Models & Agents
Daily AI models, agents, and practical developments.
Daily
~15 min
Daily AI briefing covering new model releases, agent frameworks, and practical developments. From GPT and Claude to OpenClaw and Agent Zero — stay on top of the most exciting developments of our generation.
For developers building with AI, professionals adopting AI tools, and anyone who wants to stay ahead of the most transformative technology of our generation.
Latest Episode
Loading...
Recent Episodes
Ep 33: MetaComp just released the world's first dedicated AI agent governance framework built specifically for regulated financial services.
Tue, Apr 21, 2026
Ep 32: Qwen3.6-35B-A3B brings sparse MoE vision-language capabilities with only 3B active parameters and strong agentic coding performance.
Fri, Apr 17, 2026
Ep 31: Google DeepMind's Gemini Robotics-ER 1.6 upgrade delivers enhanced embodied reasoning and instrument reading for real-world robot control.
Wed, Apr 15, 2026
Ep 30: Aaron Levie declares the enterprise AI shift from chatbots to agents is now underway, moving beyond the "Chat Era."
Mon, Apr 13, 2026
Ep 29: Knowledge distillation now compresses full ensembles into single deployable models while preserving their collective intelligence.
Sat, Apr 11, 2026
Ep 28: Meta’s Muse Spark and a production-grade compiler-as-a-service approach for agents headline a day heavy on practical agent infrastructure.
Thu, Apr 09, 2026
Ep 27: Gemma 4 delivers massive gains across European languages while a 25.6M Rust model achieves 50× faster inference via hybrid attention.
Tue, Apr 07, 2026
Ep 26: AutoAgent autonomously optimizes its own harness using the same model to reach #1 on Terminal-Bench and financial modeling in under 24 hours.
Sun, Apr 05, 2026
Ep 25: Google drops Gemma 4, claiming the strongest small multimodal open model yet with dramatic gains across every benchmark compared to Gemma 3.
Fri, Apr 03, 2026
Models & Agents - Episode 24 - April 01, 2026
Wed, Apr 01, 2026
Ep 23: Alibaba Qwen just dropped Qwen3.5-Omni, a native end-to-end multimodal model built for text, audio, video, and realtime interaction.
Tue, Mar 31, 2026
Ep 22: Naver's Seoul World Model grounds video generation in real Street View geometry from over a million images and generalizes to other cities without fine-tuning.
Sun, Mar 29, 2026
About Models & Agents
Daily AI briefing covering new model releases, agent frameworks, and practical developments. From GPT and Claude to OpenClaw and Agent Zero — stay on top of the most exciting developments of our generation.
Hosted by Patrick in Vancouver.
Resources & Further Reading
AI Labs & Research
OpenAI
GPT, DALL-E, and Sora — research and blog posts from the makers of ChatGPT
openai.com/research
Anthropic
Claude model family — constitutional AI, safety research, and responsible scaling
www.anthropic.com/research
Google DeepMind
Gemini, AlphaFold, and fundamental AI research — one of the world's top AI labs
deepmind.google/research/
Meta AI (FAIR)
Llama open-source models, computer vision, and fundamental research
ai.meta.com/research/
Mistral AI
European AI lab building efficient open-weight models — Mixtral and Mistral Large
mistral.ai
arXiv AI
Latest AI research preprints — papers drop here before journal publication
arxiv.org/list/cs.AI/recent
Developer Tools & Frameworks
Hugging Face
The GitHub of ML — 500K+ models, datasets, and Spaces demos. Essential for any AI developer
huggingface.co
LangChain
Framework for building LLM-powered applications — chains, agents, and retrieval pipelines
www.langchain.com
LlamaIndex
Data framework for LLM apps — connect your data to language models with RAG pipelines
www.llamaindex.ai
Ollama
Run open-source LLMs locally — Llama, Mistral, Phi, and more on your own hardware
ollama.com
Vercel AI SDK
TypeScript toolkit for building AI-powered web applications with streaming UI
sdk.vercel.ai
Weights & Biases
ML experiment tracking, model versioning, and dataset management for AI teams
wandb.ai
Benchmarks & Leaderboards
LM Arena (Chatbot Arena)
Crowdsourced LLM rankings — users vote on which model gives better responses
lmarena.ai
Open LLM Leaderboard
Hugging Face's benchmark for open-source models — MMLU, ARC, HellaSwag scores
huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard
Papers With Code
ML papers with code implementations and state-of-the-art benchmarks across tasks
paperswithcode.com
Artificial Analysis
Compare LLM providers on speed, cost, and quality — essential for API selection
artificialanalysis.ai
Newsletters & Analysis
Latent Space
AI engineering podcast and newsletter — deep dives with builders and researchers
www.latent.space
The Decoder
Daily AI news focused on practical developments, model releases, and tools
the-decoder.com
Import AI
Jack Clark's weekly AI newsletter — policy, research, and industry analysis
importai.substack.com
The Batch (deeplearning.ai)
Andrew Ng's weekly AI newsletter — news, insights, and research highlights
www.deeplearning.ai/the-batch/
Simon Willison's Weblog
Prolific AI practitioner's blog — LLM tools, prompt engineering, and practical AI
simonwillison.net
Open-Source Ecosystem
Ollama
Run Llama 3, Mistral, Phi, and other open models locally with one command
ollama.com
LM Studio
Desktop app for running local LLMs — GUI for model discovery, download, and chat
lmstudio.ai
vLLM
High-throughput LLM serving engine — the standard for production model deployment
github.com/vllm-project/vllm
llama.cpp
Run LLMs on CPU/GPU with minimal resources — the engine behind most local AI apps
github.com/ggerganov/llama.cpp
Open WebUI
Self-hosted ChatGPT-like interface for local models — works with Ollama out of the box
github.com/open-webui/open-webui
Recommended Tools
Free tier
Claude
Anthropic's AI assistant — strong at reasoning, coding, and long-context analysis
Free tier
ChatGPT
OpenAI's conversational AI — GPT-4o with vision, code, and browsing capabilities
Free tier
Cursor
AI-first code editor — autocomplete, chat, and codebase-aware assistance built on LLMs
Free
Hugging Face
Try 500K+ models in your browser — text, image, audio, and multimodal demos
Free
Ollama
Run open-source LLMs locally — one command to download and chat with Llama, Mistral, Phi
Free tier
Perplexity
AI-powered search engine — asks follow-up questions and cites sources for every answer
Key Concepts
What is an LLM?
A Large Language Model (LLM) is an AI system trained on vast amounts of text data to understand and generate human language. Models like GPT-4, Claude, Gemini, and Llama are LLMs. They work by predicting the most likely next token (word piece) in a sequence, but this simple mechanism produces remarkably capable systems that can write code, analyze documents, reason through problems, and hold conversations.
What is an AI agent?
An AI agent is a system that uses an LLM as its 'brain' to autonomously plan and execute multi-step tasks. Unlike a simple chatbot that responds to one message at a time, an agent can break down complex goals, use tools (web search, code execution, APIs), observe results, and iterate. Examples include coding agents (Cursor, Claude Code), research agents (Perplexity), and browser agents that navigate websites on your behalf.
What is RAG?
Retrieval-Augmented Generation (RAG) is a technique that gives LLMs access to external knowledge by retrieving relevant documents before generating a response. Instead of relying solely on training data, a RAG system searches a database (using vector embeddings), finds relevant passages, and includes them in the LLM's context. This reduces hallucinations and lets you build AI that can answer questions about your own documents, codebase, or data.
What is fine-tuning?
Fine-tuning is the process of further training a pre-trained LLM on a specific dataset to specialize it for a particular task or domain. For example, you might fine-tune Llama on medical literature to create a healthcare-specific model. It's more expensive than RAG but can teach the model new behaviors, styles, or domain expertise that prompt engineering alone can't achieve. Most developers start with RAG and only fine-tune when necessary.
What is MCP (Model Context Protocol)?
MCP is an open protocol (created by Anthropic) that standardizes how AI models connect to external data sources and tools. Think of it as a USB-C port for AI — instead of building custom integrations for every tool, MCP provides a universal interface. An MCP server can expose databases, APIs, file systems, or any tool, and any MCP-compatible AI client can use them. It's rapidly becoming the standard for agent tool connectivity.
Latest from the Blog
Read in-depth articles from recent episodes.
April 21, 2026
Ep 33
7 min
MetaComp just released the world's first dedicated AI agent governance framework...
MetaComp just released the world's first dedicated AI agent governance framework built specifically for regulated financial services.
Read article →
April 17, 2026
Ep 32
7 min
Qwen3.6-35B-A3B brings sparse MoE vision-language capabilities with only 3B acti...
Qwen3.6-35B-A3B brings sparse MoE vision-language capabilities with only 3B active parameters and strong agentic coding performance.
Read article →
April 15, 2026
Ep 31
6 min
Google DeepMind's Gemini Robotics-ER 1.6 upgrade delivers enhanced embodied reas...
Google DeepMind's Gemini Robotics-ER 1.6 upgrade delivers enhanced embodied reasoning and instrument reading for real-world robot control.
Read article →Get Models & Agents in your inbox
New episode alerts delivered straight to your email — no spam, unsubscribe anytime.
More from Nerra Network
Ten daily shows keeping you informed.