Google says Gemini 3.5 Flash is now available across Gemini, Search, developer tools, and enterprise products. The launch signals a broader push toward AI agents that can plan, code, and work across longer workflows.
With unprecedented efficiency, localized reinforcement learning, and a massive SpaceXAI partnership, Cursor’s newest model is reshaping how artificial intelligence tackles complex, real-world coding.
Unmatched Efficiency...
The renowned AI researcher and OpenAI co-founder joins Anthropic's pre-training team, signaling a major strategic shift in the race to build the next generation...
In a world-first agreement, the Mediterranean nation is offering premium AI access to all residents and citizens abroad—but education comes first.
A Landmark Agreement: OpenAI and...
Shanghai AI Lab’s SU-01 model proves that massive parameter counts aren't the only path to elite mathematical and scientific problem-solving.
A Paradigm Shift in AI...
A controversial shift away from strict, case-by-case data approvals raises major cybersecurity and privacy alarms, putting public trust in the National Health Service to...
Google DeepMind says AlphaEvolve, a Gemini-powered coding agent, has moved from research demo to practical algorithm designer across genomics, power grids, mathematics, Google infrastructure, and commercial optimization. The evidence is broad, but some claims still need careful caveats.
Anthropic says Claude Mythos Preview can find and exploit serious software vulnerabilities at a level that changes cyber defense. Project Glasswing turns that dangerous capability toward critical software, but the claims still need careful caveats.
This week in AI brought Claude interpretability research, larger compute deals, low-latency voice infrastructure, finance agents, on-device AI privacy concerns, multimodal models, faster inference claims, geopolitics, security failures, and AI-assisted game development.
Anthropic says Natural Language Autoencoders can turn model activations into readable text, giving safety teams a new way to audit hidden model behavior while leaving major reliability and replication questions open.
