OpenAI introduced GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano models with a 1-million-token context window, positioning them against offerings from Google and Anthropic. The models excel at coding and follow OpenAI's ambition to create an agentic software engineer. While GPT-4.1 scored well on benchmarks, it still faces reliability challenges with longer inputs.

Meta CEO Mark Zuckerberg testified in the FTC's antitrust trial seeking to force the company to divest Instagram and WhatsApp. The FTC argues Meta holds nearly 80% of the personal social networking market, while Meta contends this definition is too narrow and excludes competitors like TikTok. The case hinges on proving Meta's monopoly power and illegal actions to maintain dominance.

Nvidia is establishing over one million square feet of manufacturing space in Arizona and Texas to produce AI chips domestically. Production of Blackwell chips has begun at TSMC's Phoenix plants, while supercomputer manufacturing facilities are being built with Foxconn in Houston and Wistron in Dallas. Mass production is expected within 12-15 months, with Nvidia aiming to produce up to $500 billion of AI infrastructure in the US over four years.

Salesforce's AI tool Agentforce now generates 20% of their APEX code in production. Rather than replacing developers, AI is transforming their roles toward more strategic work. Developers are adopting 'vibe coding' - giving high-level directions to AI instead of precise instructions. This shift requires new testing approaches and elevates developers as business partners.

Google DeepMind developed DolphinGemma, an AI model that helps interpret dolphin vocalizations and generates dolphin-like sounds. Built on Google's Gemma series, this efficient model will power a platform on Pixel 9 smartphones for the Wild Dolphin Project this summer, enabling researchers to generate synthetic vocalizations and listen for matching replies.