find a solution for this so I decided to try C2Rust on Lua to see if it works. Turnout everything working as expected so I start to refactor the transpiled code to make it idiomatic to Rust.

prompt-to-proof is an open-source toolkit to (1) measure LLM streaming latency and throughput and (2) run a small, reproducible code eval, with hash-chained receipts you can verify. It targets OpenAI-style /chat/completions (works with OpenAI or local vLLM/llama.cpp).

‘Maybe I should’ve practiced for this’ was my first thought when we had our baby a few months ago, and she cried hard through the night in the first few weeks.

I created this website to simulate potential crying intensities and maybe help new and expecting parents practice some of those scenarios to be ready for them.

Understood. Hacker News requires a different approach. The audience is technical, skeptical of hype, and appreciates conciseness. The tone should be direct and informative, letting the project's technical merit speak for itself.

The convention is to post a link with a title, and then add a comment immediately with more details. Here are the two components for your post.

1. The Hacker News Title Use the Show HN: prefix. The title needs to be a factual, one-sentence summary of what the project is.

Title: Show HN: SHAde – A tool that turns Git commit hashes into deterministic art

2. The First Comment (from you, the OP) This is where you provide the essential context. No fluff. Get straight to the motivation, the technical details, and the link.

Comment:

Hi HN,

I created a command-line tool called SHAde. I was looking for a more aesthetic way to represent the history of a software project than the standard git log and was inspired by the idea of procedural generation. The goal was to create a unique, deterministic 'visual fingerprint' for every commit.

Here's how it works: It takes the SHA-256 hash of a Git commit and uses it to seed a pseudorandom number generator. This PRNG then drives the generation of geometric patterns (currently bars and spirals) and the selection of colors.

A key detail was the color generation. Instead of picking random RGB values, which often results in harsh combinations, it generates palettes in perceptually uniform color spaces (CIELAB/LCH). This helps ensure the results are more visually coherent. The entire process is stateless and deterministic, so a given commit hash will always produce the exact same image.

The tool can:

Render single commits to SVG or PNG.

Generate a browsable HTML gallery of a repository's recent history.

Create an animation showing the visual transition between two commits.

It's written in Python and is open-source on GitHub: https://github.com/LalwaniPalash/SHAde

I'd be interested to hear any feedback or answer any questions.

Hey HN! I've been frustrated with how complicated most digital download apps are for Shopify, so I built Alva Digital Downloads with a focus on simplicity and actually responsive human support.

What it does:

Auto-delivers digital files (PDFs, videos, ebooks, courses) after purchase Built-in fraud protection that holds suspicious orders for review (a common problem with digital products is they get delivered before the actual payment fails) No bandwidth limits even on the free plan Time limits, download counts, and IP restrictions to protect content Works with Shopify's native checkout and customer accounts

What makes it different: After trying dozens of digital download apps, most were either too complex or had invisible support teams. I focused on making setup as simple as it can be and providing real human support. The free plan is actually useful (1.5GB storage, unlimited bandwidth) - not just a trial.

Try it out:

If you have a Shopify store: Install directly from https://apps.shopify.com/alva-digital-downloads For more info and a demo link check out https://www.alvadigitaldownloads.com/

Tech stack: Built with Node.js, React, and uses CloudFlare R2 + Workers for file storage and delivery. I launched this in August and would love feedback on the UX, pricing, or features you'd want to see. Happy to answer any questions about building Shopify apps or the challenges of digital product delivery!

Super Web Scraper is your all-in-one solution for effortless web scraping, lead generation, and data extraction. With just one click, you can extract emails, phone numbers, addresses, images, links, product details, and more from any website. No coding required. Fast, accurate, and powerful.

I’ve been building an iOS app experiment called Neural Siege, a text-based system that imagines a future where rogue AI factions dominate the world, and humans fight back through dialogue rather than weapons.

Players face AI “bosses” that use different persuasion tactics—sarcasm, logic traps, meme-driven manipulation, psychological pressure—and must outwit them in conversation. Victories and losses affect a shared “war map” that tracks the state of the Resistance.

My goals are twofold:

Explore whether interactive simulations can make people more aware of how persuasive AI can be.

See if these kinds of systems could have value beyond entertainment, as a way to study human–AI dynamics.

I’d love feedback from this community:

- Could this type of experiment be useful in AI safety or human resilience research?

- Or is it better to treat it purely as a dystopian narrative for entertainment?

Hey HN,

While working with different LLMs in Go at Marble (open source Fraud & compliance automation platform), we got tired of juggling the specific SDKs and request/response structs for OpenAI, Anthropic, Cohere, etc. It makes switching providers or even just testing a new model more work than it should be.

To fix this, we built `llmberjack`. It’s a lightweight, no-frills Go library that provides a single, common interface over multiple LLM backends.

It's *not* a complex framework like LangChain. There are no agents, chains, or magic. It does just one thing: give you unified, typed, request builder functions so you can swap out providers with a one-line config change while still taking advantage of provider-specific bells and whistles. The goal is to avoid vendor lock-in and keep application code clean and simple.

Here’s a quick example:

```go gpt, err := openai.New(openai.WithApiKey("...")) gemini, err := aistudio.New()

llm, err := llmberjack.New( llmberjack.WithDefaultProvider(gpt), llmberjack.WithProvider("gemini", gemini), llmberjack.WithDefaultModel("gpt-5"), )

type Output struct { LightColor string `json:"text" jsonschema_description:"Color of the traffic light" jsonschema:"enum=red,enum=yellow=enum=red"` }

resp, _ := llmberjack.NewRequest[Output](). WithText(llmberjack.RoleUser, "What is the traffic light color indicating for cars to stop?"). Do(context.Background(), llm)

obj, _ := resp.Get(0)

fmt.Println("Traffic light is ", obj.LightColor) ```

It’s open source and the code is straightforward. We’d love to get feedback or contributions.

Check it out on GitHub: https://github.com/checkmarble/llmberjack

We've been wanting to give agents a phone number for sms / WhatsApp chats for very long, and today we finally built it. Dm me if you want to test it out -- we are still working on the infra / backend to generate more numbers. And if you have experiences building telecommunication infra or happen to be wanting the same thing I'd love a chat. Appreciated!

I built this because I was tired of juggling multiple API keys, dealing with different pricing models, and rewriting code just to test or use various AI models from different providers.

What is it? Wisdom Gate is a simple, OpenAI-compatible API gateway. You use one API key and just change the model parameter in your existing code to switch between models like GPT-5, Claude Sonnet 4, Gemini 2.5 flash, Nano Banana, and more.

The Hook: To get people to try it and gather feedback, I'm offering completely free and unlimited access to the DeepSeek models (V3 & R1) until January 1, 2026.

How it works (e.g., with n8n): Because it's OpenAI-compatible, you can use it directly with your existing libraries or tools like n8n just by changing the API key and setting the Base URL to https://wisdom-gate.juheapi.com/v1. It's a very straightforward setup.

Business Model: My goal is to offer the premium models (GPT, Claude, etc.) at competitive, simple pay-as-you-go rates, while keeping a free tier and powerful community models like DeepSeek free for as long as possible.

I'd love to get your feedback on the platform, the developer experience, and any other models you'd like to see integrated.

Thanks for checking it out!