Software gets malleable when describing it is enough (ttvl.co)

Recent advancements in large language model (LLM) technology are transforming the landscape of software engineering, making it easier and quicker for developers to create custom tools. Instead of spending significant time coding for individual projects, engineers can now describe desired software behaviors in natural language, allowing the AI to generate, execute, and correct the code autonomously. This shift not only lowers the barrier for software creation but also introduces a sense of flexibility, enabling developers to mold software as if it were a physical material tailored to specific needs. This evolution in programming is particularly significant for the AI/ML community as it opens the door to ambient computing—where everyday environments can become responsive and intelligent. The challenge remains in creating systems that can seamlessly interface with real-world contexts, fulfilling unique, small-scale requirements that traditional products cannot address. By logging experiences from initial implementations, such as dynamic responses in a studio setting, there’s potential for developing a cohesive platform that allows users to dictate behaviors that their environments adapt to. If realized, this could lead to a revolutionary way of interacting with both software and our physical spaces, reminiscent of the innovative concepts explored in Dynamicland but accessible to a wider audience.