MatrixNet: A Blueprint for a New Internet Architecture
Hi everyone,
Fair warning, this is a long post, so I've added a TL;DR at the very end for those short on time.
I know the concept has its problems, but I believe with the right minds, we can find the right solutions.
I'd like to share a conceptual framework for a different kind of internet or network at least, one designed from the ground up to be decentralized, censorship‑resistant, and hyper‑compressed. This isn't a finished product or a formal whitepaper. It’s a thought experiment I’m calling MatrixNet for now, and I'm sharing it to spark discussion, gather feedback, and see if it resonates.
The current web is fragile. Data disappears when servers go down, links rot, and valuable information is lost forever when a torrent runs out of seeders. What if we could build a system where data becomes a permanent, reconstructable resource, independent of its original host? Imagine if it were theoretically possible to hold a key to the entire internet in just 1 TB of data, allowing you to browse and download vast amounts of information completely offline.
## The Core Idea: Data as a Recipe
Imagine if, instead of shipping a fully built Lego castle, we only shipped a tiny instruction booklet. The recipient could build the castle perfectly because they, like everyone else, already owned the same universal set of Lego bricks.
MatrixNet operates on this principle. All data, websites, files, videos, applications, are not stored or transferred directly. Instead, it is represented as a "Recipe": a small set of instructions that explains how to reconstruct the original data using a shared, universal library of "building blocks."
Let's break down how this would work, step by step.
## Phase 1: Forging the Universal Matrix
The foundation of the entire system is a massive, static, and globally shared dataset called the Matrix.
### Gathering Public Data
We start by collecting a vast and diverse corpus of public, unencrypted data. Think of it as a digital Library of Alexandria:
- The entirety of Wikipedia.
- Open‑source code repositories (like all of GitHub).
- Public domain literature from Project Gutenberg.
- Common web assets (CSS frameworks, JavaScript libraries, fonts, icons).
- Open‑access scientific papers and datasets.
- Common data assets (videos, images).
### Creating the Building Blocks
This public dataset is then processed. The goal isn't to create a colossal file, but the most efficient and small Matrix possible.
The dataset is:
1. Broken down into small, fixed‑size chunks (e.g., 4 KB each).
2. Connected to a hashed index for fast retrieval, and all duplicates are removed.
The result is the Matrix: a universal, deduplicated collection of unique data “atoms” that forms the shared vocabulary for the entire network. Every peer would eventually hold a copy of this Matrix, or at least the parts they need. It is designed to be static; it is built once and distributed, not constantly updated.
The bigger it is, the more efficient it is at representing data, but the more impractical it becomes. We need to find the right balance—perhaps start with 10 GB / 100 GB trials. I foresee that with just 1 TB we could represent the entirety of the internet using some tricks described later.
## Phase 2: Encoding Information into Recipes
Now, let's say a user wants to share a file, document, photo, or even an entire application/website. They don't upload the file itself; they encode it.
### Chunking the Source File
The user's file is split into its own 4 KB chunks.
### Finding the Blocks
For each chunk, the system searches the Matrix for the most similar building block (using the hash table as an index).
- If an identical chunk already exists in the Matrix (common for known formats or text), the system simply points to it.
- If no exact match is found, it identifies the closest match—the Matrix chunk that requires the fewest changes/transformations to become the target chunk.
### Creating the Recipe
This process generates a small
Hi everyone,
Fair warning, this is a long post, so I've added a TL;DR at the very end for those short on time.
I know the concept has its problems, but I believe with the right minds, we can find the right solutions.
I'd like to share a conceptual framework for a different kind of internet or network at least, one designed from the ground up to be decentralized, censorship‑resistant, and hyper‑compressed. This isn't a finished product or a formal whitepaper. It’s a thought experiment I’m calling MatrixNet for now, and I'm sharing it to spark discussion, gather feedback, and see if it resonates.
The current web is fragile. Data disappears when servers go down, links rot, and valuable information is lost forever when a torrent runs out of seeders. What if we could build a system where data becomes a permanent, reconstructable resource, independent of its original host? Imagine if it were theoretically possible to hold a key to the entire internet in just 1 TB of data, allowing you to browse and download vast amounts of information completely offline.
## The Core Idea: Data as a Recipe
Imagine if, instead of shipping a fully built Lego castle, we only shipped a tiny instruction booklet. The recipient could build the castle perfectly because they, like everyone else, already owned the same universal set of Lego bricks.
MatrixNet operates on this principle. All data, websites, files, videos, applications, are not stored or transferred directly. Instead, it is represented as a "Recipe": a small set of instructions that explains how to reconstruct the original data using a shared, universal library of "building blocks."
Let's break down how this would work, step by step.
## Phase 1: Forging the Universal Matrix
The foundation of the entire system is a massive, static, and globally shared dataset called the Matrix.
### Gathering Public Data
We start by collecting a vast and diverse corpus of public, unencrypted data. Think of it as a digital Library of Alexandria:
- The entirety of Wikipedia.
- Open‑source code repositories (like all of GitHub).
- Public domain literature from Project Gutenberg.
- Common web assets (CSS frameworks, JavaScript libraries, fonts, icons).
- Open‑access scientific papers and datasets.
- Common data assets (videos, images).
### Creating the Building Blocks
This public dataset is then processed. The goal isn't to create a colossal file, but the most efficient and small Matrix possible.
The dataset is:
1. Broken down into small, fixed‑size chunks (e.g., 4 KB each).
2. Connected to a hashed index for fast retrieval, and all duplicates are removed.
The result is the Matrix: a universal, deduplicated collection of unique data “atoms” that forms the shared vocabulary for the entire network. Every peer would eventually hold a copy of this Matrix, or at least the parts they need. It is designed to be static; it is built once and distributed, not constantly updated.
The bigger it is, the more efficient it is at representing data, but the more impractical it becomes. We need to find the right balance—perhaps start with 10 GB / 100 GB trials. I foresee that with just 1 TB we could represent the entirety of the internet using some tricks described later.
## Phase 2: Encoding Information into Recipes
Now, let's say a user wants to share a file, document, photo, or even an entire application/website. They don't upload the file itself; they encode it.
### Chunking the Source File
The user's file is split into its own 4 KB chunks.
### Finding the Blocks
For each chunk, the system searches the Matrix for the most similar building block (using the hash table as an index).
- If an identical chunk already exists in the Matrix (common for known formats or text), the system simply points to it.
- If no exact match is found, it identifies the closest match—the Matrix chunk that requires the fewest changes/transformations to become the target chunk.
### Creating the Recipe
This process generates a small
JSON file called a Recipe—the instruction booklet. For each original chunk it contains:
- A pointer to the base building block in the Matrix (its hash).
- A transformation—a tiny piece of data (e.g., an XOR mask) that describes how to modify the Matrix block to perfectly recreate the original chunk. If the match is exact, the transformation is empty.
#### Example Recipe (conceptual)
The Recipe itself is just data, so it can be chunked, encoded, and given its own link. This allows nesting: a website's Recipe could link to Recipes for its images, CSS, etc.
Because links point to recipes (e.g.,
### Handling Encrypted Data
Encrypted files have high entropy and appear as random noise, so finding matching chunks in a public‑data Matrix is practically impossible.
- We Do Not Expand the Matrix: It stays static and contains only publicly available data; we never pollute it with encrypted material.
- Approximate & Transform: For each encrypted chunk we perform a nearest‑neighbor search to find the Matrix block that is mathematically closest (i.e., has the smallest bitwise difference).
- The Difference Is the Key: The system records the exact difference between the chosen Matrix block and the encrypted chunk using operations such as XOR, byte reordering, or other lightweight transformations. These transformation instructions are stored in the recipe.
Reconstruction: Retrieve the specified Matrix block, apply the recorded transformation, and you obtain the original encrypted chunk bit‑for‑bit. In this way the encrypted data is effectively “steganographically” embedded within innocuous public blocks, while the heavy lifting (the transformations) lives in a tiny Recipe file.
## Phase 3: A Truly Decentralized Web (Even Offline)
When files are represented only by recipes, the whole architecture of the web can change.
- Links Point to Recipes: Hyperlinks no longer resolve to IP addresses or domain names; they reference the hash of a Recipe.
- Offline Browsing: If you have the Matrix stored locally (e.g., on an external drive), you can browse huge portions of the network completely offline. Clicking a link simply fetches another tiny Recipe, which then reconstructs the target content using the local Matrix. Your browser becomes a reconstructor rather than a traditional downloader.
- The Network Is the Data: Going “online” merely means syncing the universal Matrix and exchanging new Recipes with peers.
### Solving Classic P2P Problems
1. Seeder Problem: In BitTorrent, a file disappears when there are no seeders. In MatrixNet, files never truly die because the Matrix is a permanent commons seeded by everyone. As long as a tiny Recipe exists somewhere (and it’s easy to back up or publish), the full file can be resurrected at any time.
2. Storage & Bandwidth Inefficiency: Sharing a 1 GB file traditionally requires transferring the whole gigabyte. With MatrixNet you only need to transfer a few kilobytes—the Recipe. The heavy data (the Matrix) is already widely replicated, so bandwidth usage drops dramatically.
## Challenges and Open Questions
- Computational Cost: Finding the “most similar chunk” for every 4 KB piece is CPU‑intensive. Viable solutions will likely need:
- Locality‑Sensitive Hashing or other ANN (approximate nearest neighbor) techniques.
- GPU/FPGA acceleration
- A pointer to the base building block in the Matrix (its hash).
- A transformation—a tiny piece of data (e.g., an XOR mask) that describes how to modify the Matrix block to perfectly recreate the original chunk. If the match is exact, the transformation is empty.
#### Example Recipe (conceptual)
{
"filename": "MyProject.zip",
"filesize": 81920,
"chunk_order": ["hash1", "hash2", "hash3", "..."],
"chunk_map": {
"hash1": {
"matrix_block": "matrix_hash_A",
"transform": "XOR_data_1"
},
"hash2": {
"matrix_block": "matrix_hash_B",
"transform": null // Exact match
},
"hash3": {
"matrix_block": "matrix_hash_C",
"transform": "XOR_data_2"
}
// … and so on for every chunk
}
}
The Recipe itself is just data, so it can be chunked, encoded, and given its own link. This allows nesting: a website's Recipe could link to Recipes for its images, CSS, etc.
Because links point to recipes (e.g.,
matrix://reddit…), clicking a hyperlink triggers decoding of a recipe file that then decodes the real website or data. The webpage will contain other links pointing to further recipes, creating a chain of reconstruction instructions.### Handling Encrypted Data
Encrypted files have high entropy and appear as random noise, so finding matching chunks in a public‑data Matrix is practically impossible.
- We Do Not Expand the Matrix: It stays static and contains only publicly available data; we never pollute it with encrypted material.
- Approximate & Transform: For each encrypted chunk we perform a nearest‑neighbor search to find the Matrix block that is mathematically closest (i.e., has the smallest bitwise difference).
- The Difference Is the Key: The system records the exact difference between the chosen Matrix block and the encrypted chunk using operations such as XOR, byte reordering, or other lightweight transformations. These transformation instructions are stored in the recipe.
Reconstruction: Retrieve the specified Matrix block, apply the recorded transformation, and you obtain the original encrypted chunk bit‑for‑bit. In this way the encrypted data is effectively “steganographically” embedded within innocuous public blocks, while the heavy lifting (the transformations) lives in a tiny Recipe file.
## Phase 3: A Truly Decentralized Web (Even Offline)
When files are represented only by recipes, the whole architecture of the web can change.
- Links Point to Recipes: Hyperlinks no longer resolve to IP addresses or domain names; they reference the hash of a Recipe.
- Offline Browsing: If you have the Matrix stored locally (e.g., on an external drive), you can browse huge portions of the network completely offline. Clicking a link simply fetches another tiny Recipe, which then reconstructs the target content using the local Matrix. Your browser becomes a reconstructor rather than a traditional downloader.
- The Network Is the Data: Going “online” merely means syncing the universal Matrix and exchanging new Recipes with peers.
### Solving Classic P2P Problems
1. Seeder Problem: In BitTorrent, a file disappears when there are no seeders. In MatrixNet, files never truly die because the Matrix is a permanent commons seeded by everyone. As long as a tiny Recipe exists somewhere (and it’s easy to back up or publish), the full file can be resurrected at any time.
2. Storage & Bandwidth Inefficiency: Sharing a 1 GB file traditionally requires transferring the whole gigabyte. With MatrixNet you only need to transfer a few kilobytes—the Recipe. The heavy data (the Matrix) is already widely replicated, so bandwidth usage drops dramatically.
## Challenges and Open Questions
- Computational Cost: Finding the “most similar chunk” for every 4 KB piece is CPU‑intensive. Viable solutions will likely need:
- Locality‑Sensitive Hashing or other ANN (approximate nearest neighbor) techniques.
- GPU/FPGA acceleration
for bulk similarity searches.
- Possible machine‑learning models to predict good candidate blocks.
- Dynamic Content: Real‑time applications, databases, and live streaming don’t fit neatly into static recipes. Additional layers—perhaps streaming recipes or mutable matrix extensions—would be required.
- Integration with the Existing Internet: Adoption hinges on low entry barriers (e.g., browser plugins, easy Matrix bootstrapping). Bridging mechanisms to fetch traditional HTTP resources when a recipe is unavailable will ease transition.
## Final Thoughts: A Paradigm Shift
MatrixNet invites us to rethink data sharing as reconstruction rather than copying. It envisions a future where our collective digital heritage isn’t locked in fragile silos but woven into a shared, permanent fabric.
- What if files never die, because their pieces already exist everywhere, just in a different shape?
- What if the only thing we need to share is how to rebuild information, not the information itself?
These questions are powerful. I’m sure there are flaws and challenges I haven’t covered—your critiques, ideas, and expertise are welcome.
Let’s collaborate to build a new internet that empowers users rather than corporations or governments.
If you’re a software engineer, cryptographer, network/security specialist, machine‑learning researcher, or simply passionate about decentralized systems, please reach out. I’ve created a GitHub repo for the community to start prototyping:
https://github.com/anedsa/Matrix-Net
For this I’m seeking collaborators to help run and grow this project, if you’d like to contribute, please DM me.
---
## TL;DR
MatrixNet = hyper‑compressed, decentralized web.
- Problem: Current web is fragile, censored, and bandwidth‑inefficient; data vanishes when servers go down.
- Idea: Share only a tiny Recipe (a few KB) that tells a device which chunks from a shared Matrix to pull and how to tweak them to recreate the original file.
- Benefits: Massive bandwidth savings, permanent availability (as long as the Recipe exists), censorship resistance, and offline browsing if you store the Matrix locally.
- Catch: Finding similar chunks is computationally heavy; dynamic content needs extra layers—but it’s a promising thought experiment for a more resilient web.
Feel free to comment, critique, or join the effort!
Edit: post missing a section
https://redd.it/1mq27by
@r_opensource
- Possible machine‑learning models to predict good candidate blocks.
- Dynamic Content: Real‑time applications, databases, and live streaming don’t fit neatly into static recipes. Additional layers—perhaps streaming recipes or mutable matrix extensions—would be required.
- Integration with the Existing Internet: Adoption hinges on low entry barriers (e.g., browser plugins, easy Matrix bootstrapping). Bridging mechanisms to fetch traditional HTTP resources when a recipe is unavailable will ease transition.
## Final Thoughts: A Paradigm Shift
MatrixNet invites us to rethink data sharing as reconstruction rather than copying. It envisions a future where our collective digital heritage isn’t locked in fragile silos but woven into a shared, permanent fabric.
- What if files never die, because their pieces already exist everywhere, just in a different shape?
- What if the only thing we need to share is how to rebuild information, not the information itself?
These questions are powerful. I’m sure there are flaws and challenges I haven’t covered—your critiques, ideas, and expertise are welcome.
Let’s collaborate to build a new internet that empowers users rather than corporations or governments.
If you’re a software engineer, cryptographer, network/security specialist, machine‑learning researcher, or simply passionate about decentralized systems, please reach out. I’ve created a GitHub repo for the community to start prototyping:
https://github.com/anedsa/Matrix-Net
For this I’m seeking collaborators to help run and grow this project, if you’d like to contribute, please DM me.
---
## TL;DR
MatrixNet = hyper‑compressed, decentralized web.
- Problem: Current web is fragile, censored, and bandwidth‑inefficient; data vanishes when servers go down.
- Idea: Share only a tiny Recipe (a few KB) that tells a device which chunks from a shared Matrix to pull and how to tweak them to recreate the original file.
- Benefits: Massive bandwidth savings, permanent availability (as long as the Recipe exists), censorship resistance, and offline browsing if you store the Matrix locally.
- Catch: Finding similar chunks is computationally heavy; dynamic content needs extra layers—but it’s a promising thought experiment for a more resilient web.
Feel free to comment, critique, or join the effort!
Edit: post missing a section
https://redd.it/1mq27by
@r_opensource
GitHub
GitHub - anedsa/Matrix-Net: A proof‑of‑concept platform where files are stored not as raw bytes but as tiny JSON “recipes” that…
A proof‑of‑concept platform where files are stored not as raw bytes but as tiny JSON “recipes” that point to a shared, immutable Matrix of public data blocks. By reconstructing content from this gl...
Emoji stenography
I have created this tool to encode messages in emojis then decode them back https://github.com/Teycir/EmojiSmuggler
https://redd.it/1mq5hbc
@r_opensource
I have created this tool to encode messages in emojis then decode them back https://github.com/Teycir/EmojiSmuggler
https://redd.it/1mq5hbc
@r_opensource
GitHub
GitHub - Teycir/EmojiSmuggler
Contribute to Teycir/EmojiSmuggler development by creating an account on GitHub.
My first open source project : ClearTx
Hey folks,
I built ClearTx, an open-source tool to organize and track your UPI transactions without sending your data to any server.
Works completely offline — your data stays with you
Simple tagging & filtering for accounts, merchants, or purposes
Clean UI for quick insights
Export reports whenever you need
Repo link: ClearTx
Would love feedback, feature suggestions, or contributions from fellow devs!
https://redd.it/1mq7ixj
@r_opensource
Hey folks,
I built ClearTx, an open-source tool to organize and track your UPI transactions without sending your data to any server.
Works completely offline — your data stays with you
Simple tagging & filtering for accounts, merchants, or purposes
Clean UI for quick insights
Export reports whenever you need
Repo link: ClearTx
Would love feedback, feature suggestions, or contributions from fellow devs!
https://redd.it/1mq7ixj
@r_opensource
GitHub
GitHub - kalel-commits/cleartx
Contribute to kalel-commits/cleartx development by creating an account on GitHub.
DevTool+ - A VSCode extension that provides common developer tools with well-designed UI
https://github.com/fuzionix/devtool-plus
https://redd.it/1mq5qfo
@r_opensource
https://github.com/fuzionix/devtool-plus
https://redd.it/1mq5qfo
@r_opensource
GitHub
GitHub - fuzionix/devtool-plus: A VSCode extension that provides common developer tools directly in code editor
A VSCode extension that provides common developer tools directly in code editor - fuzionix/devtool-plus
Writing a book in the age of open source: The power of engineering applied to writing
https://blog.incrementalforgetting.tech/p/sculpting-a-book-the-chisel?r=1tixy7&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false
https://redd.it/1mq9n2b
@r_opensource
https://blog.incrementalforgetting.tech/p/sculpting-a-book-the-chisel?r=1tixy7&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false
https://redd.it/1mq9n2b
@r_opensource
blog.incrementalforgetting.tech
Writing a book in the age of open source
The power of engineering applied to writing
I made a telegram bot template
I made this template for python-telegram-bot which covers almost every integral part of a telegram bot in addition to some nice decorators and utils. After about 6 years of python telegram bot development (not full time) I can finally say this template is indeed perfect, at least for me. Hope it'll be of use for you too
https://github.com/zmn-hamid/TeleTemplate
https://redd.it/1mqckdj
@r_opensource
I made this template for python-telegram-bot which covers almost every integral part of a telegram bot in addition to some nice decorators and utils. After about 6 years of python telegram bot development (not full time) I can finally say this template is indeed perfect, at least for me. Hope it'll be of use for you too
https://github.com/zmn-hamid/TeleTemplate
https://redd.it/1mqckdj
@r_opensource
GitHub
GitHub - zmn-hamid/TeleTemplate: a python telegram bot template with database integration
a python telegram bot template with database integration - zmn-hamid/TeleTemplate
🎬 FrameExtractionTool - Extract Perfect Frames from Videos with SwiftUI
**Hey Everyone!**
I just released my latest side project - **FrameExtractionTool** \- a simple iOS app for extracting high-quality frames from videos.
**📱 What it does:**
* Video Selection: Pick any video from your photo library
* Frame-Perfect Playback: Custom video player with precise timeline control
* Frame Marking: Mark specific moments during playback
* High-Quality Extraction: Save frames at original video resolution
* Custom Albums: Organize extracted frames in custom photo albums
**🛠️ Built with:**
* **SwiftUI** \+ **AVFoundation**
* **GitHub Actions** for automated builds
⚠️ **Important Disclaimer:**
This is a **very barebone app** as a side project of mine. The main goals were to:
* Learn how AI can help build apps
* Play around with SwiftUI and modern iOS development
* Experiment with SF Symbols and Icon Composer
* Explore automated CI/CD with GitHub Actions
**This app is very heavily developed using AI.** Bugs are expected! 🐛
**🎯 Why I built this:**
I often needed to extract specific frames from videos for presentations, memes, or reference images. And I don't see a same app that offers similar functionality for free. Therefore, I tried using AI and built it myself.
**🔗 Links:**
* **GitHub**: [FrameExtractionTool](https://github.com/CasperOng/FrameExtractionTool/)
* **Releases**: Check the releases page for unsigned IPA files.
**🤝 Contributing:**
Feel free to:
* Open issues for bugs 🐛
* Submit pull requests with fixes 🔧
* Suggest new features 💡
* Roast my (AI's) code (gently please) 😅
**TL;DR**: Made a simple frame extraction app with SwiftUI as an AI-assisted learning project. It works, has bugs, and is open source. Come try it! 😄
https://redd.it/1mqa4ib
@r_opensource
**Hey Everyone!**
I just released my latest side project - **FrameExtractionTool** \- a simple iOS app for extracting high-quality frames from videos.
**📱 What it does:**
* Video Selection: Pick any video from your photo library
* Frame-Perfect Playback: Custom video player with precise timeline control
* Frame Marking: Mark specific moments during playback
* High-Quality Extraction: Save frames at original video resolution
* Custom Albums: Organize extracted frames in custom photo albums
**🛠️ Built with:**
* **SwiftUI** \+ **AVFoundation**
* **GitHub Actions** for automated builds
⚠️ **Important Disclaimer:**
This is a **very barebone app** as a side project of mine. The main goals were to:
* Learn how AI can help build apps
* Play around with SwiftUI and modern iOS development
* Experiment with SF Symbols and Icon Composer
* Explore automated CI/CD with GitHub Actions
**This app is very heavily developed using AI.** Bugs are expected! 🐛
**🎯 Why I built this:**
I often needed to extract specific frames from videos for presentations, memes, or reference images. And I don't see a same app that offers similar functionality for free. Therefore, I tried using AI and built it myself.
**🔗 Links:**
* **GitHub**: [FrameExtractionTool](https://github.com/CasperOng/FrameExtractionTool/)
* **Releases**: Check the releases page for unsigned IPA files.
**🤝 Contributing:**
Feel free to:
* Open issues for bugs 🐛
* Submit pull requests with fixes 🔧
* Suggest new features 💡
* Roast my (AI's) code (gently please) 😅
**TL;DR**: Made a simple frame extraction app with SwiftUI as an AI-assisted learning project. It works, has bugs, and is open source. Come try it! 😄
https://redd.it/1mqa4ib
@r_opensource
GitHub
GitHub - CasperOng/FrameExtractionTool
Contribute to CasperOng/FrameExtractionTool development by creating an account on GitHub.
Anyone interested in an interesting project for an anti-bot?
All of you here likely know the dead internet theory, it’s especially bad on places like Reddit, twitter, comment sections etc.
I was thinking, maybe it’s time to try and get a group of folks together and build an open source bot detector, there has too be some way to train a program to detect likely bot activity with fairly high confidence.
Here’s why it needs to be open source and crowdsourced: we need huge amounts of data to train on human accounts and bot accounts.
But imagine a world where you can call on a Reddit bot, or twitter bot (ironic I know) and it will scan a account, then give a confidence score of how likely the account is run by a bot.
I’m fairly new into programming and ML, but I’m learning. I am however a technology consultant, meaning it’s literally my job to think of new ideas and ways to use tech, like this, then figure out how to make it happen.
So that’s what I’m doing now.
https://redd.it/1mq5xfi
@r_opensource
All of you here likely know the dead internet theory, it’s especially bad on places like Reddit, twitter, comment sections etc.
I was thinking, maybe it’s time to try and get a group of folks together and build an open source bot detector, there has too be some way to train a program to detect likely bot activity with fairly high confidence.
Here’s why it needs to be open source and crowdsourced: we need huge amounts of data to train on human accounts and bot accounts.
But imagine a world where you can call on a Reddit bot, or twitter bot (ironic I know) and it will scan a account, then give a confidence score of how likely the account is run by a bot.
I’m fairly new into programming and ML, but I’m learning. I am however a technology consultant, meaning it’s literally my job to think of new ideas and ways to use tech, like this, then figure out how to make it happen.
So that’s what I’m doing now.
https://redd.it/1mq5xfi
@r_opensource
Reddit
From the opensource community on Reddit
Explore this post and more from the opensource community
Bodhveda - open source notifications for developers
I wanted to add notifications to one of my products and I couldn't find a solution that was open source and I could self host but most are closed source, except Novu and are expensive $1 to $5 per 1,000 notifications.
So I built Bodhveda \- an open-source notification platform that lets developers add in-app notifications to their products in minutes — not weeks. Whether you’re launching your first product or scaling to millions, Bodhveda handles delivery, preferences, and analytics so you can focus on what matters.
GitHub - https://github.com/MudgalLabs/bodhveda
Website - https://bodhveda.com
Docs - https://docs.bodhveda.com
https://redd.it/1mqgjyb
@r_opensource
I wanted to add notifications to one of my products and I couldn't find a solution that was open source and I could self host but most are closed source, except Novu and are expensive $1 to $5 per 1,000 notifications.
So I built Bodhveda \- an open-source notification platform that lets developers add in-app notifications to their products in minutes — not weeks. Whether you’re launching your first product or scaling to millions, Bodhveda handles delivery, preferences, and analytics so you can focus on what matters.
GitHub - https://github.com/MudgalLabs/bodhveda
Website - https://bodhveda.com
Docs - https://docs.bodhveda.com
https://redd.it/1mqgjyb
@r_opensource
Bodhveda
Bodhveda - Notifications for developers
Bodhveda is a notification platform that lets developers add in-app notifications to their products in minutes — not weeks. You send. We deliver.
What are you building right now?
Tell us what your open-source project is about. Let’s check out each other’s projects
https://redd.it/1mqph1s
@r_opensource
Tell us what your open-source project is about. Let’s check out each other’s projects
https://redd.it/1mqph1s
@r_opensource
Reddit
From the opensource community on Reddit
Explore this post and more from the opensource community
I was tired of dealing with image-based subnoscripts, so I built Subnoscript Forge, a cross-platform tool to extract and convert them to SRT.
Hey everyone,
Like many of you who manage a media library, I often run into video files with embedded image-based subnoscripts (like PGS for Blu-rays or VobSub for DVDs). Getting those
into the universally compatible .srt format was always a hassle, requiring multiple tools and steps.
To solve this for myself, I created Subnoscript Forge, a desktop application for macOS, and Linux that makes the process much simpler.
It's a tool with both a GUI and a CLI, but the main features of the GUI version are:
* Extract & Convert: Pulls subnoscripts directly from MKV files.
* OCR for Image Subnoscripts: Converts PGS (SUP) and VobSub (SUB/IDX) subnoscripts into text-based SRT files using OCR. It also handles ASS/SSA to SRT conversion.
* Batch Processing: You can load a video file and process multiple subnoscript tracks at once.
* Insert Subnoscripts: You can also use it to add an external SRT file back into an MKV.
* Modern GUI: It has a clean, simple drag-and-drop interface, progress bars with time estimates, and dark theme support.
The app is built with Go and the Fyne (https://fyne.io/) toolkit for the cross-platform GUI. It's open-source, and I'm hoping to get some feedback from the community to
make it even better.
You can check it out, see screenshots, and find the installation instructions over on GitHub:
**https://github.com/VenimK/Subnoscript-Forge**
I'd love to hear what you think! Let me know if you have any questions or suggestions.
https://redd.it/1mqpmkb
@r_opensource
Hey everyone,
Like many of you who manage a media library, I often run into video files with embedded image-based subnoscripts (like PGS for Blu-rays or VobSub for DVDs). Getting those
into the universally compatible .srt format was always a hassle, requiring multiple tools and steps.
To solve this for myself, I created Subnoscript Forge, a desktop application for macOS, and Linux that makes the process much simpler.
It's a tool with both a GUI and a CLI, but the main features of the GUI version are:
* Extract & Convert: Pulls subnoscripts directly from MKV files.
* OCR for Image Subnoscripts: Converts PGS (SUP) and VobSub (SUB/IDX) subnoscripts into text-based SRT files using OCR. It also handles ASS/SSA to SRT conversion.
* Batch Processing: You can load a video file and process multiple subnoscript tracks at once.
* Insert Subnoscripts: You can also use it to add an external SRT file back into an MKV.
* Modern GUI: It has a clean, simple drag-and-drop interface, progress bars with time estimates, and dark theme support.
The app is built with Go and the Fyne (https://fyne.io/) toolkit for the cross-platform GUI. It's open-source, and I'm hoping to get some feedback from the community to
make it even better.
You can check it out, see screenshots, and find the installation instructions over on GitHub:
**https://github.com/VenimK/Subnoscript-Forge**
I'd love to hear what you think! Let me know if you have any questions or suggestions.
https://redd.it/1mqpmkb
@r_opensource
GitHub
GitHub - VenimK/Subnoscript-Forge: Subnoscript Forge is a GUI written in Go for subnoscripts
Subnoscript Forge is a GUI written in Go for subnoscripts - VenimK/Subnoscript-Forge
Shark WebAuthn library for .NET
Hello everyone,
Over the past few months, I have been working on a server-side implementation of the WebAuthn standard for .NET as an alternative to existing solutions.
You can check out the project here: https://github.com/linuxchata/fido2
I’d love to hear what you think. Do you see any areas for improvement? Are there features you’d like to see added? Any kind of feedback, advice, or questions are appreciated.
Thanks in advance!
https://redd.it/1mqsw2p
@r_opensource
Hello everyone,
Over the past few months, I have been working on a server-side implementation of the WebAuthn standard for .NET as an alternative to existing solutions.
You can check out the project here: https://github.com/linuxchata/fido2
I’d love to hear what you think. Do you see any areas for improvement? Are there features you’d like to see added? Any kind of feedback, advice, or questions are appreciated.
Thanks in advance!
https://redd.it/1mqsw2p
@r_opensource
GitHub
GitHub - linuxchata/fido2: Shark WebAuthn library for .NET
Shark WebAuthn library for .NET. Contribute to linuxchata/fido2 development by creating an account on GitHub.
Build the buddy that gets you! We open-sourced a complete AI voice interaction system!
Hey everyone, we just open-sourced Buddie: a complete, AI-powered voice interaction system we built from the ground up, so you can create your own AI buddy.
It's a full-stack platform for developers, hackers, and students, including custom hardware, firmware, and a mobile app. Therefore, you can use our solution to create various forms of AI devices, such as earphones, speakers, bracelets, toys, or desktop ornaments.
What it can do:
Live transcribe & summarize meetings, calls, or in-person chats.
Get real-time hints during conversations .
Talk to LLMs completely hands-free.
Context-aware help without needing to repeat yourself.
We've put everything on GitHub, including docs, to get you started. We're just getting started and would love to hear your ideas, questions, or even wild feature requests. Let us know what you think!
https://redd.it/1mqxxjs
@r_opensource
Hey everyone, we just open-sourced Buddie: a complete, AI-powered voice interaction system we built from the ground up, so you can create your own AI buddy.
It's a full-stack platform for developers, hackers, and students, including custom hardware, firmware, and a mobile app. Therefore, you can use our solution to create various forms of AI devices, such as earphones, speakers, bracelets, toys, or desktop ornaments.
What it can do:
Live transcribe & summarize meetings, calls, or in-person chats.
Get real-time hints during conversations .
Talk to LLMs completely hands-free.
Context-aware help without needing to repeat yourself.
We've put everything on GitHub, including docs, to get you started. We're just getting started and would love to hear your ideas, questions, or even wild feature requests. Let us know what you think!
https://redd.it/1mqxxjs
@r_opensource
Reddit
From the opensource community on Reddit
Explore this post and more from the opensource community
Anyone else got charged a few cents by GitHub for an open-source repo?
I just noticed something odd and wanted to check if it’s only me.
On July 27, 2025, I opened a support ticket with GitHub after receiving an invoice that showed my public open-source repository being billed under “metered” usage. From what I understand, public repos shouldn’t trigger these charges.
I only got a reply on August 12, and the next day they explained it was a bug: some users were charged a couple of cents for metered billing products, even when they shouldn’t have been. They reversed the charge and said they’re working on a fix.
That’s fine — but now I’m wondering: how many other people saw a tiny $0.02 or $0.03 charge and didn’t bother contacting support?
Has anyone else here noticed small, unexpected charges for public repos recently?
https://redd.it/1mqz5sr
@r_opensource
I just noticed something odd and wanted to check if it’s only me.
On July 27, 2025, I opened a support ticket with GitHub after receiving an invoice that showed my public open-source repository being billed under “metered” usage. From what I understand, public repos shouldn’t trigger these charges.
I only got a reply on August 12, and the next day they explained it was a bug: some users were charged a couple of cents for metered billing products, even when they shouldn’t have been. They reversed the charge and said they’re working on a fix.
That’s fine — but now I’m wondering: how many other people saw a tiny $0.02 or $0.03 charge and didn’t bother contacting support?
Has anyone else here noticed small, unexpected charges for public repos recently?
https://redd.it/1mqz5sr
@r_opensource
Reddit
From the opensource community on Reddit
Explore this post and more from the opensource community
Curl Keeps Cars Rolling: How a Tiny Open-Source Tool Powers Millions of Vehicles
https://ostechnix.com/curl-runs-in-top-car-brands/
https://redd.it/1mqzy8d
@r_opensource
https://ostechnix.com/curl-runs-in-top-car-brands/
https://redd.it/1mqzy8d
@r_opensource
OSTechNix
Curl Runs In The World's Top 47 Car Brands [August 2025 Report] - OSTechNix
Learn how curl, a small open-source tool, is built into hundreds of millions of cars, including models from the world’s top 47 car brands.
A secure and private Android call recording app without root?
For Android 10 or higher.
https://redd.it/1mr3jwv
@r_opensource
For Android 10 or higher.
https://redd.it/1mr3jwv
@r_opensource
Reddit
From the opensource community on Reddit
Explore this post and more from the opensource community
Build Your own AI Agents
We've released Denser Agent as an open-source project! You can build your AI agents with weather forecast, meeting scheduling and database analytics capabilities.
GitHub: https://github.com/denser-org/denser-agent/
Youtube tutorial & Demo: https://www.youtube.com/watch?v=3\_KledHS-WM
Happy building on your AI Agents! 🛠️
https://redd.it/1mr4e7g
@r_opensource
We've released Denser Agent as an open-source project! You can build your AI agents with weather forecast, meeting scheduling and database analytics capabilities.
GitHub: https://github.com/denser-org/denser-agent/
Youtube tutorial & Demo: https://www.youtube.com/watch?v=3\_KledHS-WM
Happy building on your AI Agents! 🛠️
https://redd.it/1mr4e7g
@r_opensource
YouTube
Build Powerful AI Agents with Denser - Open Source, Tutorial & Demo
Learn how to build production-ready AI agents with simple, extensible code! In this tutorial, we demonstrate the Denser Agent framework and show you how to create AI agents with real-world capabilities including weather forecasting, meeting scheduling,…
Zulip 11.0: Organized chat for distributed teams
https://blog.zulip.com/2025/08/13/zulip-11-0-released/
https://redd.it/1mr6795
@r_opensource
https://blog.zulip.com/2025/08/13/zulip-11-0-released/
https://redd.it/1mr6795
@r_opensource
The Zulip Blog
Zulip 11.0: Organized chat for distributed teams
We’re excited to announce the release of Zulip Server 11.0, containing hundreds
of new features and bug fixes: message reminders, support for channels without
topics, channel folders, and so much more! Over 3,300 new commits have been
merged across the project…
of new features and bug fixes: message reminders, support for channels without
topics, channel folders, and so much more! Over 3,300 new commits have been
merged across the project…
Project: Unstructored -> structured
I’m building an open-source AI Agent that converts messy, unstructured documents into clean, structured data.
The idea is simple:
You upload multiple documents — invoices, purchase orders, contracts, medical reports, etc. — and get back structured data (CSV tables) so you can visualize and work with your information more easily.
Here’s the approach I’m testing:
1. inference_schema
A vLLM analyzes your documents and suggests the best JSON schema for them — regardless of the document type.
This schema acts as the “official” structure for all files in the batch.
2. invoice_data_capture
A specialized LLM maps the extracted fields strictly to the schema.
For each uploaded document, it returns something like this, always following the same structure:
>
3. generate_csv
Once all documents are structured in JSON, another specialized LLM (with tools like Pandas) designs CSV tables to clearly present the extracted data.
💬 What do you think about this approach? All feedback is welcome
https://redd.it/1mr9rms
@r_opensource
I’m building an open-source AI Agent that converts messy, unstructured documents into clean, structured data.
The idea is simple:
You upload multiple documents — invoices, purchase orders, contracts, medical reports, etc. — and get back structured data (CSV tables) so you can visualize and work with your information more easily.
Here’s the approach I’m testing:
1. inference_schema
A vLLM analyzes your documents and suggests the best JSON schema for them — regardless of the document type.
This schema acts as the “official” structure for all files in the batch.
2. invoice_data_capture
A specialized LLM maps the extracted fields strictly to the schema.
For each uploaded document, it returns something like this, always following the same structure:
>
3. generate_csv
Once all documents are structured in JSON, another specialized LLM (with tools like Pandas) designs CSV tables to clearly present the extracted data.
💬 What do you think about this approach? All feedback is welcome
https://redd.it/1mr9rms
@r_opensource
Reddit
From the opensource community on Reddit
Explore this post and more from the opensource community