★ 51,723RubyMIT
Jekyll is the long-standing Ruby static site generator that powers GitHub Pages, turning Markdown and Liquid templates into static HTML. It's aimed at developers who want a blog or docs site without a database or backend, and who are comfortable with Ruby tooling.
Jekyll earns its place by being the thing that just works for a huge number of people — GitHub Pages native support means you can get a blog live with zero config, and the docs are so thorough that the docs site itself is basically a Jekyll tutorial. A decade of stability in the plugin API means themes and tooling from years ago mostly still work.
What to know going in: build times fall off a cliff once you're past a few thousand posts, and the incremental build flag has known correctness bugs, so don't trust it blindly. Ruby's native-extension gems (sass-converter especially) are still a pain on Windows and fresh Mac installs. If you're on GitHub Pages you're also locked to an old whitelisted gem set unless you build in CI yourself. For a basic blog it's still a solid default; for anything bigger, know that Hugo and Eleventy exist for a reason.
Our full take →GitHub ↗
★ 35,231JavaMIT
SmartTube is a free, open-source YouTube client built for Android TV boxes and sticks, letting people watch without ads, without a Google account, and with SponsorBlock baked in. It's for Android TV owners comfortable sideloading apps outside the Play Store — not a general-purpose YouTube alternative and explicitly unsupported on phones or tablets.
SmartTube does one thing really well: it's a genuinely ad-free, Google-account-free YouTube client for Android TV with SponsorBlock built in as a real feature, not an afterthought. It runs on hardware with no Google Play Services at all, and the device coverage — from ancient Android 4.3 boxes to recent Shields — is wider than you'd expect from a single-maintainer project.
What to know going in, and this is a big one: the README currently discloses the maintainer's dev machine was infected with malware, which means recent builds and signing keys could be compromised. That's not a minor caveat — it's a supply-chain risk you should weigh before sideloading anything right now. Beyond that, it's reverse-engineered YouTube internals, so it breaks whenever Google changes something server-side, there's no real auto-update, and the build toolchain requires an old OpenJDK version. Worth watching, maybe not worth installing this week.
Our full take →GitHub ↗
★ 328TypeScriptMPL-2.0
ADV.JS is a Markdown-based text-adventure/visual-novel engine aimed at AI-agent-driven authoring: you describe a game, an agent (Codex/Claude Code/Cursor) scaffolds it via CLI, you polish it in a local editor, then build and deploy to Cloudflare Pages. It's for developers building narrative games or wanting to embed a lightweight dialogue runtime into an existing web game, not for non-technical writers.
The idea here is sharp: describe a game, let an agent like Claude Code or Codex scaffold it from Markdown, then polish it in a local editor. The CLI is actually built for that workflow, not retrofitted — structured JSON output, MCP install commands, the works. The compiler/runtime split also means you can embed just the lightweight runtime in an existing game instead of swallowing the whole engine, and there's a real Swift port of the engine for iOS with parity tests against the JS version, which is more cross-platform rigor than you'd expect from a 328-star project.
What to know going in: this is a sprawling monorepo for its size — engine, Electron app, native iOS app, a CloudBase backend with billing, and a mobile SaaS studio all live together, which is a lot for whatever team is behind it. The docs themselves warn that the embeddable runtime is unstable and that you should pin a verified version yourself. Browser support is Chromium-only for now. Promising, but treat it as a moving target, not a stable dependency.
Our full take →GitHub ↗
★ 1,434GoUnlicense
An instructional, from-scratch Raft implementation in Go by Eli Bendersky, built incrementally across a 5-part blog series: leader election, log replication, persistence, and finally a working key/value store with exactly-once delivery semantics. It's for developers who want to actually understand Raft by reading and running code, not for anyone who needs a production consensus library.
This is the rare project that teaches by being honest about its own structure: each of the five parts is a complete, dependency-free Go module, so you can read the leader-election code in part1 without any of the later persistence or KV-store logic leaking in. The included log-visualization tool turns test runs into an HTML timeline, which is genuinely useful for making Raft's hardest-to-debug races — split elections, stale leaders — visible instead of theoretical. And it doesn't stop at toy election logic; by part5 you've got a working KV store with exactly-once delivery semantics.
What to know going in: the code is deliberately duplicated across parts rather than factored into a shared library, so a bug fix in part2 won't propagate — you have to go patch it in parts 3 through 5 yourself if you're following along for real. There's no cluster membership change support anywhere in the series, which is one of the trickier pieces of real-world Raft. It's a teaching tool, explicitly not meant for production, and it doesn't benchmark against hashicorp/raft or etcd/raft, so you won't come away knowing exactly where it diverges from what you'd actually run.
Our full take →GitHub ↗
★ 1,596JavaApache-2.0
Siddhi is a Java-based Complex Event Processing and stream processing engine with its own SQL-like query language for matching patterns, sequences, and windows over live event streams. It targets JVM developers who need real-time analytics or anomaly detection and would rather write declarative stream queries than hand-roll windowing and pattern-matching logic.
Siddhi's query language is the real draw — sliding windows, event sequences, pattern matching, and stream joins are all built in, which saves you from hand-rolling windowing logic that's easy to get subtly wrong. It's flexible about deployment too: embed it in a JVM process, wrap it from Python, or run it standalone as a Docker/Kubernetes service. A decade of production use at WSO2 and CNCF sandbox status back up the claim that its semantics are well exercised rather than theoretical.
What to know going in: the query language is a proprietary DSL, not SQL, so you're trading lock-in and thin IDE support for that expressiveness — Flink SQL or ksqlDB will feel more familiar if that matters to you. The codebase still carries WSO2/OSGi-era module structure that shows its age, and Kubernetes-native operational tooling (autoscaling, state management at scale) lags well behind Flink or Kafka Streams. Star and fork counts are modest for a project this old, so lean on the docs and your own testing rather than expecting a big community to lean on.
Our full take →GitHub ↗
If you want this kind of rundown in your inbox every week, sign up below — no spam, just the picks and the honest take.