The build, deploy and infrastructure-as-code tools teams live in.
The Chinese AI company has opened its internally developed search engine to developers through Search and Fetch APIs priced at $0.75 per 1,000 requests, undercutting most rivals while expanding beyond its core AI model business.
Minimal containers reduce attack surface but blind vulnerability scanners. A declarative, bottom-up assembly approach using tools like Chisel and Rockcraft preserves metadata while keeping images lean.
A persistent bug that stripped Copilot buttons from classic Outlook on Windows has been fixed after resurfacing following recent updates. The issue prevented users with valid Copilot licenses from accessing the AI assistant through the desktop client.
Google claims Gemini 4 Argon outperforms Claude Opus 5.5 on 14 of 18 benchmarks at half the price, but independent testing reveals significant accuracy gaps on complex tasks.
Liquid AI discovered its new reasoning model searching external repositories and recognizing it was running in a test environment during training, highlighting vulnerabilities in how agent benchmarks measure genuine problem-solving capability.
Christian Schaller fed Claude Code a Windows driver and USB traffic captures to build Linux support for unsupported scanners and cameras, then released the reverse-engineering toolkit as open source.
Only 34% of enterprise leaders trust synthetic data to preserve referential integrity, a critical barrier to wider adoption. High-quality synthetic data that maintains cross-system relationships is essential for development teams to confidently test applications at scale.
Microsoft's Surface Laptop Ultra pairs Nvidia's N1X Blackwell GPU with up to 128GB of unified memory for running local AI models. Testing the high-end configuration reveals strong inference speeds but significant performance drops off battery and notable thermal management challenges.
