GitHub Trending: Zed, NVIDIA Auto-Ontology and the frameworks that dominate the week

The list of GitHub's most popular repositories this week reveals an ecosystem split between high-performance development tools, autonomous agent infrastructure, and the consolidation of major web frameworks. Unlike previous rankings where agents and generative AI dominated the top, this time the toolmakers — Zed, Prometheus, React, and Angular — took center stage, suggesting the market is shifting from experimentation to production.

What each repository does and why it's trending

1. Zed (zed-industries/zed) — Rust

~292 stars today | 93,430 total

The code editor created by the founders of Atom and Tree-sitter has finally reached maturity. What distinguishes Zed from competitors isn't just the speed promise — every new editor claims that — but the architecture: it runs on a proprietary engine called CodeEditor that decouples rendering from extension execution, allowing large file editing without freezing. The Rust language isn't a detail: it guarantees absence of data races at compile time and enables Zed to use native threads for syntax parsing without blocking the UI. The growth of 93k stars in just over two years indicates the community is still getting accustomed to the dual model (open-source core, multiplayer features as a paid service).

2. NVIDIA Auto-Ontology (NVIDIA/auto-ontology) — Python

~2,005 stars today | 8,790 total

This is the project that surged the most this week. Auto-Ontology is a framework for building semantic ontology layers over structured data, enabling AI agents to "reason" about relationships between entities instead of merely generating plausible text. The difference from traditional RAG is that while RAG retrieves relevant documents and feeds them to an LLM as context, auto-ontology builds knowledge graphs where each node is a typed entity with properties — and queries are generated deterministically via RDF/SPARQL before touching the model. NVIDIA is positioning this as a centerpiece of its enterprise agents, which explains the violent rise. In practical terms: if you need a system where "the cost of error" is high (medical diagnosis, regulatory compliance, industrial control), having an ontology graph as an intermediate layer between raw data and the LLM drastically reduces hallucinations.

3. React (react/react) — JavaScript

~244 stars today | 252,180 total

React remains the most-starred framework on GitHub. That's not new — it's the second most popular repository on the platform, behind only TypeScript. What's worth noting is that, with React Server Components as a stable API and Next.js 15+ mature as a full-stack framework, React has evolved from "the UI library" into the backbone of entire web applications. The steady star growth reflects corporate adoption: companies that migrated away from React toward "more modern" alternatives between 2022-2024 are returning, realizing that React's tooling ecosystem (Vite, TanStack, Remix as an official fork) offers the best balance between control and productivity.

4. Angular (angular/angular) — TypeScript

~128 stars today | 101,270 total

Angular, which went through one of its largest refactorings by removing Zone.js as an optional dependency and introducing reactive signals in production, is gaining new momentum. The paradigm shift is significant: instead of a framework that automatically manages the change detection cycle (and which frequently caused excessive change detection in large applications), Angular 17+ delivers a model where each component declares its reactive dependencies explicitly. This results in fewer re-renders, better debuggability, and a more predictable API. The growth to 101k stars shows that the enterprise community is convinced of the direction.

5. Ruby on Rails (rails/rails) — Ruby

~152 stars today | 58,960 total

Rails continues as one of the most active frameworks on GitHub. Recent contributions focus on Hotwire improvements (the HTML streaming library that replaces the need for complete SPAs), Asset Pipeline performance, and Ruby 3.3+ stabilization. What's interesting is that Rails has repositioned itself as the ideal framework for internal tools and MVPs: the Hotwire + importmaps combination allows building interactive interfaces without compiling JavaScript, reducing deploy and maintenance complexity. Startups that previously used Rails for their core product now use it as an admin layer, while the "pretty" frontend runs in React or Svelte separately.

6. Symfony (symfony/symfony) — PHP

~89 stars today | 31,340 total

Symfony remains the backbone of the PHP enterprise ecosystem. With the 7.x version consolidating the component system (each component usable independently), Symfony has become more flexible: you can use only the Serializer or only the HttpClient without dragging the entire kernel. The PHP community is growing — PHP 8.3 introduced refined match expressions and more expressive typed properties, significantly reducing the amount of runtime validation code needed. For those coming from the JavaScript/Python ecosystem, Symfony offers the same level of structure: dependency injection, typed controllers, and an events system that works like middleware.

7. Prometheus (prometheus/prometheus) — Go

~140 stars today | 67,090 total

Prometheus remains the gold standard for observability. What keeps it at the top isn't innovation but irreversible adoption: it's the standard for metrics in Kubernetes and cloud-native, and migrating to another system would be catastrophic for most teams. Recent contributions focused on storage efficiency (more aggressive series compaction) and PromQL improvements for complex aggregations. Prometheus's key strength is its pull model: the metrics server actively fetches data, eliminating the firewall and NAT issues that push-based systems face.

8. Google Skills (google/skills) — Python

~532 stars today | 22,770 total

Google's repository containing "agent skills" for Google products and technologies. It's a package providing structured integrations for AI agents to access Google APIs (Google Search, Google Maps, Google Drive, Gmail, etc.) through typed functions and validated schemas. The idea is that, instead of an agent needing to "read the documentation" for each API to discover how to interact, the skills provide ready-made functional descriptions — what input each function expects, which parameters are required, and what the output format is. This reduces the integration overhead of AI agents in real-world workflows.

9. mpv (mpv-player/mpv) — C

~36 stars today | 37,830 total

The most popular open-source video player continues receiving codec updates and performance improvements. mpv is often the engine behind popular GUI applications like VLC (yes, VLC uses mpv as a backend for certain decoding operations) and Stremio. The reason it's trending is recent support for hardware decoding on NVIDIA/AMD GPUs via VA-API, significantly improving 4K/HDR playback on modest hardware. For Linux users, mpv remains the reference in image quality and memory consumption.

10. Rust (rust-lang/rust) — Rust

~186 stars today | 121,340 total

The Rust compiler keeps growing. Recent contributions focus on borrow checker optimizations (which can now infer more complex borrows without requiring explicit annotations from the programmer) and improvements to the macro system (which now supports conditional metaprogramming without expanding code at compile time). Rust has consolidated not only as a systems language but as a language for tooling: Zed, Starling (game framework), and much of the modern development tooling ecosystem are written in Rust for one reason: compile time is predictable and there's no garbage collector.

What the set reveals

Three observations that recur in this week's data:

1. The editor war is reaching maturity. With Zed leading in growth, Neovim consolidated, and Cursor paving the way for LLM-based editors, the code editor has evolved from a peripheral tool to the central interface of the development workflow. Whoever controls the editor controls the experience of whoever writes code. The fact that two of the three main contenders (Zed and Neovim) are written in Rust isn't coincidence — it's a statement that performance is the new selling point.

2. Autonomous agent infrastructure is professionalizing. NVIDIA's Auto-Ontology and Google Skills represent two complementary approaches to the same problem: how to make AI agents do things in the real world instead of just chatting. One builds knowledge graphs for structured reasoning; the other provides typed wrappers for existing APIs. Together, they suggest that the next wave of AI won't be about bigger models but about better integration systems.

3. Traditional web frameworks are in consolidation phase. React, Angular, Rails, and Symfony — none are "rising" in terms of aggressive growth. They're at the top because they're safe choices: companies needing long-term support, a mature ecosystem, and qualified teams. The creative energy of web development isn't in the main frameworks anymore, but in the layers on top of them: TanStack for React, Hotwire for Rails, Symfony Components for PHP. The frameworks have become solid ground — innovation happens on what you build on top.

The question: in a world where the code editor is the central place in development, and AI agents are integrating into the workflow, what will the developer's role be in two years? Perhaps the answer lies in Rust: languages that give total control over memory and concurrency are the ones that adapt best to complex integration scenarios — and those who write in them are writing the tools that agents will use.

Sources: GitHub Trending, Zed Official, NVIDIA Auto-Ontology, Google Skills

✓ Independent sources cross-checked and verified before publishing