Space Luisa Fernanda
Space, astronomy, rockets and the physics of it all.
Tuesday, September 29, 2026 · UTC
The Sky & Telescopes

Four AI repositories gained over 4000 combined stars in 24 hours on GitHub

All four machine learning and decision model repositories recorded large daily gains on 24 September 2026.

Arjun Nair
Astronomy Editor
Share on X
Stands on 4 first-party readings of the GitHub activity record, not reporting.
Four public open source repositories recorded large one-day star gains on 24 September 2026. By this paper's arithmetic, the four added a combined 4006 stars over the preceding 24 hours. GitHub data shows NandhaKishorM/laya rose 2742 stars to 22,650 [4]. jev-chat/jev-chat-jarvis gained 781 stars to reach 5,957 [1]. mizorewww/laya-mlx added 307 stars to hit 6,181 [3]. zai-org/ZCode increased 176 stars to 6,691 [2]. All four repositories are tagged for machine learning, local inference or decision modelling. No note explaining the coordinated gain was present on any repository page at the time of reading.
Proof4 sources · 1 publisher · signed
What this stands on
  1. jev-chat/jev-chat-jarvis stars: 5,957 stars, Kotlin: 装在手机上的对话副驾：在 QQ / X / 飞书里读懂对方、给出候选回复、一键填入输入框，发不发由你。非侵入，只读屏幕，不 hook 不改包。 [accessibility-service, android, chat-assistant, llm, qq] as of 2026-09-24 (was 5,176 on 2026-09-23). · DRM3 GitHub activity the fact record
  2. zai-org/ZCode stars: 6,691 stars, TypeScript: Z.ai's coding agent harness. Powerful, intelligent, extensible. as of 2026-09-24 (was 6,515 on 2026-09-23). · DRM3 GitHub activity the fact record
  3. mizorewww/laya-mlx stars: 6,181 stars, Python: Native MLX runtime for Laya typed decision models - 7-14 ms short decisions on M3 Max. No text generation, PyTorch, or cloud API. [apple-silicon, decision-model, inference, laya, local-ai, machine-learning, mlx, modernbert, system-one, typed-decisions] as of 2026-09-24 (was 5,874 on 2026-09-23). · DRM3 GitHub activity the fact record
  4. NandhaKishorM/laya stars: 22,650 stars, Python: Non-autoregressive System 1 decision engine. Typed choice, score and yes/no decisions over any text in a single forward pass, in 100+ languages, with a router t [calibration, classification, decision-model, huggingface, jev, modernbert, multilingual, nlp, python, pytorch, routing, ] as of 2026-09-24 (was 19,908 on 2026-09-23). · DRM3 GitHub activity the fact record
Read directly from the GitHub activity record, not from anyone's report of it.