Software doesn't need to be readable anymore. it needs to be explainable (ghuntley.com)

In a recent discussion at the AI:Engineer World Fair, AI21 Labs highlighted a transformative shift in software development, emphasizing that software no longer needs to be human-readable but should be explainable to humans. This change stems from how AI technologies can automate code generation, rendering traditional coding methods obsolete. The economic viability of AI is also under scrutiny, with concerns that current adoption rates may not justify the heavy investments in AI infrastructure and talent. Professionals are now tasked with identifying where advanced AI capabilities are genuinely necessary, signaling a need for efficiency and cost-effectiveness in AI utilization. The conversation suggests a potential resurgence of unikernels and programming languages optimized for machine agents rather than human readability. This shift has important implications for software architecture, presenting an opportunity to reconsider long-standing practices and potentially simplify systems by minimizing the complexity traditionally attributed to human operation. Languages with stronger type systems, like Rust and Haskell, are highlighted for their ability to ensure stability and facilitate agent-driven maintenance. As programming languages evolve, there's a growing belief that they can be tailored more toward machine understanding, allowing for greater agility and innovation in the tech landscape.