Last two months
AI software created in the last two months that is still moving and rising fast. Each entry says what the repo is, what it does, and who it is for.
The top three keep more context so you can see why they are rising and where they fit.
Rebuild the object in a reference image as a code-only, procedural, quality-gated, animation-ready Three.js model. Token-efficient image-to-3D.
img2threejs rebuilds a reference image as reviewable, procedural Three.js code rather than a mesh, texture pack, or photogrammetry dump. The repo is a Python pipeline with quality gates, biased toward token-efficient, animation-ready output. Use it when an agent should ship a 3D object you can still diff.
A 2.78-trillion-parameter Kimi K3 running inference on a single CPU in 8.24 GB of RAM. Portable C99: no BLAS, no framework, no GPU.
A portable C99 runtime for Kimi K3 inference: no BLAS, no framework, no GPU. It claims a 2.78T MoE can run on one CPU in about 8GB of RAM. The repo is the engine and build files, not a weight dump. Use it to test whether a huge model can land on consumer hardware.
Opinionated Oxlint rules for rejecting low-evidence TypeScript and JavaScript patterns
anti-slop is an opinionated Oxlint ruleset that rejects low-evidence, low-signal TypeScript and JavaScript. It is a vendored rules pack, not another formatter. Use it when agent-generated filler is flooding the tree and CI should stop it first.
On 2026-09-05, GitHub Search results for created:>2026-07-07, pushed:>2026-07-07, stars:1000..80000 and topics such as llm, ai-agents, mcp, claude-code, codex, openclaw, and skill were merged and deduplicated. Only AI software, workflows, or structured skills created in the last 60 days, pushed in 60 days, and containing a real source tree were kept. Forks, archives, awesome lists, courses, prompt dumps, and model-weight repositories were excluded. At most 2 items share an owner and 8 share a category. velocity = stars / max(age_days, 21) to 3 decimals; star_delta is filled only for repositories that also appeared in the previous snapshot.
Included repositories were created in the last 60 days, pushed in the last 60 days, have 1,000 to 80,000 stars, and are ranked by rising velocity.
Awesome lists, reading lists, courses, forks, archived repositories, pure documentation, model-weight dumps, and projects without runnable software are excluded.
An administrator-authorized agent collects and submits the full dataset. Every publication is one indivisible 30-item snapshot.