EPFL scientists breakthrough 'error recycling' method teaches AI to learn from its own mistakes, enabling unlimited-length video generation that maintains quality for minutes instead of the previous 30-second limit without requiring extra computing power.
Claude AI now powers 4% of all GitHub commits and could reach 20% by 2026, as Anthropic's revenue growth outpaces OpenAI while major enterprises like Accenture deploy the AI agent to thousands of professionals, threatening traditional software business models.
Google unveils Developer Knowledge API that grants AI development tools direct access to Firebase, Android, and Cloud documentation in machine-readable format, enabling automated searches and retrieval of implementation guidance while promising expanded features including structured code samples and broader documentation coverage.
AI startup Fundamental emerges from stealth with massive $255M Series A funding at $1.2B valuation, developing breakthrough Nexus technology that analyzes enormous structured datasets traditional AI cannot handle, already securing seven-figure Fortune 100 contracts.
Anthropic unveils Claude Opus 4.6 featuring revolutionary 'agent teams' that enable multiple AI agents to autonomously divide and tackle tasks in parallel, boasting enhanced coding abilities, 1 million token context window, and benchmark-leading performance while competing directly with OpenAI's multi-agent technology.
Reddit's AI-powered search feature explodes from 1 million to 15 million weekly users in 2025, with CEO Steve Huffman positioning it as a major revenue opportunity while announcing plans to eliminate the distinction between logged-in and logged-out users by Q3 2026 through AI personalization.
OpenAI releases GPT-5.3 Codex, a groundbreaking AI coding model that runs 25% faster than its predecessor and remarkably helped create itself through self-debugging capabilities, launching just minutes after competitor Anthropic in a dramatic race to market.
Microsoft releases breakthrough research identifying three telltale signatures that reveal when AI models have been secretly compromised with malicious backdoors, launching an open-source detection tool as security experts warn the true scale of poisoned artificial intelligence systems threatening enterprises remains dangerously unknown.
Gartner predicts half of companies cutting jobs for AI will rehire workers as the technology proves insufficient, with major firms like HP, Salesforce, and Klarna discovering AI systems still require human oversight and cannot match human judgment in most roles.
New Anthropic research reveals AI systems fail complex tasks through confusion and random errors rather than malice, with scaling improvements actually worsening performance on difficult problems while reducing bias, suggesting future AI failures will resemble industrial accidents caused by incoherence.