A month and a half ago, Embarcadero introduced Kai, the AI assistant for RAD Studio. Since there have now been several updates and blog posts in quick succession, I thought it would be useful to summarize them here.
For anyone who hasn't looked at it yet, Kai is an AI assistant built specifically for Delphi and C++Builder developers. Rather than being a generic chatbot, it integrates directly into the RAD Studio IDE, understands projects, source code, forms, and the development workflow, and supports both cloud AI services and local LLMs.
Quality First: Kai 1.0.1
The first update after the initial release focused almost entirely on quality. Version 1.0.1 improved the overall quality, stability, and usability of the product, delivering a more reliable day-to-day experience for RAD Studio developers. It also added support for authentication when connecting to local AI servers such as Ollama and LM Studio.
More Polish in Kai 1.0.2
The second maintenance release continued along the same path. Kai 1.0.2 improved compatibility with different Windows locales and international character sets, fixed several IME problems for Asian languages, and generally made the product more reliable. It also added much improved support for completion with local LLMs, including support for FIM (Fill-In-the-Middle) tokens.
Beyond Code Generation
One blog post that's quite interesting is the one by Antonio. It explores using Kai to modernize existing Delphi applications by refreshing their user interfaces. Many long-lived business applications still have layouts that clearly reflect the era in which they were originally written. While the underlying code may still be perfectly good, the user experience often shows its age.
Using AI as a design assistant -- to suggest more modern layouts, improved spacing, better color schemes, or updated visual styles -- is an interesting direction. It's not about replacing developers or designers; it's about giving them a starting point and helping accelerate work that is often repetitive and difficult to begin from a blank page.
I suspect this is only the beginning of how AI will influence application modernization. Code generation is useful, but helping improve existing applications may ultimately prove even more valuable when done within the IDE itself.
Looking Ahead
Kai is in a fast moving space, AI for developers. The rapid cadence of maintenance updates, the focus on developer feedback, and the gradual expansion into new AI-assisted workflows indicate a product that will continue to grow over time. Expect we'll have plenty more to talk about in the coming months.