GitHub AI Daily Trending Shakeup: vercel-labs/ui Claims Top Spot as Agent UI and Creative Tools Surge
Overtaking dsh-Go after its nine-day streak, vercel-labs/ui led daily GitHub star gains. Creative agent tooling spanning WebGPU infinite canvases and video reve
In a daily tracking snapshot of artificial intelligence repositories on GitHub published on November 15, 2026, vercel-labs/ui captured the top spot by accumulating 929 new stars over a 24-hour window. The surge displaced ccch1mneyyy/dsh-Go, a mobile DeepSeek runtime client that had held the leading position for nine consecutive days, down to second place with 918 new stars.

Image source: https://synon.co.jp/article/news/2026-11-15-github-ai-daily
The dataset, compiled by Synon using the GitHub Search API between 19:07 and 19:08 UTC on November 14, 2026, measures the net change in repository stars compared against the same time window 24 hours prior. Total star additions across the top 15 AI repositories reached 6,216, reflecting a slight dip from 7,279 on the previous day. Across the broader leaderboard, specialized developer libraries that empower autonomous software agents to render, manipulate, and inspect user interfaces and rich media demonstrated the strongest momentum.
vercel-labs/ui Overtakes dsh-Go After a Nine-Day Run
The newly crowned leader, vercel-labs/ui, describes itself as an open-source behavior library enabling software agents to assemble next-generation user interfaces (TypeScript, licensed under Apache-2.0). Having fallen outside the previous day's rankings with an estimated gain of 195 stars, it recorded 929 confirmed stars over the measurement period, lifting its cumulative total to 4,449.
Architecturally, the project focuses on headless components and accessibility primitives compliant with ARIA standards, stripping out prescriptive styling so that automated systems can reliably bind functional behaviors. It supports the Model Context Protocol (MCP) and its browser-native WebMCP variant, while integrating across frontend frameworks such as React, Vue, Svelte, Solid, and standard Web Components. This framework provides an operational baseline for AI agents that need to transition beyond textual conversational interfaces, safely triggering interactive web controls like forms, toggles, and modal dialogues.
The runner-up, ccch1mneyyy/dsh-Go, added 918 stars to bring its cumulative count to 22,660, falling short of first place by just 11 stars. Built with Flutter and Dart, the application enables developers to run DeepSeek Harness directly on mobile devices or manage remote PC interactions. Meanwhile, the official deepseek-ai/deepseek-harness repository ranked third (+607 stars, 281,042 total). Combined, these two projects in the DeepSeek Harness ecosystem accounted for 1,525 stars, representing approximately 25 percent of the total star growth recorded across the top 15 repositories.
Creative Tooling Expands Toward Infinite Canvases and Video Decomposition
A pronounced trend across the snapshot is the expansion of autonomous agents beyond straightforward code generation into graphic layout construction and automated video production.
Ranked fourth and fifth were two agent-driven visual media tools. JuneYaooo/rea-video climbed from seventh to fourth place, adding 571 stars to reach 8,476 total. Designed as an agent workflow skill, rea-video inspects filmed footage, AI-generated clips, or programmatically animated sequences frame by frame, breaking them down into structured breakdown documentation (VIDEO.md) that details reproduction instructions. In fifth place, lemomo-ai/lemo-opuscar picked up 457 stars (cumulative 16,105). Targeted at Claude Code, it supplies 99 distinct animation styles crafted purely through code routines without relying on generative video foundation models.
Entering the list for the first time at seventh place was berkantay/gpucanvas, which logged 394 stars. Created on November 12, 2026, the TypeScript repository leverages WebGPU and WGSL shaders to render thousands of dynamic DOM elements onto an infinite canvas. Beyond enabling human developers to select, drag, resize, and apply custom shader effects, the system exposes coordinate hooks allowing autonomous agents to parse canvas elements and point to interactive target nodes.
With heydamir/unslop—a Claude workflow skill that refines generic AI-generated layouts into polished production interfaces—taking eighth place (+352 stars, 15,941 total), one-third of the top 15 entries focused directly on visual interface construction and video workflows.
Diversification of Workflows and Claude Code Skills
Lower positions on the leaderboard reflected a pivot toward practical workflow orchestration rather than pure model weight execution.
Developer tooling held firm, with yetone/magpie taking sixth place (+441 stars, 36,514 total) by providing a desktop menu bar utility to swap backend models across active coding agents. Expo's cross-platform runtime exact placed ninth (+271 stars, 4,156 total), while the longstanding community prompt repository f/prompts.chat (formerly Awesome ChatGPT Prompts, created in December 2022) returned to the top ten with 252 additions, reaching 177,331 stars.
The remaining spots highlighted specialized workflow automation. Background terminal agent KKKKhazix/agent-autoloop placed eleventh (+215 stars), followed by local meeting transcription tool FrigadeHQ/yogurt in twelfth (+205 stars). Two newly indexed Claude Code extensions also gained traction: EricJujianZou/rederive, which analyzes codebases by reconstructing core architectural decisions, ranked thirteenth (+204 stars), and dansugc/reverse-engineer-marketing placed fifteenth (+198 stars) by inspecting outbound web links to uncover underlying marketing strategies.
Conversely, specialized local hardware utilities that led previous charts saw cooled interest: SSD-based inference loader wici-ai/inferssd dropped out with 35 daily additions, and shanselman/gputaskman fell below the threshold with 160 stars.
Methodology and Measurement Caveats
The figures reported in this snapshot originate from deterministic GitHub Search API queries rather than GitHub's internal Trending portal. GitHub's official Trending view balances page traffic, issue discussions, pull requests, and commit velocity under a proprietary scoring model, whereas this analysis strictly isolates star growth differentials across a 24-hour delta.
Furthermore, for repositories created within 48 hours of the sampling period, the entire cumulative star count is treated as daily growth due to the absence of a prior baseline. The 394 stars credited to gpucanvas (seventh place) reflect its total accumulated count since its November 12 creation. The analysis team also noted that 158 suspected distribution exploits and 81 duplicate repos mimicking popular agent projects were purged from the candidate pool prior to computing rankings.
Sources
- Synon Daily Tracking Report: GitHub AI Trending 2026-11-15
- vercel-labs/ui Repository: GitHub - vercel-labs/ui
- ccch1mneyyy/dsh-Go Repository: GitHub - ccch1mneyyy/dsh-Go
- berkantay/gpucanvas Repository: GitHub - berkantay/gpucanvas