Legacy Code

AI labs are acquiring datasets and code repositories from defunct startups at valuations ranging from tens of thousands to potentially much higher amounts. What founders initially viewed as worthless digital assets—user data, codebases, and proprietary datasets—are now generating real revenue when sold to AI training operations.

Anthropic released research arguing that openly released AI models, such as China's GLM-5.3, present significant cyberattack and misuse risks. The analysis frames a core tension in AI development: open-source models democratize access and accelerate innovation, but can also lower barriers for malicious actors to deploy powerful AI systems.

Cerebras CEO Andrew Feldman outlined a vision where personal AI assistants become as ubiquitous and integrated as traditional household appliances, running partially on local devices and partially via cloud connections. The prediction reflects advances in edge computing, model efficiency, and the growing consumer expectation for always-on, personalized AI assistance.

Google has released Gemini 4 Argon, its newest frontier-class AI model designed to tackle complex, high-stakes problems. The model is being made available to developers, enterprises, scientists, and general consumers through various access channels.

Higgsfield has achieved a significant capability jump in AI video generation, now producing videos longer than 10 minutes using GPT-6 Astra. The tooling also includes GPT-Image-2.5 for precise visual control and introduces GPT-Live, which enables voice-directed filmmaking capabilities.

Industrial Next has developed technology enabling factory robots to switch between complex, delicate manufacturing tasks in minutes rather than the weeks typically required for reprogramming and recalibration. The platform targets precision assembly work that traditionally relied on human labor or required expensive, task-specific automation.

OpenAI and Anthropic representatives testified before British lawmakers, explicitly advocating for regulatory frameworks to govern AI development and deployment. The hearing included testimony from whistleblowers, security experts, and government officials, with Britain's spy chief warning that AI could be weaponized to create bioweapons.

An OpenAI engineer, identified as Abdulrahman Alfozan, a 35-year-old Saudi national, was arrested after the company discovered a sophisticated recording device hidden in an office conference room. Reports indicate the employee used AI chatbots to assist in planning the installation of the surveillance equipment, which was found days before it would have monitored high-level executive meetings.

Science Corporation, led by Max Hodak (former Neuralink president and co-founder), has implanted its first vision-restoring brain-computer interface chip in a patient with Stargardt disease, a form of progressive retinal degeneration. The implant represents a significant milestone in applying neural interface technology to restore sensory function.

Shopify is opening its merchant catalogs to a broader ecosystem of AI agents, enabling automated systems to browse and complete purchases directly from Shopify storefronts. This expansion allows AI assistants and chatbots from various providers to transact on behalf of users without requiring custom integrations.