Silverthread

DAN STURTEVANT: So, my name is Dan Sturtevant. I'm the CEO and co-founder of Silverthread. My background originally was technical over the course of my career. I was a developer in cryptography and device driver development for Linux. I did supercomputing. I worked in this simulation of-- like a simulation of systems in the automotive and aerospace industries.

And then I went back to MIT to get into system design and management. Over the course of my time at MIT, both through the masters and the PhD program, I connected with a few professors named Alan McCormack and Carlos Baldwin, who had some really interesting ways of measuring the technical health of software systems.

Alan McCormack and Carlos Baldwin, both at Harvard Business School, are two of the co-founders of Silverthread who did a really-- a lot of really interesting research that we've built our company on. They came up with ways of measuring the health of software systems using graph theory and then also tying that to various business outcomes. Carlos wrote a really interesting book called Design Rules: The Power of Modularity back in 2002 that a lot of this is based on. And I thought that it really explained a lot of my experiences and code base that I worked in previously.

I had some places where I worked where development went very well. And I had other places where I worked where it didn't. And what I found using their techniques is that their automated ways of measuring the health of code bases were highly predictive of the challenges that I had had earlier in my career. When I graduated, we all decided to start Silverthread to commercialize that technology based on a lot of demand that was coming up at that time among people in industry and defense.

[MUSIC PLAYING]

DAN STURTEVANT: So Silverthread's technology is based on research that had been done at MIT and Harvard over the course of about 15 years. Professor Alan MacCormack and Carliss Baldwin had developed ways of measuring the technical health of software systems and did a variety of research publications on those topics. Later on, when I started working with them, we were statistically connecting those technical health measures to business outcomes, specifically things like, how does that relate to a defect density or developer productivity?

And what does that mean for agility of a software development organization, or risk, or waste, or cost? Based on the ability to couple a deep understanding of the technical health of software systems and the business impact that it has, we can work with people to bridge the gap between a technical understanding of what you're building and the business implications so that you can make better decisions.

So CodeMRI is a technology that we developed originally at MIT and Harvard, and now have commercialized. It is a way of measuring the technical health of software systems, both from an architectural standpoint and from a code quality standpoint. We've also done a lot of statistical work linking those measures to business outcomes, such as developer productivity, and defect density, and others.

What we can do now is, given a new code base, we can adjust that code into the platform, run it through the predictive models so that we can show you what the economics are of working in the system, as well as give you a detailed understanding of the technical help. We could also tap into someone's version control systems, issue tracking systems, other management systems, to calibrate those models instead of having predictive models if they want to make more refined adjustments to the policies that they have internally.

CodeMRI also contains technologies that help people fix systems. That's been developed more recently. But we've developed tools to give people perceptiveness into the technical health of the system, be able to set controls and rules, so that they prevent new problems and identify the best thing to do in the system to break apart existing problems. So people have used this to prevent issues and then go on a path where they are improving, both their technical health and the economics of their organization, and measuring both along the way.

One example of an impact we had was an organization that we've worked with recently that had a 2000 file Java code base that had been around for about a decade. It was in reasonable shape, but it had definitely developed some challenges. We did predictive analytics that showed the amount of waste and cost and risk associated with future development in that code base.

Based on our combined assessment of both the technical architecture and of the economic implications, that team was able to make the business case within their organization to get funding to pay down that technical debt. They were then, over the course of about four months, able to spend 25% of their development labor attacking the health issues in that code base, completely eliminated them. And when we did the assessment after the fact, that we found that they had become three times as productive as a team.

[MUSIC PLAYING]

[MUSIC PLAYING]

DAN STURTEVANT: A lot of people are focused on agile software development. But what they find is that as their companies grow and as their code bases scale up, agility breaks down. And often that happens because the code base becomes non modular, it becomes tied up in ways that are architecturally unsound, making it so that everyone's got to talk to everyone because everyone's actions can break everyone else. It's not scalable. We've worked with companies to help them make their code bases modular, which effectively makes it so they have architectural agility.

And what we found with them is a good agile process combined with architectural health, makes them actually agile in practice not just in theory. Software is a very difficult thing to manage inside of an organization. It's the one thing that a CFO really can't measure like you can any other kind of asset that you have. You can measure the depreciation schedule of computers in factories, and you can use quantitative evidence around the productivity of a manufacturing line. But really understanding the economics of your software development organization, right now it's a black box.

A CFO is really in a bad place. We, by combining information about the technical architecture of a code base and the economic impact, we can quantify the relationship between those things so that the CFO and the architect can have productive conversations about how much money you should spend on improving the technical health of systems, how much you should spend on new features, how much you should spend on bug fixes, what kind of posture you should have internally so that you can be most effective as an organization.

Silverthread tools report information about a software system in, both, technical and in business terms. And presents that information in a way that's consumable by both audiences so that they can have better conversations with each other right now the architects in an organization and the CFO in an organization it's hard for them to communicate effectively we bridge that gap. So that they can become more agile over time.

[MUSIC PLAYING]

DAN STURTEVANT: Silverthread as a company is focused on helping people gain a much better understanding of the technical health of their systems and the economic impact. We have tools to help people do that assessment, and we also have care tools to help people fix their systems.

We've developed partnerships, and we continue to want to develop new ones in several key areas, one of which is we need people to be able to take our diagnostics and use those in the field. There are several places we can't go internationally, or into [? SCIFs ?] inside the Defense Department. There are several people who have access to code and want to be able to use our results to help their clients.

Other partners will want to use our tools to help someone repair a code base that is challenged. So they use our tools while doing the work on top of it to improve the health of a system that they are working on with a client. Silverthread has a data vault of 10,000 systems that we can compare a code base against when we're looking at its health.

We've built that up over the course of about 15 years of R&D, both at Harvard and MIT, and we continue to add to it as we work with new clients so that we can give our customers a well-rounded view on their code base, but also where it sits from an industry perspective. One of the things to understand is that code bases that we look at all have technical health challenges of one kind or another.

Software's only been around for 50 years. No one's figured out how to do it perfectly. So often our customers are apprehensive about seeing their results. And what people really need to know is that we're all in this together.

Software is a challenge, and most of the systems have some difficulties, but can be improved. And being able to see where you sit relative to others is healthy, and it's important as an organization.

[MUSIC PLAYING]

DAN STURTEVANT: Silverthread has worked with several people across industry, and several verticals, including healthcare, including automotive, aerospace, medical devices. Basically people who have complex systems that have the potential to do great good and to do great harm if they malfunction. We do a deep technical assessment and then can predict the risk that might be involved with shipping something that can hurt people.

Several people have used our assessment to get a better handle on what their risks are, and then used our tools that helped them fix those systems to mitigate those risks. Based on that, people have become more agile and had better economics within their organization. But they've also become more confident in their ability to ship complex systems and do so reliably.

Silverthread has had several projects with the United States Air Force. The most recent of which is a project to help them improve the health of eight systems that are in various states of development. We're going to be working with their developers to help them understand the health of their system, to understand the economic impacts, to make improvements to the architecture of their code bases, to lock it down so that new challenges don't arise in those systems. And then to monitor the resulting economic benefits and risk related benefits that they have.

[MUSIC PLAYING]

[MUSIC PLAYING]

DAN STURTEVANT: It's a real honor to be invited to be part of STEX 25. We're flattered by the opportunity. I think that we've grown a lot over the past five years. We've gotten some good proof points, both in the DOD and in industry, that show that there's a real appetite for people who want to better understand the health of their code bases and how they impact the economics of their organization. And it's great to have that kind of validation.

The ILP has been a great source of leads for us. We've been speaking with several ILP members at many of the conferences they have, both here at MIT, across the country, and internationally. It's great to get out in the world and see what people are doing, what challenges they have, and have the ability to interact with executives at a very high level. We have several pilots going with current ILP members and are very grateful for the opportunity.

[MUSIC PLAYING]

[MUSIC PLAYING]

DAN STURTEVANT: Silverthread is a software economics company. We work with executives at the top of some very large software development organizations. They might have thousands of simultaneous software development projects and have a poor understanding of which ones are going well and which ones might have some challenges. We have found that the technical health of these growing code bases that are being developed is a strong driver of the economics within each of those teams.

It impacts developer productivity, IT impacts defect density, it impacts the risk that that software poses when it's shipped into the field in some of the safety critical systems out there. We help executives understand, in an unbiased way, the health of their systems and the impact that's having on their economics across the organization.

We also have tools that teams can use to improve the health of those code bases and measure the resulting benefit to risk and economics so that executives can see what is going across their portfolio in real time and monitor the improvements so that they can continuously improve.

[MUSIC PLAYING]