Show HN за 30 октября 2025 г.

0github.com is a pull request viewer that color-codes every diff line/token by how much human attention it probably needs. Unlike PR-review bots, we try to flag not just by "is it a bug?" but by "is it worth a second look?" (examples: hard-coded secret, weird crypto mode, gnarly logic, ugly code).

To try it, replace github.com with 0github.com in any pull-request URL. Under the hood, we split the PR into individual files, and for each file, we ask an LLM to annotate each line with a data structure that we parse into a colored heatmap.

Examples:

https://0github.com/manaflow-ai/cmux/pull/666

https://0github.com/stack-auth/stack-auth/pull/988

https://0github.com/tinygrad/tinygrad/pull/12995

https://0github.com/simonw/datasette/pull/2548

Notice how all the example links have a 0 prepended before github.com. This navigates you to our custom diff viewer where we handle the same URL path parameters as github.com. Darker yellows indicate that an area might require more investigation. Hover on the highlights to see the LLM's explanation. There's also a slider on the top left to adjust the "should review" threshold.

Repo (MIT license): https://github.com/manaflow-ai/cmux

Hey HN, I’m Ilias, 19, from Paris.

I built Ellipticc Drive, an open-source cloud drive with true end-to-end encryption and post-quantum security, designed to be Dropbox-like in UX but with zero access to your data, even by the host.

What’s unique:

Free 10GB for every user, forever.

Open-source frontend (audit or self-host if you want)

Tech stack:

Frontend: Next.js

Crypto: WebCrypto (hashing) + Noble (core primitives)

Encryption: XChaCha20-Poly1305 (file chunks)

Key wrapping: Kyber (ML-KEM768)

Signing: Ed25519 + Dilithium2 (ML-DSA65)

Key derivation: Argon2id → Master Key → encrypts all keypairs & CEKs

Try it live: https://ellipticc.com

Frontend source: https://github.com/ellipticc/drive-frontend

Would love feedback from devs and security folks — particularly on encryption flow, architecture, or UX.

I’ll be around to answer every technical question in the comments!

idea’s simple: what’s your company culture really like? no fluff. no employer branding (aka glassdoor). just real employee takes.

ratetheculture.com

Hi HN, I’m Sol (YC06). I built BestInterest (https://bestinterest.app) solo to help co-parents communicate peacefully after divorce. It was just featured in WIRED (https://www.wired.com/story/ai-emotional-spellcheck-difficul...) - without any PR.

The idea came from personal experience — a painful divorce and challenging co-parenting communication. Courts often tell co-parents to keep things business-like and child-focused, which sounds simple but is brutally hard in practice. I realized AI can do what humans often can’t — remove emotion from the loop. It started as a simple theory — that AI could actually prevent emotional abuse in digital communication.

I’d taken a break from tech after years at Google (I was a PM there), but eventually brushed off the dusties and decided to build it myself.

THE STACK

Google Cloud + Firebase + Gemini (with some OpenAI functionality still in place) + Twilio. Front end: FlutterFlow. I bootstrapped everything — no funding, no team at first, just persistence, a supportive partner, and late nights after my kids were asleep. One upside of co-parenting: suddenly, a lot of kid-free time to think/code.

Early on, I knew I wanted an advisor with deep expertise in abuse recovery. During my own healing, Dr. Ramani Durvasula’s YouTube videos were life-changing, so she topped my “never-going-to-happen” list. I cold-emailed her — and to my surprise, she said yes.

Yesterday, WIRED featured our story: “Divorced? With Kids? And an Impossible Ex? There’s AI for That.” Side note: in the article, our leading competitor acknowledged using users’ personal correspondence for training data — which was… surprising.

It’s surreal seeing something that began with personal pain now helping others in such a profound way. I get emails every week from customers saying the app has changed their lives. It’s incredibly gratifying.

I’m learning as I go — building in a space this sensitive has challenged me in many ways and shown just how deeply this kind of technology is needed.

I was “fortunate,” in a strange way, to have lived this pain firsthand; it helped me understand what was needed for my niche. AI has just as much potential to create harm or false information as it does to bring light to dark places — protecting victims and helping people find safety in their communication.

Happy to talk about any of these:

- Bootstrapping a consumer AI app solo

- Restarting life as an entrepreneur after kids, a divorce, and an eight-year hiatus

- Transitioning to being a full-time dad (9 years ago)

- Growing a real subscriber base in a niche without a marketing budget — leveraging AI and SEO

- Building for a legally and emotionally complex community

- Using AI to protect against abuse — designing filters that help without over-censoring

- Breaking into a quasi-regulated industry where many assume court approval is required just to operate

AMA — happy to talk about the journey, the challenges, or anything else that resonates.

In https://news.ycombinator.com/item?id=45755788 there was mention of an IRC server using gawk, but no code. So I wrote an IRCd in pure bash.

I built a full backend-as-a-service platform — solo, in TypeScript — to fix what I hated about Supabase and Appwrite.

The Problem - Supabase: Too rigid. One schema model. No flexibility. - Appwrite: Too loose. No RLS. Security is manual.

The Solution: 3 Schema Types | Type | Use Case | Security | |------------|-----------------------------------|----------| | `Document` | Rapid prototyping (Appwrite-style) | Manual | | `Managed` | Secure apps (auto RLS + permissions) | *Secure by default* | | `Unmanaged`| Full SQL power (raw Postgres) | None |

### Killer Features - *APIs better than PostgREST*: Join tables *without FKs*, filter by nested columns - *Dashboard* (like Supabase Studio) – full CRUD, RLS editor, file browser - *Type-safe SDK* – zero config, full autocomplete - *Bun runtime* – faster than Node.js - *Self-host in 2 minutes* with Docker

```bash git clone https://github.com/Nuvix-Tech/nuvix cd nuvix && docker compose up -d ```

A small hobby project for browsing playlists from different countries.

Started this because configuring Fabric8 for EKS was driving me crazy. Building a client that auto-detects your environment (IAM roles, profiles, local configs) so you don't have to wire it all up manually. Still pretty early and building for EKS right now. Long way to go, but putting it out there early to see if others have hit the same problems or want to help shape where this goes.

I built DeepShot, a machine learning model that predicts NBA games using rolling statistics, historical performance, and recent momentum — all visualized in a clean, interactive web app. Unlike simple averages or betting odds, DeepShot uses Exponentially Weighted Moving Averages (EWMA) to capture recent form and momentum, highlighting the key statistical differences between teams so you can see why the model favors one side. It’s powered by Python, XGBoost, Pandas, Scikit-learn, and NiceGUI, runs locally on any OS, and relies only on free, public data from Basketball Reference. If you’re into sports analytics, machine learning, or just curious whether an algorithm can outsmart Vegas, check it out and let me know what you think: https://github.com/saccofrancesco/deepshot

Hi HN,

I realized there is a lot of duplicate effort in managing both git and Linear issues.

So made a simple CLI that feels like git, but actually modifies the Linear Issues as well.

I have 3 use cases right now:

1. lit switch "description of issue"

- Runs a search through Linear for issues matching the description - If multiple hits, will ask to disambiguate - Assigns issue to you, marks as in progress

git checkout the branch name (creates it if it doesn't exist)

2. lit commit "commit message/issue comment"

- Figures out correct issue based on branch - leaves a comment on the issue - git commit -m

3. lit checkout "Issue Title" -d "Description of Issue" -t f - Parses arguments: title, description (optional), issye type [bug, feature, improvement] (optional) - Creates new Linear Issue - Generates the Linear automation friendly branch name (exactly how Linear does it in the UI) - Does git checkout -b LinearbranchName

Give it a go and let me know what you think! Also happy to publish to brew if there is enough interest.

Happy Building!

I built my first macOS app to automatically open csv, xls files in Google Sheets.

I work as marketing, revops person and often have to combine data from different platforms for reporting purposes.

Google made the import flow super broken with too many clicks in between. So I built a simple solution that saves me some time.

Sharing it here, you can test it out for free. No subscription bullshit, one time payment to get unlimited usage if you like it.

Happy double clicking!

We're two developers who got hit twice by LLM cost problems and built LangSpend to fix it.

First: We couldn't figure out which features in our SaaS were expensive to run or which customers were costing us the most. Made it impossible to price properly or spot runaway costs.

Second: We burned 80% of our $1,000 AWS credits on Claude 4 (AWS Bedrock) in just 2 months while building prototypes of our idea but we had zero visibility into which experiments were eating the budget.

So we built LangSpend — a simple SDK that wraps your LLM calls and tracks costs per customer and per feature.

How it works: - Wrap your LLM calls and tag them with customer/feature metadata. - Dashboard shows you who's costing what in real-time - Currently supports Node.js and Python SDKs

Still early days but solving our problem. Try it out and let me know if it helps you too.

- https://langspend.com - Docs: https://langspend.com/docs - Discord: https://discord.gg/Kh9RJ5td

I've been following the development of posit arithmetic on and off for the past 2 years or so. A couple months ago I decided to try my hand at writing a software implementation.

For those who don't know, posit arithmetic is a new(ish) floating point format that offers _several_ advantages over IEEE floats, in particular at lower precisions. It generally has better precision and simpler design, and offers the quire, a way to compute dot products very fast and with no rounding errors (much can be said about its often surprising uses); this makes it suitable for HPC, neural networks, etc. I have some links to references in the README, if you're interested in learning more. I find it really interesting!

This crate aims to be a fully complete and correct standard-compliant implementation. Even though it's WIP, it already has a decent chunk of functionality: you can define types with arbitrary size and exponent size, convert to/from ints/floats, do + - × ÷, use the quire, etc. The main thing missing are elementary functions (exp, log, sin, etc.).

It's also very very fast, according to my benchmarks probably the fastest free implementation atm! If anyone knows of a faster one please get in touch, I'd love to add it to the benchmark suite.

Correctness is extensively tested too, exhaustively where possible, and with proptest where not.

Finally, the code is well-documented, including the internal algorithms, which might possibly make it (I hope!) a useful learning tool.

Thanks for reading!

I built and open sourced navcat, a new JavaScript 3D pathfinding library.

It helps you build web games, simulations, and creative websites that require pathfinding and navigation in 3D environments.

It supports navigation mesh generation from 3D geometry, navigation mesh querying, and has some higher level APIs for simulating a crowd of agents on a navmesh.

I built this after previously building recast-navigation-js, a webassembly port of a state of the art c++ library for navigation. But limitations with WASM <-> JavaScript interop and a desire to extend recastnavigation's functionality drove me to building navcat.

Say hi to the cats - https://navcat.dev/

examples: https://navcat.dev/examples github: https://github.com/isaac-mason/navcat npm: https://npmjs.com/package/navcat

This is our minimal example of a headless e-commerce backend. It uses @unchainedshop/platform as a node.js e-commerce framework.

I made a community plugin to get notes out of Granola and into Obsidian.

It syncs the transcript and the generated note.

Thanks to the AI coding robots for doing 90% of the typing.

If it's your jam contributions are welcome: https://github.com/tomelliot/obsidian-granola-sync