Daily · star velocity · 2026-10-28

rea stays on top as the 28 October edition ranks the settled 27 October window

The 28 October edition moves to the complete 27 October window: rea leads again, hallmark and ai-design-skills argue about taste in the middle, DeepTutor and elephant-goldfish-model arrive, and LinkSwift, dsh-our-free-model and VoiceStudio hold their spots.

Fastest-growing this period

Counts come from GitHub's official stargazers/history endpoint: the stars gained on each day of the window. Project briefs and the reading notes below are editorial additions.

1

morluto/rea : Point your agent at any app and learn how it works
TypeScript
A toolkit that lets your coding agent reverse engineer software without source code: it inspects native binaries through Hopper or Ghidra, JavaScript and Electron apps, .NET assemblies and websites, then explains how a feature works with the evidence and the limits behind each conclusion. Analysis runs locally, and the same tools work straight from your terminal.

+5,579
2

Gathin23/cleanyourtv : Debloat any Android TV over Wi-Fi and swap in a clean launcher
Python
A kit that turns an ad-filled Android TV home screen into a calm, Apple TV style one. Over the same Wi-Fi, its ADB scripts back up the TV, disable telemetry, sponsored rows and preloaded apps without uninstalling anything, then install a tiny launcher with a dock of four apps. A restore script undoes every step, and tested lists ship for Xiaomi, TCL, Sony and more.

+2,442
3

Nutlope/hallmark : A design skill that lints pages for the tells of machine-made UI
CSS
A design skill for Claude Code, Codex, Cursor and OpenCode that builds pages which refuse to look machine-made. It deals three themes from a catalog of forty-eight materials, then checks the result in a real browser against the tells of generated UI, from the eyebrow above every heading to a sign-up form at the end of every page. A fresh-context reviewer gets the last word.

+2,338
4

HKUDS/DeepTutor : A private tutor that reads your own books, videos and notes
Python
A self-hosted learning workspace from the HKU group behind LightRAG. It chats you through a subject, generates quizzes and mastery drills, answers questions about YouTube or Bilibili videos down to the timestamp, and builds knowledge bases on GraphRAG or LightRAG. Memory is layered and editable, so its picture of you stays visible, and you bring your own model provider.

+2,254
5

Ebony-Vinyl/dsh-our-free-model : A dsh plugin that puts frontier models one install away, no login or API key
JavaScript
A plugin for the DeepSeek Harness shell: install it and the model picker fills with free frontier models like DeepSeek V4.1 Flash and Kimi K3, no login or key needed. Free lanes come straight from OpenCode's Zen and Kilo AI's public gateways, re-pulled every refresh. A local OpenAI-compatible relay port lets other tools call the same models, and usage stats stay on the machine.

+2,167
6

hmjz100/LinkSwift : A userscript that pulls real download links out of nine Chinese cloud drives
JavaScript
A Tampermonkey script forked from a popular download helper and rebuilt to keep working as Chinese cloud drives change. It extracts real direct links from nine services including Baidu, Aliyun, Quark and 123 Cloud, hands them to the downloader you choose, and adds UI tweaks plus companion scripts for membership features. Which drives still work varies as the sites update.

+2,061
7

debpalash/VoiceStudio : Try voice design, transcription and dubbing in one workspace
Python
A local audio workspace for voice design, transcription, dubbing and longer narrated projects. Useful for trying several speech workflows with your own material in one application, with engine choice and hardware determining the supported languages and output quality.

+1,983
8

unclebob/ub-mode : Make it work, then make it right, with gates that score the code

An agent skill that pairs software-craft rules with deterministic quality gates. TDD, root-cause fixes and small ordered commits sit on one side; companion tools score cyclomatic complexity, structural duplication and mutation-test survivors on the other, so your agent refactors against measured thresholds instead of taste.

+1,871
9

elayadesign/ai-design-skills : Design rules your AI coding tool can actually follow

A collection of design skills for AI coding tools such as Claude Code, Cursor and Codex. Drop a skill's rules file into your tool and it builds a landing page with a full visual system, or turns a static design into a pixel-exact animated showcase exported as MP4 and GIF. The design values sit in one editable section, so restyling a skill means changing a few lines.

+1,798
10

google/elephant-goldfish-model : Draft the spec with full context, then attack it with none
Python
An open-source Google plugin for design-before-code work. A deep-context agent drafts the specification for a feature you are about to build, then zero-context reviewer subagents read that draft cold and challenge its unstated assumptions, and you make the trade-offs and approve before any code exists. It installs into Claude Code, Codex or Antigravity.

+1,748

Ranking window: 2026-10-27 UTC, full day.

What I would look at first

Reading notes updated 2026-10-28; the ranking above preserves its original reporting window.

This edition moves to the fully settled 27 October window. morluto/rea still leads by a wide margin: hand your coding agent a binary, a website, an Electron app or a .NET assembly and it comes back with a readable reconstruction where every claim carries its evidence and limits. Gathin23/cleanyourtv holds second, stripping sponsor rows, telemetry and ads off an Android TV or Google TV on your own network and swapping in a launcher you can roll back from backup.

The middle of the chart is an argument about taste. Nutlope/hallmark audits rendered pages in a real browser for the tells of generated UI, while elayadesign/ai-design-skills writes the design rules your agent must follow. google/elephant-goldfish-model moves the fight upstream: a deep-context agent drafts the spec for a feature, then zero-context reviewers read that draft cold and poke at the assumptions it never wrote down. unclebob/ub-mode is discipline of a different kind, pairing TDD and root-cause habits with tools that score complexity, duplication and mutation-test survivors. HKUDS/DeepTutor points agents at learning instead, tutoring from your own books and videos with timestamp-grounded answers and editable memory.

The practical picks barely moved. hmjz100/LinkSwift keeps pulling real download links out of nine Chinese cloud drives, Ebony-Vinyl/dsh-our-free-model still fills the DeepSeek Harness picker with free frontier models through public gateways, and debpalash/VoiceStudio remains the local choice for voice cloning and dubbing. If you only open two: a bloated television points to cleanyourtv, and an agent whose output you cannot verify points to elephant-goldfish-model.

rea stays on top as the 28 October edition ranks the settled 27 October window

Fastest-growing this period

morluto/rea : Point your agent at any app and learn how it works

Gathin23/cleanyourtv : Debloat any Android TV over Wi-Fi and swap in a clean launcher

Nutlope/hallmark : A design skill that lints pages for the tells of machine-made UI

HKUDS/DeepTutor : A private tutor that reads your own books, videos and notes

Ebony-Vinyl/dsh-our-free-model : A dsh plugin that puts frontier models one install away, no login or API key

hmjz100/LinkSwift : A userscript that pulls real download links out of nine Chinese cloud drives

debpalash/VoiceStudio : Try voice design, transcription and dubbing in one workspace

unclebob/ub-mode : Make it work, then make it right, with gates that score the code

elayadesign/ai-design-skills : Design rules your AI coding tool can actually follow

google/elephant-goldfish-model : Draft the spec with full context, then attack it with none

What I would look at first