OpenWorker Harness SDK puts the software around an AI model in focus
Andrew Ng’s November 13 edition of The Batch introduces an open agent harness. Context, tools and permissions are the central engineering questions.
THE SIGNAL FEED
LLM.
Topic feed
From model selection to agent skills: local inference, fine-tuning and verified task-specific configurations. Recipes and documentation ↘
llama.cpp 1.0 sets clearer compatibility boundaries for local LLM software
ggml-org has released version 1.0.0 of the llama.cpp and ggml libraries. Public-interface versioning is the central change for application integrators.
GitHub shares a practical guide to AI agent skills
Repeatable engineering workflows can be captured in files and reused.
Intel and RadixArk collaborate on open SGLang inference
The collaboration targets the software stack used to deploy models on Intel platforms.
Claude Fable 5.2 and Mythos 5.2 target demanding AI work
Anthropic introduces models for coding, science and complex technical tasks.
Microsoft brings its multi-model Copilot harness to Foundry
The announcement extends a shared agent workflow approach through a managed developer platform.
AI illustrationAnthropic donates Agent Skills to the Agentic AI Foundation
An open standard for agent skills moves to an industry foundation.
AI illustrationClef on Ollama shows what a task-specific decision model looks like
Cloudflare’s multimodal model evaluates typed questions and candidate outcomes. Its specialization goes beyond supplying a different system prompt.
No stories in this direction yet. Official documentation and recipes are available below.
Shape a model around your task
Skills and system instructions define a workflow; RAG supplies external context; LoRA and fine-tuning change model parameters. These are different adaptation methods: choose for the task and evaluate the result.
MODELS
Task-specific models ↗
Releases, licenses and model capabilities.
Official resources ↗LOCALRun locally ↗
Ollama, llama.cpp, GGUF and quantization.
Official resources ↗FINETUNINGLoRA & fine-tuning ↗
Adapters, datasets and training recipes.
Official resources ↗SKILLSSkills & agents ↗
Instructions, tools and specialized workflows.
Official resources ↗RAGRAG & memory ↗
Bring your documents and context into an application.
Official resources ↗EVALUATIONEvaluation & serving ↗
Evaluation, performance, cost and deployment.
Official resources ↗