The 5 coolest open-source projects I've discovered in 2026

There are hundreds of open-source apps out there, especially now in the era of vibe coding—it is actually becoming a bit of a problem.

However, there are still more conventional open-source projects out there, and some of them are quite interesting. Here are a few of my favorite open source projects from this year.

Ladybird

An independent browser engine built from the ground up

Credit: Ladybird

Credit: Ladybird

Credit: Ladybird

Credit: Ladybird

Credit: Ladybird

Close Credit: Ladybird

Credit: Ladybird

Credit: Ladybird

Credit: Ladybird

Credit: Ladybird

Chromium, which powers Google Chrome, Microsoft Edge, Brave, and more, largely dominates the browser space. There are really only two major alternatives: Firefox and Safari.

That is why it is so interesting to see a brand-new browser appear.

Ladybird is an open-source browser that does something pretty unusual in this day and age: it has its own rendering and JavaScript engines. It doesn't rely on the Blink, WebKit, or Gecko lineage, making it the first new engine to emerge in more than 10 years. It isn't just a vibe-coded passion project either, it has attracted serious sponsors, including Shopify, Cloudflare, Proton, Jetbrains, and other groups and individuals.

The movement in 2026 has been fairly impressive for such an ambitious project. We've seen file downloads, browsing history, and sandboxing-by-default (a critical security feature), and more added, and an increasing amount of the codebase—which was previously written in C++—has been migrated to Rust.

Related 9 Rust apps that are faster than the Linux tools they replace They can even replace classic Linux tools and let you play old PC games.

If you want, you can give Ladybird a try by pulling the repo and building it, but it isn't ready for mainstream use yet.

Lightpanda

A headless browser built for AI agents and automation

Credit: Lightpanda

If you are building scraping pipelines or AI agent workflows, you've probably run into the issue of headless Chrome being an unrepentant memory hog.

Lightpanda aims to address that problem by stripping out all the human-centric features a browser contains and only contains the parts that an AI actually needs. Their own benchmarks suggest it can be up to 11x faster (a claim I didn't rigorously test) and uses significantly less RAM. My own testing suggests that the RAM claim is true—Lightpanda only used up a few hundred megabytes, while Chrome regularly consumes several gigabytes.

As an added bonus, existing Playwright scripts should work with minimal tweaking, though it is still a work in progress, so some instability is to be expected.

If you're interested in optimizing your performance, it is a good option to try.

Lightpanda on Windows uses WSL.

Ollama

The simplest way to run open LLMs on your own machine

Credit: Corbin Davenport / How-To Geek / Ollama

Local AI models are increasingly useful, and I now use a few dozen different AI-powered applications on my desktop PC.

It used to be that getting an AI server up and running on your PC was a bit inconvenient, but Ollama makes it almost trivially easy. All you need to do is download and install Ollama, then you can download and run dozens of different models with a single command.

You should use Ollama if you need a private assistant for sensitive documents, if you're prototyping AI features offline, or if you have custom vibe-coded apps that can leverage AI features, like image cognition. Just keep in mind that you'll need decent RAM or VRAM to get the most out of it.

Open WebUI

A self-hosted front end for local models

Ollama handles the stuff behind the scenes and is basically "the brain." Open WebUI is basically the opposite—it provides a convenient interface that gives you chat access to all of your local models.

It comes with everything you actually need: retrieval augmented generation (RAG) support, a prompt library, and multi-user accounts.

It can also be connected to cloud-based models if you want to do that, though you lose any privacy benefits.

Continue

An open-source AI coding assistant you control

Continue provides a Copilot-style experience in VS Code and JetBrains, but with one major difference: you can choose the model. In the case of a local hoster, that means you connect it directly to your Ollama instance.

I've been using it with the Qwen models specifically.

Continue is helpful when you want AI assistance, but for one reason or another, you can't send code or data to a third party like OpenAI or Anthropic.

It takes a bit more work to get running, but once it does, it can actually be pretty competent. Just don't try treating it like Fable 5—it should be used for less ambitious things instead.

Open-source is better than it ever has been

If you're feeling overwhelmed by the number of cloud-based subscriptions you have, there has never been a better time to try self-hosted alternatives than there is today. Ollama and Open WebUI provide convenient AI access, and there are dozens of great apps out there that can be used to replace commercial services.

I'd recommend starting simple. FInd something simple that you've always wanted to self-host, like a note-taking app or Ollama, and start there.

Once you get started, the follow-up apps get easier and easier.