The October 2026 edition of AI Tools Radar is out. It's the second monthly edition after 2026-09, built on 1 November 2026. A new edition is built automatically on the 1st of each month.
What the radar is
The radar at arsentev.ai/github selects and reviews open-source AI projects from GitHub, Hugging Face and Hacker News, one page per project. A script takes a daily snapshot of every selected project from the public GitHub and Hugging Face APIs. Removed repositories stay in the data with their status, so the sample does not drift toward survivors.
What's in edition 2026-10
There are 757 projects, selected between 11 June and 31 October 2026. The previous edition had 590. By source it's GitHub 584, Hugging Face 163 and Hacker News 10. By kind it's 594 repositories, 135 Hugging Face models and 28 Hugging Face Spaces. Selected per month: June 70, July 134, August 204, September 182, October 167.
The median GitHub star count at the moment of selection is 650.5, up from 542 in the previous edition. The top languages are Python 196, TypeScript 143, JavaScript 64, Rust 37 and Swift 24. On Hugging Face the top tasks are text-generation 46, image-text-to-text 40 and text-to-image 16.
The snapshots cover 51 days, from 12 September to 1 November 2026. At the last snapshot 579 of the 594 GitHub repositories were still available, 14 return 404 (deleted) and 1 is archived. 65.1% were pushed within the last 30 days, down from 70.4% in the previous edition.
Stars and licences
For the 415 repositories observed over the whole window, the median star gain between 12 September and 1 November is 187. The largest gains are cloudflare/security-audit-skill +52,205, deepseek-ai/deepseek-harness +46,005, DietrichGebert/ponytail +45,021, stablyai/orca +44,555 and affaan-m/ECC +37,101.
A median of 187 against a top gain of 52,205 means a handful of projects take most of the attention.
On licences, 15.5% of the repositories have no licence or one that GitHub does not recognise. The counts are MIT 308, Apache-2.0 143, NOASSERTION 52, no licence 38, AGPL-3.0 23 and GPL-3.0 10. That's roughly one in six repositories without a clear licence, so check the licence field before you build on a trending repo.
Limits and data
The sample is curated, not random. It reflects what the radar picked for people building with AI coding tools and agents, so it says nothing about GitHub as a whole. Star counts before 12 September are not observed.
The data is CC BY 4.0, in three files: items.csv, snapshots.csv and stats.json. Every number in this post comes from the build script over the CSV files.
This edition: https://doi.org/10.5281/zenodo.23852917
The series DOI always points to the latest edition: https://doi.org/10.5281/zenodo.22730450
Mirror on Hugging Face: https://huggingface.co/datasets/arsentev-ai/ai-tools-radar
Evgenii Arsentev, MD, PhD, Founder & CEO of Dad4Kids. The radar pages are at arsentev.ai/github.
AI assistance: drafted and edited with an AI tool; the data, measurements and conclusions are my own.
Top comments (0)