OpenAI Shuts Down Sora: Why the Most Hyped AI Video Model Is Being Retired
OpenAI has announced the sudden discontinuation of Sora, its AI video generation app and API, pivoting resources toward robotics and AGI research.
Month: March 2026
xMemory: The New AI Technique That Cuts Token Costs by 50% for Multi-Session Agents
Researchers from King's College London and The Alan Turing Institute have developed xMemory, a new memory architecture that cuts token usage by nearly 50% while improving AI response quality.
Insanely Fast Whisper: Transcribe 2.5 Hours of Audio in Under 98 Seconds with OpenAI’s Latest Model
Insanely Fast Whisper is a community-built CLI tool that leverages Flash Attention 2, batch processing, and Hugging Face Transformers to transcribe 150 minutes of audio using Whisper Large v3 in just 98…
Dexter: The Autonomous AI Agent Built to Do Deep Financial Research — Like Claude Code for Markets
Dexter is an open-source TypeScript financial research agent that autonomously plans, executes, and self-validates investment analysis using live market data from income statements, balance sheets, and cash flow data.
Microsoft VibeVoice: The Open-Source Frontier Voice AI That Transcribes an Hour of Audio in One Pass
Microsoft has open-sourced VibeVoice, a comprehensive voice AI framework featuring a 7B-parameter speech recognition model that processes 60-minute audio in a single pass, multilingual support across 50+ languages, and native integration with…
Anthropic’s Claude Now Controls Your Mac: The AI Agent Race Enters a New Phase
Anthropic has released the most ambitious consumer AI agent to date, giving Claude direct control of Mac computers ??clicking buttons, opening apps, and navigating software on users' behalf. Available now for Pro…
Google TurboQuant Algorithm Cuts AI Memory Usage by 6x ??And It’s Completely Open
Google Research's new TurboQuant algorithm achieves 6x KV cache compression with zero accuracy loss, potentially cutting enterprise AI inference costs by 50% or more. The technology is now freely available.
AI Scientist-v2: SakanaAI’s Autonomous Research System Produces First AI-Written Peer-Reviewed Paper
SakanaAI's AI Scientist-v2 represents a quantum leap in automated scientific discovery, autonomously generating hypotheses, running experiments, and writing papers accepted at peer-reviewed workshops without human templates.
xMemory: How Hierarchical Memory Architecture Cuts AI Agent Token Costs by 50%
xMemory uses a four-level hierarchical memory architecture to cut AI agent token costs by nearly 50% while improving long-term reasoning accuracy.
Microsoft VibeVoice: Open-Source Frontier Voice AI for 60-Minute Transcription and 90-Minute Speech Synthesis
Microsoft's VibeVoice open-source framework delivers 60-minute transcription, 90-minute speech synthesis, and real-time TTS capabilities through Hugging Face integration.