Edition 04 covers 10 October to 9 November 2026 with both of the tracker’s instruments refreshed. The GitHub radar ranks new repositories by stars per day, read directly from GitHub’s API on 9 November; DeepSeek’s agent prototype and inference engine top it. The launch board logs models, agent tools and funding rounds, each tagged Verified (first-party page, repository or reputable press) or Reported (one secondary source); Google, OpenAI and GitHub’s Universe conference filled most of it.
AI Tracker · Edition 04 · 9 November 2026
DeepSeek Put Its Agent and Its Inference Engine on GitHub. They Drew 47,000 Stars in Under Four Weeks.
Published · v1.0.0 · Edition 04 · 28 new movers this window
Consistency & methodology notes
Every board item links one source. It is labelled Verified where that source is a first-party page, repository or changelog, or reputable press found by its own URL, and Reported where it rests on a single secondary source. Benchmark, usage and customer figures are the companies’ own unless a card says otherwise.
What changed this edition (window 10 October to 9 November): 28 new movers, 8 earlier items with a new development, and three carried over unchanged (Gemini agent, OpenAI’s Agents API and Copilot Autopilot). 15 items were retired because they predate the window, were superseded or had no new development: GPT-6 Astra, Grok 4.7, DeepSeek-V4.1-Flash, Qwen3.8-Flash-Next, Qwen3.8-Max (open weights), Muse personal agent, Gemini Spark, GitHub Copilot (unified), Agentforce long-horizon runtime, Temporal, AIR, OpenAI Agent Plugins, Muse Code, Cloudflare Wallets & cloudflare.pay, Cloudflare Kitesurf.
- GitHub star counts were re-read on 9 November from GitHub’s search API (Edition 03 could not read the API). Ruflo (ruvnet/ruflo) now has 81,324 stars, up from 67.6k on 11 August.
- Correction to Edition 03 — Graphify. Edition 03 gave Graphify as 16 stars and called claims of 85,000+ stars unverified. The 16 stars belong to a different, small repository (rhanka/graphify). The project the tracker follows, Graphify-Labs/graphify, is real and had 144,768 stars on 9 November.
- Moved or gone: OmniRoute now lives at cheaper-inference/OmniRoute (85,834 stars); madslorentzen/ai-job-search no longer appears on GitHub and is dropped from the watch list.
- Dates that differ between sources are shown as given, not resolved: GPT-6.1 Terra (OpenAI's launch posts 26 Oct; one model-tracking site lists 15 Oct); Kimi K3.1 Pro weights (Moonshot's post 4 Nov; one roundup says 5 Nov); Tencent Hy4 (Tencent 4 Nov; Artificial Analysis 3 Nov). Daytona's round is dated Tuesday 20 Oct by Fortune and Friday 23 Oct by TechCrunch; shown as a range. Traversal's release is 3 Nov (Business Wire); some press dates it 4 Nov.
- Mistral Large 4 active parameters: Mistral's 6 Oct post on X says 49B, while its blog and TNW give about 52B for the released model.
- Single-source or company-reported figures: Grok Bot's 800,000+ paid weekly users (SpaceX, via Business Insider) and the Siri AI waitlist removal (Yahoo Tech). Copilot Fusion Max benchmark claim; Greptile's 1% of GitHub PRs; Anthropic's 70+ platforms (and about 1 million skills published); Traversal's ORCA-bench 25.3%; Daytona's customer list.
- Tencent Hy4 is described as open source, but its weights were not out by 9 Nov; Artificial Analysis lists it as proprietary for now.
- Name-collision watch: 'Muse' is Meta's consumer agent, also its Muse Spark model family and its Muse Code tool. Grok Bot posts from the @bot account on X. xAI now trades as SpaceXAI. UiPath's Autopilot (its in-product assistant) is unrelated to Microsoft's Copilot Autopilot.
- Just outside the window, so not carded: Manus ($500M+, 8 Oct), Rein Security ($25M Series A, 8 Oct) and Vesta ($30M, 8 Oct) predate the window; Claude Managed Agents dynamic workflows went to beta on 9 Oct, one day before it. Legora ($550M Series E, legal AI) and Fal ($1.3B, inference) are not agent tooling. Cursor's cloud-agent canvases (16 Oct) were seen only on third-party changelog trackers.
- Missed by Edition 03: Claude Opus 5.5 (22 Sep), Claude Sonnet 5.5 (28 Sep), GPT-6.1 Sol (29 Sep), Claude Haiku 5.5 (7 Oct) and GPT-6 with Intelligent UI in ChatGPT (7 Oct).
- Freshness bias: the board prioritises roughly the last four to eight weeks. Carried-over items say what, if anything, changed.
GitHub radar: the fastest-rising new repositories
New repositories created since 25 September 2026 with at least 1,000 stars, ranked by stars per day since creation. That rewards a three-week-old project gaining fast over an older one sitting on a large total. Counts were read from GitHub’s API on 9 November 2026 by tools/ai_radar.py, which runs eleven searches (agent, llm, mcp, topic:ai and others), removes duplicates and leaves out repositories the tracker already follows.
| Repository | Stars | Stars a day |
|---|---|---|
| deepseek- Towards a More Autonomous Agent System | 21,589 | 2,399 9 days |
| ccch1mneyyy/ dsh-go: DeepSeek Harness in your pocket. Runs on the phone itself and can also control your computer remotely. (translated from Chinese) Unofficial: a phone port of DeepSeek’s harness by an individual developer, not by DeepSeek. | 17,112 | 1,141 15 days |
| deepseek- DeepInfer: High-performance, large-scale inference framework for DeepSeek models | 25,716 | 1,072 24 days |
| wheresryan22/ An agent skill for creating detailed, interactive isometric illustrations of real-world machines and complex s… No commits since 18 October; growth came in its first ten days. | 26,150 | 817 32 days |
| wangzan101/ An on-device AI agent for Android that operates apps and the web to get things done for you. | 7,852 | 714 11 days |
| KKKKhazix/ Lets a command-line agent take on and follow through tasks in the background, and sends a card only when it ne… (translated from Chinese) | 4,234 | 706 6 days |
| tangency- SolidCraft: an open-source parametric 3D CAD application built in pure Rust. Desktop, web, and automation thro… Check before relying on it: 6,710 open issues against 15,163 stars is unusual and may be automated filing. | 15,163 | 460 33 days |
| sarpstar/ Films made with code. Short films, long YouTube videos, ads, explainers and music videos: an AI coding agent w… | 1,800 | 300 6 days |
| amosblomqvist/ Agent harness for learning difficult things. | 6,996 | 291 24 days |
| qybaihe/ 🛡️ BLA (Blind LLM Agents) — Adversarial defense shield that protects binaries, Electron apps, JS bundles, Pyth… | 4,084 | 272 15 days |
A star is a bookmark: it measures attention, not use, and it can be bought. Creation date is not launch date, so a project built in private and opened to a waiting audience shows an inflated rate. GitHub search is not a census, so absence here is not evidence of absence.
Repositories the tracker already follows, 9 November 2026
| Project | Stars |
|---|---|
| Graphify Graphify- | 144,768 |
| Claude-Mem thedotmack/ | 115,170 |
| OmniRoute cheaper- | 85,834 |
| Ruflo ruvnet/ | 81,324 |
| Strix usestrix/ | 74,567 |
| Hallmark nutlope/ | 51,392 |
| Colibri justvugg/ | 49,238 |
| Codebase Memory MCP deusdata/ | 48,392 |
| Vibe Trading hkuds/ | 37,940 |
| OfficeCLI iofficeai/ | 33,157 |
| Grok Build xai- | 27,728 |
| OpenWiki langchain- | 17,836 |
Sources. Star/fork counts are read directly from the GitHub API. Launch, funding and model-release items are sourced per-item as linked in each card above and in the notes; see each card's “Source” link for the specific citation. Prior editions are linked from the AI Tracker hub.
About this article: Researched, written and edited by Umashankar Triplicane Dwarakanathan, with AI research assistance; every figure is meant to trace to the primary source cited. See the Editorial Policy for how sourcing, AI use and corrections work.
