In engineering circles, you’ll sometimes hear stories of 10x or 100x developers who are incredibly more productive than average. This doesn’t mesh well with my own experience. There certainly can be significant differences in productivity between developers, but the biggest differences in productivity are not caused by some innate genetic advantage, but are driven by people working on something they care about, with few dependencies except for ones they know well.
We’ve been doing regular hackdays at Caplin since our first hackday on ‘html5’ back in 2010. Hackdays serve as a sort of ‘1 rep maximum’ for development. We dial down a lot of the complexity that comes from making something of product quality; we dial up the commitment by having projects chosen by the people taking part, and situating it in a friendly, time-bound (24 hours) competitive context. They’re a great way to scratch an itch that has been bugging you — many of our internal debugging tools have come out of hackdays, but they also give engineers a chance to show their own vision for the future of our products and prove their ideas with working code. Many of our products have features first shown in a hackday, or in some cases have entirely come from hackday projects.
Lots of organisations do hackdays or hackday like events, and we’ve recently started to do joint hackdays with other companies. You can read about our hackday with Standard Bank here.
For Hackday 17, we wanted to explore a theme that has been generating a lot of buzz recently – AI, but not just AI in the neural networks and large language model sense, but the broader requirement that we see coming for all software — that it’s more responsive to the needs of the user, more proactive, more understanding of context. All software will need to become smarter. But how could we enable a team that are smart developers but not data scientists to make significant progress in this theme in the space of 24 hours? We knew that we needed tools and expertise to help us.
Like many good ideas, the idea to do a joint hackday with TurinTech came at a party; we’d been thinking about the best way to run a ‘Smarter’ themed hackday, but weren’t sure how to give our teams the best opportunity to succeed. When Mark Tannetta told us about TurinTech.ai and their evoML tool that can automatically generate model code and guides you through the process, it seemed like a perfect match.
TurinTech didn’t just let our teams use their tool, they ran a session a few weeks before to talk about it and how they saw AI in the finance space more generally. On the day itself, they sent over a contingent of data scientists to provide expert advice to our teams. They were a fun and knowledgeable bunch, and a great addition to the vibe of the hackday.
Jiaxuan and Dava worked very closely with the TurinTech team, and produced an impressive integration with one of our products in the 24 hours. They won second place. Our winning team produced a vision for how our products could be made smarter through self-service administration — even getting ChatGPT to help generate the theming. But I’ll let the winning teams explain their projects in their own words in future blog posts.