ChatGPT for Healthcare launches, Claude Code gets code-simplifier plugin

OpenAI introduced OpenAI for Healthcare on January 8. It includes ChatGPT for Healthcare, which became available to health systems that day and is being rolled out at several institutions.¹ According to OpenAI, the products use GPT-5.2 models evaluated with real-world clinical scenarios and physician-led testing.¹

Unlike the separately launched ChatGPT Health for consumers, the offering is aimed at people working in care, administration, and research.²¹ Described features for institutional use include role-based access and single sign-on.³ Those controls matter when an AI tool is connected to workflows involving sensitive health information. The announcement does not specify prices or how far deployment has progressed at each named institution.¹

Anthropic has released the code-simplifier agent that the Claude Code team says it uses internally. The plugin is designed to simplify code and can be installed through the official Claude plugin marketplace. The release covers that agent, not the full Claude Code application.¹

That distinction matters for developers looking for an open-source alternative to Claude Code: releasing one component does not make the product itself open source. Claude Code remained available through paid Claude subscriptions in January, including Pro, Max, Team, and Enterprise.² The plugin adds an option within the Claude Code workflow, but the announcement alone does not establish how much it improves code quality in a given project.¹

In early January, Phison introduced an extension of aiDAPTIV+ for local AI inference on PCs. The company says it can run a 120-billion-parameter mixture-of-experts model with 32 GB of system memory. The approach moves model parameters that are not currently needed into a flash-based cache. Phison demonstrated it on several types of consumer PCs, including laptops and desktops.¹²

The approach could put larger models within reach of machines that otherwise lack enough system memory. Local inference may interest businesses that do not want to send every request to a cloud service. Lower memory use does not necessarily mean fast responses, however: moving parameters between flash and memory can slow inference. Independently verified latency figures for the demonstrated configurations are not available.³²

Ford introduced an AI assistant at CES 2026 designed to answer questions about a specific vehicle. Using a photo and vehicle data, it could estimate, for example, how many bags of mulch would fit in a truck bed. Cargo-space questions are another intended use. Ford plans to bring the assistant to its Ford and Lincoln smartphone apps in 2026 and integrate it into vehicles in 2027.¹²

Ford's approach goes beyond general chatbot answers by connecting the assistant to vehicle specifications for practical questions about available space.¹ A photo-based result is an estimate, not a guarantee that a load will fit. Ford has not said which models will receive the in-vehicle version first.¹