2025年12月5日 の Show HN

I'm sure folks here have seen pastebins a thousand times. There's no innovation left in this space – and that's kind of the point.

When I wanted to self-host a pastebin, every option I found was too much. Git-based version control, OAuth, elaborate admin panels. I just wanted something I could deploy in under a minute with a CLI that actually works.

So I built pbnj (yes, like the sandwich).

What it is:

- A minimal, beautiful pastebin with syntax highlighting for 100+ languages

- One-click deploy to Cloudflare (free tier gives you ~100,000 pastes)

- CLI-first: pbnj file.py → get a URL, copied to clipboard

- Memorable URLs: crunchy-peanut-butter-sandwich instead of x7f9a2

- Private pastes with optional secret keys

- Web UI for when you're not in a terminal

What it isn't:

- No accounts, no OAuth, no git integration

- No multi-user support (fork it and run your own)

- No expiring pastes, no folders, no comments

- Not trying to replace Gist or be a "platform"

Why not just use Gist? Maybe you want to own your data. Maybe you enjoy self-hosting things. Or maybe you're a little autistic like me and just like having your own stuff.

Live demo: https://pbnj.sh GitHub: https://github.com/bhavnicksm/pbnj CLI: npm install -g @pbnjs/cli

If this scratches an itch for you, I'd appreciate a star on GitHub. Happy to answer any questions!

AI coding assistants are productive but sloppy. They produce code that looks right but:

- Imports packages that don't exist - Uses placeholder functions that do nothing - Leaks patterns from JavaScript, Java, Ruby into Python - Leaves behind dead code and duplicates - Uses mutable default arguments

I built sloppylint to catch these "AI slop" patterns before they hit production.

     pip install sloppylint
     sloppylint .

It detects 100+ patterns across categories: - Hallucinated imports (20% of AI imports reference non-existent packages) - Placeholder code (`pass`, `...`, `TODO`) - Wrong-language patterns (.push(), .equals(), .forEach()) - Mutable defaults, bare excepts, dead code

This isn't a replacement for traditional linters - it catches the specific mistakes AI makes that humans wouldn't.

https://github.com/rsionnach/sloppylint

My team and I built Haven. It is a stripped down browser that only connects to verified financial institutions and blocks everything else. No extensions, no injected scripts, no overlays, no random third party code. It is not a general browser, it is just a controlled environment for banking and investing.

Why? I have spent years building browser extensions and know exactly how much power they have. The extension model is useful, but the security model is too loose and a single malicious install can see everything, rewrite everything, and mislead a user without leaving a trace. My dad learned that the hard way after installing a fake Zoom extension and then losing money from his banking account.

In addition to my dad’s story, the number of people adding AI into their current browsers or using AI native browsers is growing and offers amazing capabilities, but the risks are growing exponentially. I wanted to build a way for him and for anyone, from sophisticated users to normal users, to bank safely.

It is currently free for consumers as we learn how people use it and determine a future revenue model for providing secure sessions. At our core we believe in privacy, which is the reason we are building this product. We are funded by trusted and established investors including Google Ventures, Valley Capital Partners, and others.

If you want to try it the app is at https://starthaven.com.

Looking for people to test this out, get product feedback and feature requests!

I got back into church a couple years ago and would try taking notes with Apple Notes. It was a struggle trying to type notes while focusing on the sermon. Honestly, it would have been easier to write it in a notebook but in the end I built this iOS app to solve that problem.

You can record audio during a sermon (or upload files), and it transcribes using Whisper, then generates summaries, flashcards, and reflection questions tailored to Christian content.

The backend is Spring Boot + Kotlin calling OpenAI's API. Instead of deploying the backend through one of the cloud providers directly I decided to go with Railway. Users are notified with push notifications when their transcription and summary are completed. The iOS app uses SwiftUI and out-of-the-box SwiftUI components.

I worked with Spring Boot + Java a few years back when in fintech so it was cool to try writing something in Kotlin. I'm also a full-time Flutter dev that has been trying to get into Native iOS development and felt like I found a good use case for an app.

Currently only available in the US/Canada App Store. There is a free 3-day trial that you can use to give the app a go.

The goal was helping Christians retain more from sermons and build stronger biblical literacy.

Happy to answer questions about the architecture, AI prompting approach for Christian content, or anything else.

App Store link: https://apps.apple.com/us/app/bible-note-journal/id675227330...

Hey HN, I've just launched https://bananaprompts.fun/, a collection of carefully crafted prompts designed specifically for Nano Banana Pro, the AI model that excels at advanced image generation. Whether you're into photorealistic seascapes, cyberpunk portraits, or isometric claymation cityscapes, these prompts are optimized to help you create stunning visuals with minimal tweaking.

It's built for AI creators, artists, and devs who want reliable starting points to experiment with Nano Banana Pro. I'd love your feedback on the prompts, UI, or ideas for expansion—maybe more categories or community submissions?

Check it out and let me know what you think!

Hi HN,

There are several tools that help you search through the monthly "Who is Hiring" posts on Hacker News. The primary difference with this one is it includes chat, semantic search as well as a semantic map visualization (select "business" from the dropdown and expand to get a sense of how this can be used). Behind the scenes it uses LLM instructions in batch to extract, format, tag the job posts, computes UMAP after everything settles while of course making everything searchable.

You can use the basic text search to quickly filter the results or alternatively use semantic search (toggle via the button in the search bar).

Finally, you can chat with the job postings as well (click the Chat button). It has a basic RAG type pipeline but also includes some tools which make it possible to ask broader questions like "What are the general themes in the job postings this month?" and dig down from there.

Anyway, I hope people find this useful. Any feedback is welcome (either here directly or feel free to use the contact page here https://nthesis.ai/contact which dog foods the same mechanism - no contact info required).

If you want to build something similar there is an API and a nice (in my opinion) CLI tool than can be used to ingest data, search or chat as well.

OLake is our open-source tool for ingesting Database & Kafka data into Apache Iceberg. We recently redesigned the write pipeline and saw ~7x throughput improvements. Sharing the architecture decisions, trade-offs, and benchmarks.

I built Codesprint because I realized my raw WPM on standard typing tests didn't translate to coding. My muscle memory for prose was fine, but I was slow with brackets, semicolons, and indentation; the stuff that actually matters in a technical interview.

Codesprint is a typing tool specifically for code syntax. Instead of random words, it uses real LeetCode snippets (Two Sum, LRU Cache, etc.) so you're practicing realistic patterns.

The Tech: It’s built with Next.js, TypeScript, Chakra UI, and Framer Motion.

One interesting challenge was the scoring logic (lib/scoring.ts). Standard WPM assumes a word is 5 characters, but code is dense with symbols. I had to tweak the engine to handle indentation and "perfect word" streaks correctly to give a metric that actually feels fair for code.

I currently support Python, Java, C++, and JS. I’m looking to add Rust and Go next.

I’d love feedback on the typing "feel", specifically how the engine handles auto-indentation vs manual tabbing.

Thanks!

I built it for myself. After a lot of iterations, this is getting some traction.

It helps me organize my personal life.

Still trying to figure out what else it needs to do, and what can be improved.

Please provide feedback!

Author here: I built this because I wanted to learn about CBOR and how it's build and how it encodes data. I work with it on a almost daily basis, its s binary format that is defined in the RFC8949 specification.

It was my first time using AI to formalize the rules from an RFC and turn them into testable, working code. That's why I also explicitly say that this tool is not currently in use for production (I'm sure it will contain bugs!), it's just like a tinkering tool for me and at one point I hope to make this into something that people can maybe even use in their production system.

For visualization, I built a hex viewer that shows how the CBOR is decoded. It's really helping me understand the format better.

Right now it's just a basic version—you can encode, decode, and diff two CBOR values. There's still a lot of work to do. I'd like to add more educational content about what CBOR actually is and offer more tools.

I work in the Cardano space where CBOR is used heavily, so one thing I want to add is a function that can recognize which Cardano era a transaction (or part of a transaction) comes from.

There is still a ton of work to do and I have more ideas in mind, but for now that's my first version and I would like to get some feedback :)

The CBOR parser for that project is already open sourced, and I also built a small test suite to validate my CBOR parser against test cases and see how I compare to different parsers. That's why I created two projects that make this app possible. They are both linked in the about page and they're called Nachos and Taco.

We built this tool after years of frustration working on massive, undocumented COBOL systems. The core problem is that simple maintenance tasks turn into high-risk, multi-day exercises because there is no clear, automated way to map program dependencies or control flow.

The Challenges We Solve:

Dependency Sprawl: A single program change often ripples through dozens of Copybooks, which themselves include others. Our tool performs a full static analysis of the codebase to generate an accurate, visual dependency graph of every program and copybook included.

Obscured Control Flow: PERFORM blocks and deeply nested IF statements make control flow difficult to trace manually. The tool generates a clear, interactive control flow graph (CFG) for any selected program. Clicking a node in the CFG jumps directly to that line in the source.

Variable Tracing: Debugging is painful when variable usage is scattered across a 5,000-line file. We map the entire lifecycle of any variable: definition, modification, and usage points, streamlining the debugging process.

The goal was to move tribal knowledge out of people's heads and onto the code itself. The tool supports adding annotations and notes directly onto the flow graph nodes, linking critical business context directly to the source of truth.

We focused on generating highly accurate, navigable visualizations that senior developers can use to quickly understand and safely refactor complex logic.

We'd appreciate any feedback from others who have wrestled with legacy systems.

Hi HN,

I'm Rosa, a massage therapist for over 30 years. I noticed my clients felt relaxed after a massage, but their stress and muscle tension always came back. A one-hour massage isn't always enough to combat a long workweek.

I saw that people needed info on how to take care of their bodies between appointments, not just treatment.

That's why we built MASSAGE BY ROSA – a wellness platform to meet this need.

Here’s what we offer:

It’s a two-part deal:

-Massage Therapy: Hands-on therapy for pain relief – based on methods from South Florida.

-Online Body Therapy Courses: This is what I want to share. I turned my knowledge into video courses teaching self-massage, workstation adjustments, and ways to release tension. It’s like having a therapist help you stay well.

The Tech:

We’re keeping it basic with a static site for course content and subscriptions, which lets us focus on making great video lessons.

We're launching this to solve the problem of upkeep in physical wellness. It's for people who sit a lot and want lasting relief without constant professional help.

Check it out and tell me what you think:

Landing Page: https://massagebyrosa.com

I’m here to answer any questions about the platform, the methods, or how the business runs!

BinaryStorage is a PHP binary key/value store designed for fast and efficient data access. It supports any PHP serializable data, provides startsWith/contains searches, and includes data compaction to reduce disk usage.

GitHub: https://github.com/olivier-ls/binary-storage-php

Hey,

My name is Benjamin and as many developers, I was tired to not succeed my technical interviews, for any reasons.

I made SharpSkill.fr in order to change that.

Real Use Case, flashcards & interview simulators.

One goal : Destroy the next technical interviews.

Don't hesitate to share us your feedback.

This is a personal project that I've been working on very slowly now on the side for the past 5 years (yes, ridiculously long), using it for myself for a while and now finally have it ready to share to a wider audience. It's a web app for building printed wall calendars that include sun and moon info for each day, specific to the location you choose.

Rather than using accounts, each calendar gets its own unique URL by which it can be accessed. This way there are no accounts, no logging in, no tracking, etc. Everything is deleted 7 days after the last access, or if ordered, 18 months after the last access. There's no email list, no collecting/saving of email, etc. It's for creating and printing calendars; nothing else. This makes it free to try without worrying about getting harassed by remarketing...

I built it with nodejs and pdfkit for creating the calendar PDF that gets sent to the printer (lulu.com). I'm just using file storage of json documents to start with -- I know this is not ideal, but I started simple and lightweight and will fix things if/when needed. In the past I've been overly obsessed with using all the right tools and frameworks, but that can sometimes distract from the simple core goal, so for this I've intentionally taken a different approach (specifically since this is more of a "just for fun" project).

But I don't know -- adoption has been slow so far. Does anyone use printed calendars anymore?