AI in week 27.
29 new tools and models from 4 analysed videos; 8 confirmed available in the EU.
Videos analysed
Full body waifus, Claude Fable is back, LongCat 2.0, mind-reading AI, live video editing: AI NEWSAIAI Search· 5 Jul 2026 · 20 tools
AI News: Fable's Back But This New Model is Better?MAMatt Wolfe· 3 Jul 2026 · 14 tools
Deepseek drops another HUGE breakthroughAIAI Search· 3 Jul 2026 · 2 tools
GLM-5.2: The Complete Guide to the Best Open-Source ModelMAMatt Wolfe· 1 Jul 2026 · 4 tools
Language Models
All Language Models →Claude Sonnet 5
Model · Anthropic
Anthropic's Claude Sonnet 5 LLM, with a high-effort "Max" setting; since succeeded by Claude Sonnet 5.5.
Luna
Model · OpenAI
Smaller, efficient OpenAI model reportedly cut in price by 80% after efficiency work; details are limited.
Agents A1
Model
35B Mixture-of-Experts open-weights model from InternScience for long-horizon agentic search, engineering and tool use.
LongCat 2.0
Model · Meituan
Meituan's open-weights 1.6T-parameter mixture-of-experts LLM (48B active), aimed at agentic coding and reasoning tasks.
Image Generation
All Image Generation →Nano Banana 2 Light
Model · Google
Google's faster, cheaper variant of Nano Banana 2 image model, generating images in about 4 seconds.
MrFlow
Library
Training-free staged sampling method that speeds up pretrained flow-matching text-to-image diffusion models.
RDM
Research
Research method that trains one-step image generators by matching feature distributions under frozen encoders.
Video Generation
All Video Generation →Gemini Omni Flash
Model · Google
Faster Gemini Omni variant from Google for generating and editing video from references and text prompts.
LiveEdit
Model
Research framework for real-time, causal streaming video editing with diffusion, running at about 12.66 FPS.
NotebookLM Shorts
App · Google
NotebookLM feature that turns user sources into roughly 60-second vertical short video overviews.
Agents & Automation
All Agents & Automation →Comfy MCP
API · ComfyUI
MCP bridge from ComfyUI that lets AI agents build, edit and run ComfyUI workflows from natural language.
Genspark Claw
Agent · Genspark
Genspark's hosted OpenClaw-based agent that connects to Slack, Teams and WhatsApp, with a desktop version.
Coding
All Coding →Productivity
All Productivity →Gemini Meet Notes
App · Google
Gemini feature in Google Meet that transcribes meetings and produces summaries and action items saved to Google Docs.
Claude Science
App · Anthropic
Anthropic desktop workbench app for scientists, comparable to Claude Code and Cowork, on paid plans for Mac and Linux.
Robotics
All Robotics →UBTECH U1 series
Hardware · UBTECH Robotics
UBTECH U1 series: humanoid robots from UBTECH Robotics with realistic faces and hair, reportedly in first deliveries.
OmniContact
Research
Hierarchical humanoid loco-manipulation framework that chains meta-skills via contact flow, accepted to CoRL 2026.
Aspire
Research · Invideo
Research system from Invideo that writes robot control code, repairs failures and stores fixes in a reusable skill library.
CHORD
Research · Nvidia
NVIDIA research framework that teaches dexterous robot hands manipulation from human demonstrations via contact wrenches.
SimFoundry
Research
Research method that turns a single photo or video into a simulation-ready 3D scene for robot reinforcement learning.
Hardware & Devices
All Hardware & Devices →Research
All Research →BuseyBench
Research
Informal benchmark that scores how well AI models draw SVG images of Gary Busey, ranked on an LLM-judged leaderboard.
DSpark
Research · DeepSeek
DeepSeek research component for multi-token output, intended to speed up LLM text generation.
PhysiFormer
Model
Diffusion transformer that generates rigid and elastic 3D mesh dynamics in world coordinates from initial velocities and materials.
Brain2Qwerty v2
Research · Meta
Meta research model that decodes non-invasive MEG brain signals into typed text; v2 trained on roughly 10x more data.
MusViT
Model
Vision Transformer foundation model for sheet music, pre-trained with masked autoencoding on 9.7M IMSLP pages.
VidiHand
Research
Research method that reconstructs 4D hand motion (MANO) from video using a fine-tuned video diffusion model.
Platforms & Infra
All Platforms & Infra →X MCP Server
API · X
Hosted MCP server from X that lets MCP-compatible AI tools access the X API using a user's own account permissions.
Inference.net
API · Inference.net
OpenAI-compatible inference API and gateway for open-weight and closed models, with provider failover and pass-through pricing.