Show HN for August 21, 2025

Hey HN! I built a proof-of-concept for AI memory using Git instead of vector databases.

The insight: Git already solved versioned document management. Why are we building complex vector stores when we could just use markdown files with Git's built-in diff/blame/history?

How it works:

Memories stored as markdown files in a Git repo Each conversation = one commit git diff shows how understanding evolves over time BM25 for search (no embeddings needed) LLMs generate search queries from conversation context Example: Ask "how has my project evolved?" and it uses git diff to show actual changes in understanding, not just similarity scores.

This is very much a PoC - rough edges everywhere, not production ready. But it's been working surprisingly well for personal use. The entire index for a year of conversations fits in ~100MB RAM with sub-second retrieval.

The cool part: You can git checkout to any point in time and see exactly what the AI knew then. Perfect reproducibility, human-readable storage, and you can manually edit memories if needed.

GitHub: https://github.com/Growth-Kinetics/DiffMem

Stack: Python, GitPython, rank-bm25, OpenRouter for LLM orchestration. MIT licensed.

Would love feedback on the approach. Is this crazy or clever? What am I missing that will bite me later?

Me and Guy (@guyb3) built ChartDB to generate ER diagrams from your database without a need of any database access (via query/sql/dbml). We started with an open-source version, and after seeing a lot of use we decided to make a cloud version.

Our OSS launch (1y ago) - https://news.ycombinator.com/item?id=41339308

Now we’re launching ChartDB Cloud - built for teams:

- Embed ERDs into docs, dev portals, or Miro/Notion etc.

- Collaborate in real-time (with live cursors like Figma)

- Keep diagrams always in sync with your database

- Organize large, messy schemas without pain

- Export DDL in multiple SQL dialects (solved deterministically)

- AI assistant to brainstorm and generate new schema objects or schema changes

We designed it so working with databases feels less like a chore and more like a creative process.

Would love feedback - especially from teams dealing with messy schemas or outdated docs.

https://app.chartdb.io

I first posted Splice CAD as an in-browser tool for making cable harnesses.

Since then it’s grown in both features and scope — the direction is moving from “harness-only” toward a lightweight CAD for wiring and electrical assemblies. New functionality includes:

Editing Enhancements

- Full undo/redo to easily restore editor state

- Multi-select & group actions to move, delete, and add components

- Bulk connect tool to create straight-through, crossover, or custom wiring patterns quickly

- Multiple connections per pin allow for daisy chains, etc.

Documentation Additions

- Multi-page PDF configurator to add A2, A3, or A4 pages for engineering drawing downloads

- WireViz YAML export (generate WireViz diagrams directly: https://www.danielrojas.net/projects/wireviz)

Library and Component Additions

- Expanded beyond harnesses to include categories for more applications: connectors, cables, breakers, fuses, switches, motors, power supplies

- Magic MPN button in the Component Creator to auto-fill specs from part numbers

Try it out, no signup: https://splice-cad.com/#/harness

Docs & tutorials: https://splice-cad.com/#/tutorial/

Hey, HN! I'm Aaron. I built an open-source Lovable for Claude Code users.

Platforms like Lovable, Replit Agent, and Bolt require separate API keys and $25+/month subscriptions. But if you’re already subscribed to Claude Pro or Cursor, you can use those plans directly without extra costs.

Claudable runs entirely locally through Claude Code (Cursor CLI also supported) and provides:

- Instant UI preview (similar to Lovable)

- Web-optimized, production-ready designs

- Direct Git integration

- One-click Vercel deployment

- Zero additional API costs

It’s open source and available today. I’m actively developing it and would love community feedback on what features to prioritize next.

GitHub: https://github.com/opactorai/Claudable

Happy to answer any questions!

Hey HN! I’m Lukas Dickie the founder of Changefly and I’m truly excited to share with you our latest release of Changefly ID with Anonymized Identity & Age Verification.

By putting privacy first and using a novel approach to account protection, Changefly ID offers a path to a safer, more secure, and less-intrusive internet for everyone:

- Changefly ID for anonymous authentication + ANONYMIZED identity & age verification for services that are required to verify minimum age

- Zero-knowledge proofs, secure multi-party computation, temporary unique identifiers

- Protects against bot attacks, bot scraping, identity theft, phishing, credential stuffing, online tracking, and other evolving threats

Changefly ID + Anonymized Identity & Age Verification is like showing a bartender a “Yes, I’m over 21” hologram badge instead of handing them your driver’s license with your name, address, and birthdate. They get what they need—but none of what they shouldn’t have.

We’re just getting started and I’d love to hear what you think, what you’d like to see next, and any feedback you have.

https://changefly.com

We built Weam because we felt existing AI tools didn’t work well for teams. Everything was scattered across chats, prompts, and workflows — hard to share, harder to organize.

Weam is an open-source platform that tries to fix that.

Organize prompts, chats, and agents into “Brains” (team folders).

Run agents and even Pro Agents for workflows.

Bring your own LLM keys (works with OpenAI, Anthropic, Gemini, Llama, etc.).

Self-hosted, so you keep control of your data.

Includes RAG pipelines for document-based AI.

It’s early but we’d love feedback. Repo here: https://github.com/weam-ai/weam

Docs: https://weam.ai/

Curious if anyone has run into the same pain of making AI tools “team-friendly.”

You can scroll left and right forever

If you press the ∞ button, you can scroll up forever

Sometimes I find comments that are so good I want to follow these people and catch up on what they're posting. This is my first Chrome extension. Looking forward to feedback on ways to improve it.

Hey HN,

I'm Syket, and I've noticed a pattern: most startup failures aren't due to bad ideas, but slow/expensive technical execution.

Over 30+ projects, I've developed a framework for rapid MVP development:

Week 1-2: Core features + authentication + payments Week 3-4: Mobile app + admin dashboard + analytics Week 5-6: AI features + optimization + launch prep

Recent examples: - Taplab Agency: Now UK's largest edu creator platform (https://taplab.agency) - Unithrive: Mentorship platform serving thousands of UK students (https://www.unithrive.net) - Connect Jew: NGO management system scaling across multiple cities (https://connect-jew.vercel.app)

What I've learned about startup tech: 1. *Start with revenue generation* - build payment processing first 2. *Mobile-first design* - 80% of users are on mobile 3. *AI integration* - users expect smart features now 4. *Performance = retention* - every 100ms delay costs users

The key insight: Don't build everything. Build the minimum that generates revenue, then iterate based on real user data.

I'm curious - what's been the biggest technical bottleneck in your startup journey? Happy to share specific solutions I've implemented.

Portfolio: https://syket.io

I’ve been working on a new Go library called Tap: https://github.com/yarlson/tap. It’s a Go port of the TypeScript library Clack https://clack.cc/, bringing its simple, interactive command-line prompts to Go.

Why: I often found myself rewriting input handling, spinners, and progress bars when building CLI tools. Existing options in Go felt either too barebones or too heavyweight. Tap aims to provide a middle ground: ergonomic, event-driven prompts with minimal dependencies.

What works today:

- Text and password input

- Confirm and single-select menus (generic, type-safe)

- Spinners and progress bars

- Styled intro/outro/cancel messages

- Built-in testing utilities

What’s coming next:

- Multi-select

- Autocomplete

- Grouped prompts and themes

Status: Early stage. APIs may change. Core features are tested but it’s not production-ready yet.

Repo: https://github.com/yarlson/tap

I’d love feedback on API design, missing features, and cross-platform quirks.

I built PictureToDrawing.org - a completely FREE AI tool that converts photos into artistic drawings. No signup, no account required, just upload and go. Would love to hear some advices!

The idea for Convosphere came to me a few years ago, when I was visiting a friend in Sardinia. I had just landed at the airport and needed to get to the train station. I wanted to share a taxi with someone else but didn’t feel like going around asking people one by one. That’s when it struck me: what if there were an app that allowed me to instantly connect with everyone nearby – like a group chat for people in the same location? That became the seed of the project.

On the technical side, Convosphere is built with Flutter for cross-platform development, and the backend currently runs on Firebase+Firestore. The app is available as a web app, Android, and iOS, and it supports location-based group and one-to-one chat, events, and a trust score system for safer interactions.

You can try Convosphere at the following links:

WebSite: https://site.convosphere.app/

WebApp: https://convosphere.app/

Android: https://play.google.com/store/apps/details?id=com.convospher...

iOS: https://apps.apple.com/it/app/convosphere/id6737458247

Feedback welcome!

Hi HN

I built a small open-source CLI tool that lets you generate professional invoices directly from the terminal.

Write invoices in YAML/JSON Export to clean PDF instantly No SaaS lock-in, no vendor dependency Great for freelancers & developers

The motivation came from constantly struggling with overcomplicated invoicing software while working on side projects. I wanted something minimal, scriptable, and developer-friendly.

Repo: https://github.com/username/invoice-cli Docs: https://github.com/username/invoice-cli#readme

I’d love your feedback - Does this solve a real pain point for freelancers/developers? - What features would you expect in a minimal invoicing CLI? - Any ideas for integrations (e.g. Stripe, Notion, etc.)?

Thanks for checking it out!

I have this tendency to over-engineer things, and sometimes it can be a good learning opportunity. ddns-go is what happens when you take DDNS seriously, and try to do everything in the most efficient way possible.

On Linux it opens a netlink socket and listens for network interface changes. If you use WireGuard for all outgoing traffic, and still want to be able to handle incoming connections from the physical interface, ddns-go can even help you maintain the policy routing rules (see ip-rule(8)).

On the BSDs it opens a route(4) socket for the same thing, except it is a bit more complicated here because for historical reasons the messages do not contain IPv6 address flags. You have to call ioctl(2) to retrieve the flags, if you want to filter out deprecated and temporary IPv6 addresses.

On Windows ddns-go uses the IP Helper API to get notified for interface address changes. Of course the API does not behave quite like how the documentation describes, so there needs to be workarounds.

ddns-go also speaks asuswrt's web API and can get the WAN IPv4 address directly from the router.

As a result of working on this project, I sent 3 separate CLs to the x/sys module.

[1] https://go-review.googlesource.com/c/sys/+/597915

[2] https://go-review.googlesource.com/c/sys/+/598895

[3] https://go-review.googlesource.com/c/sys/+/603755

Hello !

I was frustrated because every time I released a new feature on my website, users couldn't see it easily. So I created ShipShipShip, a self-hostable tool that creates a changelog page where users can see what has been released, what you're working on, and give their feedback.

The repo : https://github.com/GauthierNelkinsky/ShipShipShip

It's still very fresh and have a ton of features in mind to implement, so all feedback and suggestions are appreciated

Much better Parental Control than Google Family and Kaspersky Kids. It has useful features which are needed for a digital detox which they don't have.

Hello People,

I'd like to share a small project I've been working on called docmd. It's a command-line tool written in Node.js that generates clean, fast, and lightweight static documentation sites directly from Markdown files.

The core idea was to create something that gets out of your way. I found myself wanting a documentation tool that didn't require a complex setup with React, bundlers, or a large ecosystem of dependencies. I wanted to just write Markdown and have it turn into a beautiful, functional website.

docmd is an alternative for those who find tools like Docusaurus or Mintlify to be more than they need. It has no client-side framework dependencies, resulting in lightning-fast page loads and a simple build process.

Here are the core features:

- Zero Client-Side Frameworks: Generates plain HTML, CSS, and minimal vanilla JS. This makes sites extremely fast and lightweight.

- Standard Markdown: Uses standard Markdown with YAML frontmatter for metadata. No special syntax to learn.

- Single Executable: Install it via npm, and the docmd command is all you need (init, dev, build).

- Rich Components Out-of-the-Box: Includes built-in, pre-styled components like callouts, cards, tabs, and step-by-step guides using a simple container syntax (::: callout info).

- Theming & Dark Mode: Comes with multiple built-in themes and a light/dark mode toggle that works automatically.

- Built-in Plugins: Includes essential plugins for SEO, sitemaps, and analytics without needing external packages.

- No-Style Pages: An option to create completely custom landing pages with full control over the HTML, while still using the docmd build process.

- Deploy Anywhere: The output is a standard site/ folder that you can deploy to GitHub Pages, Netlify, Vercel, or any static host.

The entire documentation site for the project is, of course, built with docmd itself. You can see a live demo here: https://docmd.mgks.dev

The project is open source, and the code is on GitHub: https://github.com/mgks/docmd

I'm here to answer any questions you might have. I would genuinely appreciate any feedback, thoughts, or critiques you have on the approach, the features, or the code itself. If you find the project useful, a star on GitHub would be a great encouragement.

Thank you for taking a look.

So I watch a lot of educational YouTube videos for work. I'd always find myself wanting to share the good stuff with my network on LinkedIn or tweet about it.

But here's the thing that drove me crazy: I'd spend ages trying to summarize a 20-minute video into a decent post. Copy some quotes, try to capture the main points, format it nicely... and it would take me like 30 minutes just to write one post.

Then I started thinking - if this is painful for me as just a consumer, imagine how much worse it is for actual content creators and YouTubers + social media managers. They're making all this great content on YouTube but then have to spend hours manually converting it for LinkedIn, Twitter, etc. just to reach different audiences.

I figured there had to be a better way, so I built this thing called ThinkYT.

You just paste a YouTube URL and it turns into posts for different platforms. No more sitting there rewinding videos trying to get the exact quote or struggling to remember what the third point was.

What used to take me 30+ minutes now takes like 2 minutes.

I would love to hear your feedback!

This is a matchsticks puzzle solver, solves any matchsticks puzzle with two digit numbers in the format x [+-x/] y = z. Upto two moves allowed.

Some sample puzzles: https://github.com/fastcodejava/Matchsticks/tree/main/images