X.COMMODELSPUBLISHED 31 March
2038894956459290963
Chaofan Shou (@Fried_rice) on X
Claude code source code has been leaked via a map file in their npm registry! Code: https://t.co/jBiMoOzt8G
AN IN-DEPTH REPOSITORY
THE MONTHLY
SIGNAL
Every model, paper, media release, agent and tool from the month—in one source-linked archive.
COMPLETE MONTHLY ARCHIVE
March 2026
A dense month for world models, agents and creative infrastructure—alongside stronger foundation models and increasingly capable real-time tools.
THE COMPLETE MONTHLY INDEX
Every Signal
79ITEMS IN MARCH 2026
Every linked paper, product release, research project and article has been checked directly. Cards use the publication's own description, imagery and published date wherever the source supplied them.
The AI Search helps discover the wider field, while every available card opens the original publication. Dates come only from publication metadata or official source APIs—never from a video. A checked date records our review, not the release. Older undated entries show their archive month.
X.COM
IDP.NATURE.COMMEDIAPUBLISHED 31 March
Evo 2 genome model
Client Challenge
Evo 2 genome model is linked to an original media publication. The page did not provide a concise preview, so open it for the author’s details, demos and results.
ARCPRIZE.ORGMODELSMONTH MARCH
3
ARC-AGI-3
ARC-AGI-3 is the first interactive reasoning benchmark for AI agents—play as humans and build agents that learn in novel environments.
CORAL79.GITHUB.IOTOOLSMONTH MARCH
ActionPlan
ActionPlan: Future-Aware Streaming Motion Synthesis via Frame-Level Action Planning
CUA-SUITE.GITHUB.IOTOOLSMONTH MARCH
Cua Suite
CUA-Suite: Massive Human-annotated Video Demonstrations for Computer-Use Agents
CUA-Suite: Massive Human-annotated Video Demonstrations for Computer-Use Agents
GITHUB.COMTOOLSMONTH MARCH
DaVinci MagiHuman
GitHub - GAIR-NLP/daVinci-MagiHuman
Contribute to GAIR-NLP/daVinci-MagiHuman development by creating an account on GitHub.
JIAZHENG-XING.GITHUB.IOTOOLSMONTH MARCH
Lumosx Home
Lumos𝒳: Relate Any Identities with Their Attributes for Personalized Video Generation
Lumos𝒳: Relate Any Identities with Their Attributes for Personalized Video Generation
MATRIX-GAME-V3.GITHUB.IOTOOLSMONTH MARCH
Matrix Game V3
Matrix-Game 3.0: Real-Time and Streaming Interactive World Model with Long-Horizon Memory
Matrix-Game 3.0: Real-Time and Streaming Interactive World Model with Long-Horizon Memory
KRISTEN-Z.GITHUB.IOTOOLSMONTH MARCH
Megaflow
MegaFlow: Zero-Shot Large Displacement Optical Flow
MegaFlow: Zero-Shot Large Displacement Optical Flow
PRISMAUDIO-PROJECT.GITHUB.IOMEDIAMONTH MARCH
Prismaudio Project
PRISMAUDIO: DECOMPOSED CHAIN-OF-THOUGHTS AND MULTI-DIMENSIONAL REWARDS FOR VIDEO-TO-AUDIO GENERATION
Video-to-Audio (V2A) generation requires balancing four critical perceptual dimensions: semantic consistency, audio-visual temporal synchrony, aesthetic quality, and spatial accuracy; yet existing methods suffer from objective entanglement that conflates…
MEDIAMONTH MARCH
Pulse Of Motion
Pulse of Motion
Xiangbo Gao · Mingyang Wu · Siyuan Yang · Jiongze Yu · Pardis Taghavi · Fangzhou Lin · Zhengzhong Tu
DANACOHEN95.GITHUB.IOTOOLSMONTH MARCH
RealMaster
RealMaster: Lifting Rendered Scenes into Photorealistic Video
TOOLSMONTH MARCH
RealRestorer
RealRestorer: Towards Generalizable Real-World Image Restoration
Towards Generalizable Real-World Image Restoration with Large-Scale Image Editing Models
WILLIAM-WANG2.GITHUB.IOTOOLSMONTH MARCH
RetimeGS
RetimeGS: Continuous-Time Reconstruction of 4D Gaussian Splatting
RetimeGS: Continuous-Time Reconstruction of 4D Gaussian Splatting
AIDEMOS.ATMETA.COMMODELSMONTH MARCH
Tribev2
TRIBE v2
A self-supervised vision transformer model by Meta AI
RESEARCH.GOOGLERESEARCHMONTH MARCH
Turboquant Redefining Ai Efficiency With Extreme Compression
TurboQuant: Redefining AI efficiency with extreme compression
Amir Zandieh, Research Scientist, and Vahab Mirrokni, VP and Google Fellow, Google Research
LUKASHOEL.GITHUB.IOMEDIAMONTH MARCH
Video To World
World Reconstruction From Inconsistent Views
World Reconstruction From Inconsistent Views
AGENTSMONTH MARCH
Worldagents
WorldAgents: Can Foundation Image Models be Agents for 3D World Models?
WorldAgents employs a multi-agent architecture to synthesize 3D-consistent worlds from 2D foundation models: a Director formulates prompts, a Generator synthesizes new views, and a Verifier curates frames from both 2D and 3DGS reconstruction space.
BLOG.GOOGLETOOLSPUBLISHED 26 March
Build With Gemini 3 1 Flash Live
Build real-time conversational agents with Gemini 3.1 Flash Live
Google is launching Gemini 3.1 Flash Live via the Live API in Google AI Studio, for building realtime voice and vision agents.
COOLFLYAIRCRAFT.COMMODELSPUBLISHED 26 March
Coolfly eVTOL
CoolFly eVTOL: Mass Production and Public Flights in 2026
CoolFly eVTOL mass production starts in late 2026. Learn what is vtol aircraft tech, how electric vtol aircraft work, and how you can join the future of flight.
COHERE.COMMODELSPUBLISHED 26 March
Transcribe
Cohere Transcribe: Open-Source Speech Recognition | Cohere
Unmatched accuracy and speed. Transcribe converts your business’ audio data into precise text for search, analytics, and automation.
BLOG.COMFY.ORGTOOLSPUBLISHED 25 March
Dynamic Vram In Comfyui Saving Local
Dynamic VRAM in ComfyUI: Saving Local Models from RAMmageddon
A new memory system that makes it possible to efficiently run the largest models on the smallest memory.
MISTRAL.AIMEDIAPUBLISHED 23 March
Voxtral TTS
Speaking of Voxtral
Mistral released a compact 4B text-to-speech model for expressive multilingual generation, low-latency streaming and rapid adaptation to new voices from short references.
AISTUDIO.GOOGLE.COMTOOLSMONTH MARCH
Aistudio
Google AI Studio
The fastest path from prompt to production with Gemini