Sam Witteveen - Videos
Back to Channel
Kimi K2.5- The Agent Swarm
In this video, I look at Kimi K2.5 the latest model from Moonshot AI and how it crushes with Agent Swarm to do tasks Site: Blog: https://www.kimi.com/blog/kimi-k2-5.html HF: https://huggingface.c...
Ollama Launch + Claude Code + GLM Flash
In this video, I look at using Claude Code with Ollama's new function called Ollama Launch along with the GLM 4.7 Flash model. Blog: https://ollama.com/blog/launch HF: https://huggingface.co/zai-...
Clone ANY Voice for Free — Qwen Just Changed Everything
In this video, we cover the new QWEN TTS open models and look at how you can do things like Voice Design & Voice Cloning with them. Blog: https://qwen.ai/blog?id=qwen3tts-0115 Colab Basic: https:...
Beating Cowork with Open Source Cowork
In this video, we look at an open-source alternative to Claude Co-Work and the interesting story that's around it. Github: https://github.com/eigent-ai/eigent Eigent AI: https://www.eigent.ai/ Ei...
Open Responses - The NEW Standard API for Open Models
In this video, I look at the Open Responses Standard that's been released by OpenAI to support open models with their Responses SDK Site: https://www.openresponses.org/ HF: https://huggingface.co/...
Qwen3 Multimodal Embeddings: Finally, RAG That Sees
In this video, I look at the recent release of Qwen3 VL embeddings and re-rankers and look at how multimodal embeddings work, including a code example. Blog: https://qwen.ai/blog?id=qwen3-vl-embe...
Google's New Universal Commerce Protocol
In this video, we look at Google's New Universal Commerce Protocol for enabling agentic commerce. What it is, how it works, and some of the places that you might end up seeing it being used. Blog:...
MiroThinker 1.5 - The 30B That Outperforms 1T Models
In this video, I look at Miro Thinker 1.5, a new model that is only 30B with 3B (based on Qwen3) active yet can do two calls out to 400 tool calls. Blog: https://research.miromind.ai/blog/introdu...
NVIDIA's 13 New Models
In this video, I go through the new models that NVIDIA announced at CES 2026. Blog - https://blogs.nvidia.com/blog/open-models-data-tools-accelerate-ai/ CES Special Presentation Wrap Up - https:/...
FunctionGemma - Function Calling at the Edge
In this video, I look at the latest release from the Gemma team, FunctionGemma, which is all about having a super-small model that you can use to do function calling at the edge. In this video, we...
Gemini 3 Flash - Your Daily Workhorse Upgraded
In this video, I go through the latest Gemini 3 Flash model from Google. With a bump in intelligence, and yet still fast and cheap, this model is set to become many people's daily workhorse. Blo...
The Gemini Interactions API
In this video, we look at the new interactions API from the Gemini API team, and how you can use it to build and do various tasks with not only Gemini models but also agents. Blog: https://blog.g...
Junie The Anti-Vibe Coding IDE
In this video, I look at Junie from JetBrains a coding agent that is Anti vibe coding but is like a professional pair programming partner. It can run in the various JetBrains IDEs that are customiz...
Mistral 3: Europe's Answer to DeepSeek or Too Little, Too Late?
In this video, we look at the latest release from Mistral, a new Mistral-3 large model and 3 Minstral models. Blog: https://mistral.ai/news/mistral-3 HF: https://huggingface.co/mistralai/collecti...
Nano Banana Pro has arrived!!
In this video, I go through the new Nano Banana Pro aka Gemini 3 Pro Image Blog: https://blog.google/technology/developers/gemini-3-pro-image-developers/ https://blog.google/technology/ai/nano-ban...
Antigravity Google's Cursor Killer
In this video, I go through Antigravity the new Agentic Coding App from Google Blog: https://antigravity.google/blog/introducing-google-antigravity For more tutorials on using LLMs and building a...
Gemini 3 Pro - The Model You've Been Waiting For
In this video, I go through Gemini 3 Pro, the model that people have been waiting for. Blog: https://blog.google/products/gemini/gemini-3 For more tutorials on using LLMs and building agents, che...
Gemini RAG - File Search Tool
In this video, I go through the latest release from the Gemini API team, which is the file search tool, or basically a simple RAG system built right into the Gemini API Blog: https://blog.google/t...
NEW Top Open Model - Kimi K2 Thinking
In this video, I look at Kimi K2 Thinking from Moonshot AI, the most recent fully open reasoning model that scores higher than GPT-5 and Anthropic for multiple benchmarks. Blog: https://moonshota...
AgentHQ by Github
In this video, I cover AgentHQ which was launch by Github at the yearly conference and which had the goal of managing all your agents AgentHQ: https://github.blog/news-insights/company-news/welcom...
LangChain Reaches 1.0 - Whats new?
In this video, I cover LangChain's recent announcements about raising money at over a $1.25 valuation and the launch of LangChain and LangGraph 1.0. LangChain Is now a Unicorn: https://blog.langch...
Is Meta killing FAIR?
In this video, I go through some of the recent news about Meta laying off 600 people that worked for Facebook AI Research. Article: https://www.axios.com/2025/10/22/meta-superintelligence-tbd-ai-...
ChatGPT Atlas - The Battle for your Browser
In this video we look at OpenAI's new browser Atlas and what it can do. Blog: https://chatgpt.com/atlas Download: https://chatgpt.com/atlas/get-started/ For more tutorials on using LLMs and build...
DeepSeek OCR - More than OCR
In this video, I look at DeepSeek OCR and show that it's an experiment in using images to compress text representations better. DeepSeek OCR Paper: https://github.com/deepseek-ai/DeepSeek-OCR/blo...
Claude Skills - SOPs For Agents
In this video, I look at a new announcement from Anthropic called Claude Skills, but also more generally at the concept of how frontier labs are creating standard operating procedures for agents to...
Haiku 4.5 - Small Beats Big
In this video, I look at the latest model from Anthropic, Claude Haiku 4.5, and see how it stacks up both in intelligence and speed to its bigger brother and its previous versions. Blog: https://...
OpenAI's Agent Builder
In this video, I go through OpenAI's Agent Builder. Blog for Agent Kit: https://openai.com/index/introducing-agentkit/ For more tutorials on using LLMs and building agents, check out my Patreon ...
OpenAI DevDay 2025 - What Hit What Missed
In this video, I go through the key announcements from the OpenAI DevDay keynote. Blog for Agent Kit: https://openai.com/index/introducing-agentkit/ For more tutorials on using LLMs and building...
Sora 2 - OpenAI's TikTok
In this video I look at the release SORA 2 and how it goes beyond just being a model, but is also an app and a social network. Blog: https://openai.com/index/sora-2/ SORA 2 feed philosophy: https...
... there's more to Sonnet 4.5
In this video I look at the release of Sonnet 4.5 in context of Anthropic's plans for the Virtual Collaborator. Blog: https://www.anthropic.com/news/claude-sonnet-4-5 Devin blog: https://cognitio...
Meta's Code World Model
In this video I look at some new research out of Meta which is a code world model and is basically an LLM trained in a different way to try and get it to understand the tokens that it's generating....
The Qwen Avalanche
Blog: https://qwen.ai/research For more tutorials on using LLMs and building agents, check out my Patreon Patreon: https://www.patreon.com/SamWitteveen Twitter: https://x.com/Sam_Witteveen 🕵️ Int...
Google's NEW Agent Money Protocol
In this video, I look at A2P, Google's new agent payments protocol. We look at some of the key facts behind it, where it sits in comparison to A2A and MCP, and how you can get started. Blog: htt...
Qwen3 Next - Behind the Curtain
In this video I go through Quentin3 Next and how it tests out a bunch of ideas. In this video, I go through Qwen3-Next and how it tests. Blog: https://qwen.ai/blog?id=4074cca80393150c248e508aa629...
Kimi K2 0905 for Agents
In this video, I look at how the Kimi K2-0905 model has been updated to be better for agentic and tool calling use cases. Model Card:https://huggingface.co/moonshotai/Kimi-K2-Instruct-0905 Colab...
EmbeddingGemma - Micro Embeddings for Mobile Devices
In this video, I look at the latest Gemma release which is EmbeddingGemma, a 300M model that can be used for doing embedding tasks like RAG and semantic similarity on phones and mobile devices at t...
The Future of AI Coding with Aja Hammerly
Recently at Google I/O Connect China, I sat down with Aja Hammerly, and we talked about the future of AI coding, how things are evolving and tips for getting better results. Firebase Studio: http...
Gemini 2.5 Flash Image is Nano Banana!!
In this video, I go through the latest Gemini 2.5 flash image model (also known as Nano Banana) and show what it can do when you combine reasoning and conversational input for really good image gen...
GPT 5 - What They Didn't Say
In this video, I look at the launch of GPT-5 and what we can work out about the system that they have released. Blog: https://openai.com/index/introducing-gpt-5/ For more tutorials on using LLMs...
OpenAI's New OPEN Models - GPT-OSS 120B & 20B
Blog: https://openai.com/index/introducing-gpt-oss/ Colab: https://dripl.ink/BLrkZ For more tutorials on using LLMs and building agents, check out my Patreon Patreon: https://www.patreon.com/SamW...
LangExtract - Google's New Library for NLP Tasks
In this video, I look at LangExtract, a library from Google that allows you to do old-world natural language processing tasks with ease using LLMs and structured outputs. Blog: https://developers...
Gemini Deep Think
In this video, we look at the latest Gemini release, Gemini DeepThink, and see what it can be used for and how it was able to reach gold medal standard in the International Math Olympiad. Blog: ht...
Ollama Gets a New App
To celebrate Ollama's 2nd birthday the cute llamas have got a new app!! Blog: https://ollama.com/blog/new-app For more tutorials on using LLMs and building agents, check out my Patreon Patreon: ...
Opal - Google Labs Killer NEW App
In this video, I look at the latest release from Google Labs, which is a new app called Opal. Opal allows you to create LLM and generative AI workflows using a drag-and-drop and description system....
SmolLMv3 - A Small Reasoner with Tool Use.
Blog: https://huggingface.co/blog/smollm3 Colab: https://dripl.ink/oFvSw For more tutorials on using LLMs and building agents, check out my Patreon Patreon: https://www.patreon.com/SamWitteveen Tw...
Kyutai STT & TTS - A Perfect Local Voice Solution?
Blog: https://kyutai.org/next/stt Blog: https://kyutai.org/next/tts GitHub: https://github.com/kyutai-labs/delayed-streams-modeling Colab: https://dripl.ink/QZevZ For more tutorials on using LLM...
GeminiCLI - The Deep Dive with MCPs
This time I do a deep dive into Gemini CLI and look at how you can use tools with it and how you can use MCPs with it to make both your development faster but also to be able to do other tasks beyo...
Introducing Gemini CLI
Blog: https://blog.google/technology/developers/introducing-gemini-cli-open-source-ai-agent/ GitHub: https://github.com/google-gemini/gemini-cli/tree/main For more tutorials on using LLMs and buil...
NanoNets OCR-s
Blog: https://nanonets.com/research/nanonets-ocr-s/ Colab: https://dripl.ink/YQEpC For more tutorials on using LLMs and building agents, check out my Patreon Patreon: https://www.patreon.com/SamWi...
Qwen 3 Embeddings & Rerankers
In this video I look at the new release from Qwen of their new Embedding and Reranking models which are start of the art and most importantly open weights models. Blog: https://qwenlm.github.io/bl...