Here’s just one segment from an excellent interview with Will Roper on the Defense Unicorns podcast with Rob Slaughter, who worked on software factory efforts like SpaceCAMP and Platform One.
There’s a little bit of conflict in what I just said: you got to get things wrong to get them right. But you also need to produce signs of hope soon enough you don’t get shut down. That’s why I was very focused on what the first thing was, the first demonstrable result. I was typically personally involved in getting to it. With that one result you can push back on the pressure that tries to collapse the initiative.
Kessel Run is a great example of how bureaucracies respond. Stand up a new organization. They get a major defense program that looks very different than the past. Level of effort funding because it’s based on a process not a product. The acquisition system looks at level of effort funding and says that’s sloppy acquisition, you don’t know what you’re buying. We said that it’s a process, that’s exactly how you do it. We manage it by metrics.
The number of times the budget for Kessel Run was under attack by multiple sources in Washington — I lost track count. It was a fight to keep it alive as a program and not turn it into something else. But it produced some flashes of newness that allowed us to push back on the collapse of the bubble and spawn new software factory after factory. We could clone that model as we proved it was doable and defensible. Ultimately that cloning process created a culture change that created a program change. I’d say the Air Force became pretty good at software compared to where it was when I entered and I’d put it over anyone else in the government.
That’s the catch 22 of a lot of programs — they can’t show early capabilities to hold off the attacks until they get money, but you can’t get money until you show progress. Kessel Run was irregular in that the AOC 10.2 program was getting cancelled and there were funds available to take a new approach.
In a normal situation, advocates would have had to create a paper plan and sell that plan to get funding in the first place. That plan would have been an easy target without some early progress and functionality like the tanker planning tool.
Other avenues to early-stage activities include the SBIR/STTR program, a few oversubscribed innovation funds, and the S&T 6.3 advanced technology development. None of these, however, would have been a good candidate to support government-led software development.
Leave a Reply