★ 868DartMIT
Nylo is an opinionated micro-framework for Flutter that imposes an MVC-ish structure (controllers, providers, events, themes, routing) on top of a standard Flutter project, plus a companion CLI (Metro) for scaffolding files. It's aimed at solo devs and small teams who want Laravel-style conventions instead of assembling their own app skeleton from scratch.
Nylo is what happens when someone decides Flutter needs Laravel's opinions. Theming, localization, a Dio networking layer with interceptors already wired, route guards for auth — all the stuff you'd otherwise assemble yourself across your first three Flutter projects is here on day one. The Metro CLI for scaffolding controllers and pages is the kind of thing that sounds minor until you've typed your fiftieth boilerplate file by hand.
What to know going in: state management is homegrown ChangeNotifier wiring, not Riverpod or Bloc, so you're opting out of the ecosystem's solutions for this one. The actual framework code lives in a separate repo from this boilerplate, so bugs and fixes mean tracking two places. And buying into Nylo's structure is a real commitment — migrating an existing app in, or back out later, isn't a weekend job.
Our full take →GitHub ↗
★ 1,597GDScriptMIT
A general-purpose Godot 4.6+ addon for building 2D grid-based metroidvania games: a visual map editor for room layout, scene linking, and minimaps, plus a persistent object-ID system for tracking collectibles and world state. Aimed at Godot developers making room/grid-based games, metroidvania or otherwise.
The Map Editor here is the standout: undo/redo, named layers, auto-detection of external map changes from git, room borders drawn right in the 2D viewport. It's the kind of in-editor tooling most Godot addons don't bother building, and it makes grid-based level design feel like a first-class Godot workflow instead of a hack layered on top.
The object ID system for persistent collectibles and switches is also genuinely useful — it replaces the usual pile of hand-rolled event flags with something structured.
Heads-up: this is a one-person project (KoBeWi), support is GitHub Discussions with no promises, and it's pinned hard to Godot 4.6+, with older projects frozen at a 1.6 tag and no path forward. The grid-room assumption is also load-bearing — if your game wants seamless, non-grid world layout, you'll be fighting this tool rather than using it.
Our full take →GitHub ↗
★ 503KotlinGPL-3.0
An Android client that bundles an end-to-end-encrypted messenger with a self-custodial crypto wallet and a light node for the Mixin Network, built in Kotlin on standard Jetpack components (Room, LiveData, Paging, ViewModel) with Hilt for DI. It's useful to study if you're interested in Android wallet architecture, Signal-protocol-style messaging, or build-integrity tooling, not as a general-purpose chat app.
This isn't really a 'try this messenger' pick — it's a 'read this build-integrity setup' pick. The Docker-based deterministic build plus a script that diffs every installed APK split against a pinned cert SHA-256, checksum-verifying even the Bundletool it depends on, is more rigor than most wallet apps bother with. The crypto paths are actually tested: Argon2, Curve25519, mnemonic derivation, TIP, Litecoin parity, all with instrumented tests. Seventy-plus Room migrations with per-version tests tell you this thing has real production mileage.
What to know going in: prebuilt closed binary blobs (mixin.aar, libwebrtc.aar) are checked into the repo, so you can verify the signed APK but can't audit or rebuild those pieces — awkward for an app whose pitch is end-to-end encryption. It's also heavy on native/NDK code across four ABIs, so forking this means NDK expertise, not just Kotlin. And despite the 'light node' language, you're still trusting Mixin's own relay infrastructure underneath everything.
Our full take →GitHub ↗
★ 4,102
A community style guide for Clojure covering formatting, naming conventions, namespace organization, and idiomatic usage patterns. It's for any team or individual writing Clojure who wants a shared reference for how code should look, and it's the de facto standard most linters and formatters (cljfmt, clj-kondo) target.
This is the rare style guide worth reading cover to cover rather than skimming for your editor's linter config. Every rule comes with the reasoning behind it, so even where you disagree you can disagree on purpose. It covers the stuff beginners actually trip on — multi-arity ordering, declare vs. restructuring, namespace aliasing — not just bracket placement, and it's been shaped by a decade of real community argument rather than one person's taste.
What to know going in: it's prose, not a spec, so you still need cljfmt or clj-kondo configured, and their rules don't map onto this guide one-to-one. A few sections (REPL rich-comment conventions, line-length rationale) read like they were written for an Emacs/CIDER workflow that newer LSP setups have mostly moved past.
Our full take →GitHub ↗
★ 228GoGPL-3.0
midgard is a self-hosted, end-to-end-encrypted clipboard sync tool: a SwiftUI menu-bar app for Mac, a Go daemon for Linux/Windows, a web page for phones (plus iOS Shortcuts), and a relay server that passes sealed blobs between your devices without ever holding the key. It's for developers who want Universal-Clipboard-style sync across mixed OSes without handing copy/paste history to a cloud vendor.
Midgard's pitch is Universal Clipboard without a cloud vendor holding your history, and the threat model actually backs that up: the relay forwards sealed blobs and buffers them only in memory until every device acks, so the server operator genuinely can't read your clipboard. QR-code device pairing handles key exchange properly instead of punting it to manual config, and there's real test coverage on the parts that matter — relay, pairing, auth hardening, end to end.
The docker-compose setup is written to join an existing Traefik network rather than assume it owns the whole box, which is the kind of detail that tells you someone has actually self-hosted this themselves.
Heads-up: the LICENSE file is just an all-rights-reserved notice despite the repo inviting PRs, which is legally murky for anyone wanting to fork it. Sign-in is hardwired to the author's own identity provider with no documented way to self-host that piece, which undercuts the self-hosting pitch a bit. It's also a one-maintainer, 228-star project, and iOS support is a web page plus Shortcuts rather than a native app.
Our full take →GitHub ↗
If you want these picks in your inbox before the page goes up, the email signup is at the bottom of the page — one email a week, no noise.