delivery. This will only be the result of these new LLM code generators being developed at Microsoft and Google. This could also be the birth of a generation of systems that optimise themselves. In 25 years we will no longer see any programmers spewing out code in Visual Studio or Emacs. That is done. We will be seeing software designed like Jordi La Forge in Star Trek - “Computer, create a program that does X” - and the LLM will generate a program that pulls in aspects of other programs to create one that “does X.” For this to happen, though, the future of computer science teaching is going to be in system design and contracted interface architecture. The software design tools that output SysML to describe software will be able to run that SysML through an LLM and have code generated and be usable with no human interventions. Can you tell us about some significant past projects involving large corporate clients that you and the Beyond Ordinary have worked on? It depends on your metric for In the 90s we saw Aspect Programming come to light. This was an innovative way of reusing software slices across new software systems without having to program the same stuff, over and over. AP was only marginally successful, but it was a great idea. There are legal issues with AP, such as copyright infringement. AP is the process by which you extract a function (aspect) out of compiled code (DLL/lib) and link it with other extracts (slices) to make a new execution path (program). We do this already with reused libraries and modular programming, but that’s done in a trusted jail environment. AP was about breaking out of jail and creating slices across code that was foreign. AP didn’t take off because many different programmers were creating the libraries, and these programmers had different styles and code contracts for the interfaces. That means the slices were not compatible and required a lot of middleware wiring to make them work. With LLM code generators the disparity of style is gone. Tomorrow we can have an LLM tool that generates the same code repeatedly. This means the interface into the aspects will be consistent, which is the key to making AP work successfully. 15 years from now, AP will re-emerge as a dominating technique for rapid application large. We’ve had a very large automotive customer for whom we helped design a humanmachine interface (HMI). This project came to us as a Java project, but we quickly discouraged this customer from that path and put them on a custom solution using Javascript and a custom build of a browser/JS engine. This proved to be an innovation in the industry that disrupted how vehicle HMIs were developed. Before we did our HMI the industry would take 5 years to deliver a new HMI for a car. The team delivered the HMI for this customer in six months. This is the standard now. We’ve done quite a bit of work in the insurance industry. If you’ve ever bought insurance online then you’ve likely used the software that I wrote. When I started working in the insurance industry around 2000, there was no use of cryptography or data protection. It was a struggle at first and a steep mountain the climb, but eventually, I convinced some early clients to adopt cryptography in their databases. All of our insurance clients, today, start with cryptography on day 1. I am proud of that struggle as it has protected millions of people from data disclosures and elevated awareness of data protection in the industry. THE CEO INTERVIEW 31
RkJQdWJsaXNoZXIy Mjk3Mzkz