July 2, 2026

Agentic Frameworks, Tooling, Skills

• 337 Skills/Plugins Repository for Coding Agents, https://github.com/trending/python?since=daily — Curated skill system and plugin collection for Claude Code, Codex, Gemini CLI, and Cursor; enables rapid integration of custom commands, references, and workflow automation into local/CLI agent stacks without custom scaffolding. [Source: https://github.com/trending/python?since=daily]

Notable Research

• ELDR: Expert-Locality-Aware Decode Routing for PD-Disaggregated MoE Serving, https://huggingface.co/papers — Dynamic expert routing that reduces interconnect pressure in disaggregated MoE inference; provides a direct architecture for optimizing local/agentic MoE serving where KV cache and compute fragmentation typically bottleneck long-horizon tool calling. [Source: https://huggingface.co/papers]
• Domain Arithmetic: One-Shot VLA Adaptation under Environmental Shifts, https://huggingface.co/papers — Enables zero-shot or single-step vision-language-action adaptation via domain arithmetic; offers a lightweight fine-tuning alternative for Kangaroo rover/robotics vision pipelines that avoids dense domain randomization or full-weight retraining. [Source: https://huggingface.co/papers]
• ABot-M0.5: Unified Mobility-and-Manipulation World Action Model, https://huggingface.co/papers — Joint mobility-manipulation world model for simulation-free actuation planning; provides a structured baseline for rover terrain traversal and end-effector control without external physics simulators or dense reward shaping. [Source: https://huggingface.co/papers]

Frontier Lab Updates

Nothing new today.

Models to Download & Try

• deepreinforce-ai/Ornith-1.0-9B-GGUF, https://huggingface.co/deepreinforce-ai/Ornith-1.0-9B-GGUF — 9B Dense architecture; ~3.5GB GGUF quantization, leaves >28GB VRAM headroom for max context or vision modules; inherits Gemma 4 + Qwen 3.5 base capabilities optimized for fast, low-overhead agent routing and coding tasks. [Source: https://huggingface.co/models?sort=trending, https://simonwillison.net/]
• nvidia/LocateAnything-3B, https://huggingface.co/nvidia/LocateAnything-3B — 4B open-weight image-text grounding model; fits comfortably in 32GB VRAM for local vision-scraper preprocessing, rover telemetry filtering, or satellite imagery object detection without cloud inference latency. [Source: https://huggingface.co/models?sort=trending]

Skipped as Already Covered

• GPT‑5.6 Sol/Terra/Luna pricing & 30-min cache architecture
• Claude Sonnet 5 tokenizer shift & pricing adjustment
• OpenClaw email-based injection defense & 6,000-test results
• Datasette Table Actions & Multi-User Auth Plugin
• Ornith-1.0-35B-GGUF (20GB) & coding benchmarks
• Qwen3.6-27B-NVFP4 18B quantization & consumer GPU efficiency