Videos
Find videos about LLM models, hardware, runtimes, benchmarks and practical AI workflows.
Breakdown of Fable 5 covering pricing, misinformation versus Mythos 5, use cases, token usage, safety constraints, coding benchmarks, hands-on testing of guardrails and a 3D game clone.
Demonstrates building and deploying an AI agent using Google's Agents CLI with Claude Code, featuring identity management, a locked-down sandbox, and a full audit trail for production-grade deployment
Compares Odysseus using SearXNG and Brave against Perplexity, plus discusses Claude Fable 5 (Mythos) versus Opus 4.8 and GPT 5.5.
Locally installs and tests DiffusionGemma, a generative model built by Google DeepMind, using the google/diffusiongemma-26B-A4B-it variant hosted on Hugging Face.
Evaluates Claude Fable 5 via multimodal coding, Windows XP-style app generation, 3D model creation, RF and microcontroller reasoning, and a physical model print test.
Evaluates Claude Fable 5 versus GPT-5.5 via a four-phase progressive coding challenge involving a Windows-style UI, browser, physics sandbox, and pseudo-3D racing game on default settings.
Compares Claude Opus 4.8 and Fable 5 using a Bevy Hantavirus Simulation test prompt to evaluate planning, implementation builds, and sudden refusal behavior against double token costs.
Demonstrates setting up a RAG system using OpenClaw, Qdrant, and RAGwire via prompts, then compares Apple and Google Q1 2026 earnings through ingested 10-Q PDFs.
Tests compare Google gemma-4-26B-A4B-it-qat-q4_0-gguf against unsloth gemma-4-26B-A4B-it-GGUF using reasoning, agency, coding, and memory evaluations on a system with 16GB VRAM and 32GB DDR4 RAM.
Compares inference speed of QWEN3.6-35B-A3B with and without Multi-Token Prediction on Apple M5 Max using Pi Coding Agent and LM Studio.
Compares Google QAT and Unsloth QAT + MTP using identical QAT base weights and MTP setup to evaluate performance differences where the quantization method is the sole variable.
Demonstrates running Gemma 4 QAT models via llama.cpp with --n-cpu-moe MoE offloading on consumer GPUs, achieving near-BF16 quality at Q4 size and ~70% less memory usage.
Claude Fable 5 is tested via World of AI Bench across coding, frontend design, 3D worldbuilding, WebGL rendering, physics simulations, vision, and agentic workflows, comparing performance against GPT-
Hands-on evaluation of Claude Fable 5 through browser workflows, C++ game simulation, SVG animation, packet visualization, scene generation, Python 3D games, simulations, frontend design, and drum kit
Hands-on review of Claude Fable 5, an Anthropic Mythos-class1 model designed for safe general use.
Compares Gemma4 12B and Gemma4 12B QAT via VPS Dashboard Setup, Tower Defense Game addition, and Chat Client Build to evaluate instruction following, UI cleanliness, reliability, and usability under f
Evaluates MiniMax M3 through browser workflows, C++ game development, scene generation, multimodal coding, simulators, frontend design, and drum kit generation.
Demonstrates step-by-step testing of MCP servers using DeepEval and the LLM-as-a-Judge concept.
A bug fix in Gemma 4's chat template resolves issues that were silently degrading multi-turn agent performance.
Demonstrates Ideogram 4 features including world knowledge, text rendering, prompt adherence, bounding boxes, comics, and integration with ComfyUI using KJ nodes.
Compares Gemma4 12B and Gemma4 12B QAT via iOS UI Clone, Weather Dashboard, and Broken Code Debug + Repair Challenge using single-file HTML, CSS, and JavaScript tests.
Demonstrates running Gemma 4 12B QAT locally via LM Studio on Windows PCs with 8GB VRAM, integrated into VS Code using the Continue extension for offline coding assistance.
Locally installs and tests SimpleMem for storing, compressing, and retrieving long-term memories using semantic lossless compression alongside Ollama.
Demonstrates deploying custom applications using Abacus Supercomputer to host local AI models accessible to users.
Compares Unsloth’s q4 QAT and regular q4 GGUF variants of Gemma 4 12B on adherence, agency, coding, and memory using a 16GB VRAM system.
Tests Kombai within Cursor for site generation, block and theme creation, tech stack integration, interactive editing, autonomous browser QA, and bug fixing.
Locally installed and tested bosonai/higgs-audio-v3-tts-4b, a text-to-speech model designed for voice chat rather than simple reading.
Sets up Himalaya email skill in OpenClaw using Gmail IMAP/SMTP and app passwords, demonstrating reading and analyzing real emails plus querying Gmail from a Discord bot.
Gemma 4 12B with QAT and QWEN 3.6 are tested as Coding Agents in VS Code on an M5 Max (128GB) using real-world coding tasks instead of benchmarks or synthetic tests.
Demonstrates DeepSeek GUI features including Code Mode, Write Mode, Kun runtime, and scheduled agents within a local-first desktop workspace.
Stacks Google's QAT quantization with llama.cpp's MTP support to run Gemma 4 12B at double the speed locally.
Compares Qwen3.6 27B Q8 and MiMo-v2.5 via iPhone Replica, Ragdoll Physics Simulator, and 3D Orbital Earth Explorer tests evaluating instruction adherence, UI cleanliness, functionality, and interactiv
Nex-AGI updates Qwen 397B to top GLM 5.1 and Claude Opus in benchmarks using Inferencer App on Mac Studio M3 Ultra 512GB with Nex-N2-Pro.
Tests Google QAT versus Unsloth Q4_0 quantized versions of Gemma 4 12B at identical file sizes to determine which approach yields better results.
News update covering Bernini, Deja View, PaGeR, Magenta Realtime, GPT Dreaming, Mamma, Reve 2, Ideogram v4, Gemma4 12B, Qwen 3.7 Plus, Cosmos 3, RTX Spark, Stable Layers, humanoid robots, Minimax M3,
Installs and tests BLS-Mini-Code-1.0 from CohereLabs locally using resources from Hugging Face.
Installs and tests nanowhale-100m, an ~110M parameter language model implementing the DeepSeek-V4 architecture.
Coding challenges test MiniMax-M3 versus Qwen3.6 27B using an iPhone-style UI, ragdoll physics simulator, and a fake Qwen3.7 30B release page to evaluate usability and polish.
Evaluates Nemotron 3 ULTRA through browser OS, scene generation, OpenCode, flight and 3D printer simulations, C++, Python, niche knowledge, roleplay, and website generation tests.
Tests Gemma 4 12B q8 vs q4 quantizations on a system with 16GB VRAM and 32GB DDR4 RAM across performance, memory, coding, and agency benchmarks using unsloth/gemma-4-12b-it-GGUF.
Tests Gemma 4 12B's encoder-free multimodal architecture, coding, reasoning, and Three.js generation via World of AI Bench, comparing it against Qwen3.6-35B-A3B.
Locally installs and tests Gemma 4 12B QAT via Ollama using coding demos, multilingual translation, creativity, and situation demonstrations.
Evaluates NVIDIA Nemotron 3 Ultra using Inferencer App on M3 Ultra 512GB hardware against state-of-the-art models under the new Open License.
Evaluates instruction adherence, UI cleanliness, functionality, and interactivity using single-file HTML/CSS/JavaScript tests: iPhone Replica, Top-Down Car Game, and Live Weather Dashboard.
Installs and tests MisoTTS, a text-to-speech model based on the Sesame CSM architecture.
Covers Seven New MAI Models, Microsoft Scout, GitHub Copilot App, Project Solara, NVIDIA Nemotron 3 Ultra, Gemma 4 12B, MiniMax M3, Codex updates, Hermes Desktop, Ideogram 4.0, and other new AI tools.
Installs OpenClaw on Windows via openclaw.ai, connects OpenRouter with poolside/laguna-m1:free model, and reviews dashboard plus agent.md, soul.md, identity.md, user.md, heartbeat.md, bootstrap.md fil
Installs and tests Mellum2-12B-A2.5B-Thinking, a post-trained reasoning-augmented assistant model trained by JetBrains.
Demonstrates using the nvidia/nemotron-3-ultra-550b-a55b:free model via OpenRouter with agents like Claude Code, OpenCode, Pi, KiloCode, VS Code, T3.code, Hermes, and OpenClaw.
Tests NVIDIA-Nemotron-3-Ultra-550B-A55B with hermes-agent.