We're building an auto-optimizing compiler for AI agents for speed & safety
https://github.com/stanford-mast/a1

https://redd.it/1owbfnf
@r_opensource

Arbiter — Open Source LLM Evaluation Library for Python

Howdy y’all!

I’ve been working on an open source evaluation library for Python called Arbiter (https://github.com/evanvolgas/arbiter).

Arbiter is an LLM evaluation framework that provides simple APIs, automatic observability, and provider-agnostic infrastructure for teams that work with AI.

It’s very much alpha software, but I would love thoughts and feedback on the library and roadmap, if anyone has anything they’d be willing to share. I’m especially curious to hear thoughts about the roadmap!



https://redd.it/1owbx52
@r_opensource

Open source purple-teaming telemetry & simulation toolkit
https://github.com/r3drun3/magnet

https://redd.it/1owahrr
@r_opensource

What would help you most from my active-development, Open-Source font’s italics? Would you appreciate placeholder italics for web apps and simpler deployment, or would you be fine waiting for my best design for that, which might take a few more months?
https://x.com/MarkFonts/status/1989074238570148232

https://redd.it/1owe08g
@r_opensource

I am building an single binary Learning Management System and looking for contributors.

Hi, I am building a single binary Learning Management System and looking for contributors.

Myself is a Moodle Admin in a University. I found Moodle hard to use and very error prone. Its codebase also has a lot of tech debt causing feature implementation extremely slow. It is using PHP so its plugins are buggy and often not useful because its is hard for develop to build plugins on top of PHP.


Therefore, I start the project Paideia LMS around a Month ago. I have been building this alone, developing, researching, writing doc, making youtube videos...

The education industry landscape is changing, with a shift to AI, the old LMS like Moodle and Canvas fails to keep up. I have hope on this LMS to replace Moodle and Canvas because it is single binary but scalable, built on modern tech like typenoscript, bun, react, payload CMS. But by the effort of myself I can only do so much.

Hopefully anyone might find the project interested and willing to help out. Any contribution or discussion is welcome.


github: https://github.com/paideia-lms/Paideia

demo: https://demo.paideialms.com/

doc: https://docs.paideialms.com/en/getting-started/

whitepaper: https://docs.paideialms.com/whitepaper-fall-2025.pdf

youtube: https://www.youtube.com/@PaideiaLMS

https://redd.it/1owiztu
@r_opensource

7 Steps to Make Your OSS Project AI-Ready
https://generativeprogrammer.com/p/7-steps-to-make-your-oss-project

https://redd.it/1owrrvp
@r_opensource

How do open-source projects gather real user references?

I maintain an open-source project that gets a steady flow of daily unique clones, often dozens per day. The point is that it is impossible to track *who* is using it and *how*. Some of those clones are probably bots and hobby users, but I'm sure part of the traffic comes from real companies and production projects.

I'd like to collect project references, not for marketing or vanity, but to understand real-world use cases, improve the roadmap, and show new users that the project is trusted in practice.

For maintainers here:

* **How do you find out who's using your work?**
* Do you rely on direct outreach, community channels, website forms, analytics, or something else entirely?
* Which approaches actually worked for you?

I added a note in the README asking users to reach out, but I'm not convinced anyone will take the initiative unless areg-sdk project is a well-known brand :)

Any insights or examples would be appreciated.

Here is the project: [Areg SDK](https://github.com/aregtech/areg-sdk) (The CTA with the note is in the README)

https://redd.it/1owt7u2
@r_opensource

When You Have Limited Runway, Is Pivoting to Services the Right Call?

so we put weam.ai on autopilot as it was open source and we were going throough funding problems we started choose automatorslab.ai to provide automation service does it loook like smart move or shoiuld we still focus on supporting our previous project, I'm a bit in a grey area right now.

https://redd.it/1owvy4f
@r_opensource

anyone else struggle with the awkward "can I share this idea?" conversation?

I was trying to write a research paper, but I was scared to share the idea or ask for feedback before publishin because I didn't wanna tell my experienced friend "hey please don't steal this" - like, I trust them completely, but you know that anxious feeling of "what if they did....."

It's such a stupid situation because:

You either don't share (and miss out on valuable feedback)
Or you share it but say nothing and just... hope for the best?

Neither option is great, and it's not about trust or being selfish - it's just about wanting to be on the same page about boundaries without making things weird.

Like, you can't just go "please don't take my idea" without sounding like you don't trust them. And NDAs are way too heavy for casual "hey what do you think of this?" conversations.

So I got frustrated and built something: **The Idea Protocol** - Let's pretend I didnt build it cuz I didn't want to seem like 'that person' in the group cuz literally that's how it feels like to set a boundary lol especially when it comes to the random ideas, maybe it's just me

It's basically like open source licenses, but for ideas. You just add one line when sharing:

> "This idea is shared under the Idea Protocol (*IP-FB*). For confidential feedback only."

There are 6 different "licenses" ranging from "strictly private" to "completely open, use however you want." So like if this becomes a thing, both the parties would just know oh okay this is your intention without making it personal - much like foss licenses - I mean I have many repos with it's own licenses but I would never be able to go legal anyway but like there's this standard in our foss world which I am so proud of,


You morally and socially ( ofc legally but yk what I mean ) respect the intent behind the license and at the same time they are not personally attacking you, it's just their preference about their code



Everyone knows the boundaries upfront, so no awkward conversations needed.

Works for everything from research ideas to "should I tell my crush how I feel?" (yes, that's a real use case I included - see the use case page lmao )

It's completely open source and not trying to replace actual NDAs - just fill the gap for casual idea-sharing where legal contracts are overkill

Anyone else deal with this problem or am I just overthinking things? cuz I legit wrote the licenses ( for making it seem like hey look it's a thing, it's nothing personal towards you ) - This is more of me wanting to see if I am the only one who overthinks to this level or maybe some of us think the same lol

https://redd.it/1owxe47
@r_opensource

OpenSigner – self-hostable key management for Web3/crypto wallets (OSS release, feedback welcome)

Hii!
in my team we just announced the open-source **OpenSigner**, a self-hostable key-management stack for **embedded Web3/crypto wallets**, and we’d love feedback from people who care about running their own infra and avoiding vendor lock-in.

here's some more info: The idea is :

* You run the infra (dockerized services + iframe).
* Keys are created client-side and split into shares (device / hot / shield).
* Signing happens in-memory with a 2-of-3 model, then wiped.
* You plug it into your existing auth (OIDC, passkeys, etc.) so users get a stable wallet without seed phrases or migrating if you ever change providers.

This is meant for teams who want “embedded wallets” UX but don’t want to hand over keys to a black-box SaaS or be locked in forever.

We’d really appreciate feedback on:

* the architecture & threat model,
* the defaults (2-of-3, components, policies),
* anything that looks over-engineered / under-thought from an ops or security POV.

Code & docs:

* Repo: [`https://github.com/openfort-xyz/opensigner`](https://github.com/openfort-xyz/opensigner)
* Docs: [`https://www.opensigner.dev`](https://www.opensigner.dev)

Happy to answer questions and iterate based on your comments.

**Would you trust this?**

Let me know your thoughts :)

https://redd.it/1owyvt7
@r_opensource

NoteVC: Version Control for Markdown
https://notevc.org

https://redd.it/1owy6vx
@r_opensource

This guy made a 3D engine horror game.. looks interesting :)

I just found this funny horror game engine.. look kind a cool.. its amazing that someone can just put together some code and make a 3D engine.. fascinating

I did not know where to share this so i just posted it here..

https://www.youtube.com/watch?v=uTpNw5sLNro

https://redd.it/1ox13r0
@r_opensource

Project Launch arkA — An open video protocol (not a platform). Early contributors welcome.

I’m building a new open-source project called arkA, and I’m looking for early contributors who want to help shape a protocol, not a platform.

arkA is trying to solve a simple problem:

Video today is centralized, algorithm-driven, and controlled.
There’s no open standard the way RSS is for blogs or podcasts.

arkA is an attempt to define:

a universal video metadata schema
an open video index format
a simple static web client that plays media from any storage
compatibility with IPFS, Arweave, S3, R2, etc.
a censorship-resistant, creator-owned distribution layer

Not a platform.
Not a startup.
Just an open protocol anyone can build clients and services on.

Early needs:

schema refinement (JSON)
first reference client (HTML/JS)
documentation
feedback on architecture
design discussions

Repo: https://github.com/baconpantsuppercut/arkA
Discussions welcome.

This is early, simple, and fun — perfect for OSS-minded builders.

https://redd.it/1ox3rsm
@r_opensource

Built an open-source WisprFlow alternative for Linux, Windows, and MacOS
https://github.com/josiahsrc/voquill

https://redd.it/1ox59gl
@r_opensource

Looking for contributors! TypeScript number utilities library
https://github.com/alyshukry/numerica

https://redd.it/1ox6uoh
@r_opensource

Polyemesis. OBS plugin to offload your streaming to Datarhei Restreamer.

I’ve been working on a new OBS plugin called Polyemesis, and I’m getting close to its first stable release. Before I promote it to 1.0.0, I’d really like help from the OBS community to put the new v0.9.0 build through real-world testing.

For anyone unfamiliar, this plugin connects OBS to Restreamer, an open-source streaming backend. The basic idea is that instead of having OBS stream directly to multiple platforms at once, you send a single feed to Restreamer and let it handle the distribution. This offloads a lot of local CPU/GPU/network cost, makes high-bitrate multistreaming more reliable, and gives you more control over formats, orientations, and platform-specific routing. If your system struggles with multistreaming or you want a cleaner, more flexible workflow, this plugin helps bridge OBS and Restreamer in a seamless way.

This 0.9.0 release is a major update. It introduces a redesigned interface that uses collapsible sections instead of tabs, respects all OBS themes through proper QPalette integration, supports macOS Universal builds, adds Linux ARM64, improves Windows compatibility, fixes a long list of memory and CURL issues, and includes a more complete test suite. The plugin went through months of debugging around authentication, headers, build systems, theme handling, Qt integration, and complex cross-platform behavior. It finally feels rock solid, but I’d like to confirm that in the wild before calling it stable.

I’m looking for people willing to test general stability, the new UI, the updated authentication flow, multistreaming performance, profile management, platform routing, and overall behavior during real workflows. If you can try it on your setup and tell me what breaks, what behaves strangely, or even what feels good, that would be incredibly helpful. I will monitor this thread and help anyone who needs guidance setting up Restreamer or getting the plugin working.

You can download the test builds here:

https://github.com/rainmanjam/obs-polyemesis/releases

If you run into issues, crash logs, theme problems, UX friction, or unexpected behavior under load, please share your findings. Everything is useful at this stage. Once this version has been hammered on a bit and confirmed stable across platforms, I’ll promote it to 1.0.0.

Thanks in advance to anyone willing to test.

https://redd.it/1ox3l87
@r_opensource

FullStacked: A local-first environment for web interfaces.
https://fullstacked.org

https://redd.it/1ox2pc3
@r_opensource

Looking for a collaborator (unpaid for now) for Image MetaHub
https://github.com/LuqP2/Image-MetaHub

https://redd.it/1oxab77
@r_opensource

Drawy, A New Whiteboard App for Linux!

This took me a long time, but after months of working during my free time, I'm extremely excited to share Drawy! It's an infinite, whiteboard desktop app written in Qt/C++.

# Motivation

Linux has had some apps with whiteboard features, like Xournal++ and Lorien. However, they have issues such as not having an infinite canvas (Xournal++) or lacking enough features (Lorien). That's why I decided to build Drawy, especially for Linux users. It's similar to Excalidraw but runs natively on your desktop, making it fast and lightweight. It's still in the alpha stage, but I have implemented key features that everyone needs:

Basic tools like pen, rectangle, ellipse, line, arrow, and text
Wacom tablet support with pressure sensitivity
Undo/redo support
Save/load files

Even though this seems very basic, it took an enormous amount of effort to develop. Drawy is still very stable to use (I've used it a lot to teach my students!)

# GitHub

The project is completely open source and licensed under the GNU General Public License V3. You can find the source code here: https://github.com/Prayag2/drawy

https://redd.it/1oxc5ep
@r_opensource

Licensing Problem

Hi everyone, I have less than one year of experience and currently work as a web developer. Recently, I was assigned to implement an algorithm that I found quite challenging (I won’t go into specifics, as it might reveal my identity). To figure it out, I looked into a library’s open source code and initially copied parts of it. While doing that, I noticed the library was licensed under MIT, which led me to research software licensing, something I wasn’t fully aware of before. After learning more, I decided not to copy the code directly. Instead, I used the idea behind the algorithm and wrote my own implementation in a different programming language, with a different structure. Now I’m unsure about the ethics and legal implications. If I re-implemented the same logic but with my own code and design, do I still need to include the MIT license for my work, or is this okay to use without attribution?

https://redd.it/1oxcxt1
@r_opensource

Just released my first open-source web app: user-flow-library (useful for dev shops and UI designers)

Hi all,

I’ve just open-sourced my first web app: github.com/alvinjchoi/user-flow-library

It’s built on Next.js and designed for UX/product folks to define & visualize user flows.

I’ve personally found it makes it a lot easier to align with clients on how their app will look and behave.

I’d appreciate any feedback, issues, or PRs. Hope someone finds it useful!



Cheers!

https://redd.it/1oxl1bn
@r_opensource