Imagine a world in which your software development teams could double the time and energy they spend on creating new code. Double.
Think about how rapidly technology has advanced in the past few decades. And then consider—developers spend an average of 55% of their time on legacy software maintenance instead of creating new code. What could technologists accomplish if they could apply more of their brainpower to innovating and solving new problems instead?
What if we could double the capacity of human creativity to solve problems through code?
We know it’s possible. That’s why Sema came to be.
The Sema Story
I am the son of an IBM programmer and a math and computer science teacher. I learned rudimentary Turtle Basic on a Commodore 64, but that was as far as my programming skills went. I did learn from my parents a love of curiosity and technology. They also taught me that logic, reason and data could be brought to bear on the most important challenges.
Throughout my career, I have sought out what I believe to be the toughest and most important data and technology problems in education, government and business to study and try to solve. After leading operational transformations at a variety of large organizations, I saw a pattern emerge. Pretty much everyone is spending an inordinate amount of time, money and energy on maintaining legacy software, or existing code.
The 55% number is an average; there are organizations who have entire teams working on nothing but software maintenance. Those software engineers are not working on new features or applications. They’re just making the old systems work the way they’re supposed to work and trying to keep users and customers satisfied—or at least not super-dissatisfied.
When I came to understand how much time developers spent on maintenance, my first hunch was that perhaps it was interesting, compelling, a puzzle to unlock. I’ve since come to understand that is rarely so. Figuring out someone else’s code—much less code written by generation after generation of developers and development teams—is tedious. They call it untangling “spaghetti code” for a reason. It’s not a puzzle, it’s a burden. And, of course, the alternative is spending that time taking on new problems and solving for the future instead. Time spent fixing old code is a misuse of all that ingenuity.
So, when I saw this problem, and saw how big it was—$250BN a year by 2020—I decided to dig in. I wondered if emerging capabilities in machine learning could take that burden off developers, freeing them for what they really wanted to be doing.
I dove into the research, reading books and literally thousands of articles on the topic. I reached out to more than 500 scientists over six months. Three of these top software maintenance and machine learning experts joined our team and became founding contributing scientists. Together with an outstanding business-side leadership team, we’ve launched Sema, which will build expert systems using deep learning and other techniques that automate full software maintenance steps in a model that understands old or otherwise patched and confusing systems, models a new system architecture, and generates a new system codebase.
Our Vision Is in Our Name
The name “Sema” is derived from the Sanskrit word Lokaksema, which means “welfare of the world.”
Sema will write code that helps fix code. In doing so, we will double the capacity for human creativity in software and unlock tremendous potential to solve even greater problems in the future.
I couldn’t be more proud of the team here at Sema and all of the folks who helped us get to this point. I’m even more excited about what’s ahead.