Why we invested in Recurse ML
TL;DR We’ve invested in Recurse ML to provide code maintenance in a single command
How it started
Jack Jackson and Armins Stepanjans, met at Entrepreneur First in London as part of the April to September 2023 cohort. A couple of months later, they landed on the idea for Recurse ML and have been building ever since.
Entrepreneur First reached out to Joe on the wider Playfair team to put Recurse ML on our radar. The remarkable traction Jack and Armin’s solution has seen in just a few months, and the scale of the problem they are solving, made it clear that this was an opportunity we couldn’t ignore.
The problem with maintaining and updating software in large organisations
Dependencies are external libraries or components that a software project relies on to function. Pre-written pieces of code, they provide specific functionalities, such as logging, data processing, or database connectivity, without requiring developers to write these features from scratch.
The problem is that in large organisations, developers can spend 18–24 months updating and maintaining these dependencies. This task is challenging due to the volume and complexity of code in legacy systems. Manually managing these updates is labour-intensive, time-consuming, prone to errors, and notoriously boring. It can delay the implementation of new features and impact overall productivity.
Advancements in AI could potentially double the amount of code engineers can generate, making it even more time-consuming to keep dependencies up to date. The need for efficient dependency management will only become more critical in the future.
Recurse ML’s solution
Recurse ML automates the dependency management process, using machine learning to:
- Automatically identify outdated dependencies
- Seamlessly update these dependencies
- Ensure compatibility through automated testing
For example, if a bank uses Java-based applications, Recurse ML software can automatically detect which parts of the code need updates and perform those updates seamlessly. This saves time, reduces the risk of human error, and ensures that the software is always up-to-date.
Here’s how it works:
Playfair’s thesis
An exciting, valuable enterprise opportunity with direct ROI
Resolving dependency issues for large enterprises is a critical, yet unsolved, challenge. The resulting efficiencies and cost savings could provide substantial value and a compelling ROI for clients. They can directly equate the time saved on dependency management to developer salary costs. Given these significant time and cost savings, Recurse ML’s solution could lead to annual contract values worth millions per client.
Real pull from the market
Top-tier banks and major asset management firms are already interested in Recurse ML. Given that the business is only months old and the product is still in its early stages, this highlights the severity of the dependency management problem, and Jack and Armin’s commercial and technical credibility.
Strong differentiation
Most competitors in the dependency management space focus on identifying outdated dependencies and automating the creation of pull requests for updates. However, they do not handle the actual code changes required for these updates. Developers still need to manually review, test, and integrate the changes.
Recurse ML stands out by offering automated code generation, which accounts for 90–95% of the dependency upgrade process. This significantly reduces the manual effort developers need to spend on these tasks. Recurse ML is a first-mover here, going beyond current solutions that only flag vulnerabilities or automate simple version updates to actually writing the code for developers.
Finally, Recurse ML are the only solution that can adapt to enterprise codebase complexity. Rule-based approaches (e.g. highly customised environments, custom JDK-class extensions, custom JVM implementations, bespoke coding styles) just don’t work.
Smart, driven founders
Jack and Armin deeply understand the dependency management issue. They’ve achieved impressive early traction with enterprise customers, driven by their technical expertise and ability to build effective solutions. They have the drive and vision to build a new entrant in this space and attract talent.
Jack read for a DPhil in Cyber Security from the University of Oxford before suspending it to found Recurse ML. His ability to engage with top-tier clients highlights his deep understanding of enterprise customer needs and his ability to drive business growth.
Armin earned an MPhil with distinction from the University of Cambridge. His experience spans from developing scalable software systems to leading engineering teams, making him well-versed in technical leadership.
The rise of AI-generated code
The founders convinced us that as foundational models and large language models (LLMs) evolve, the speed of software development will more than double. This increased productivity will exacerbate the already large dependency problem.
What’s next?
Recurse ML’s vision is to create a new programming paradigm, where codebase maintenance is fully abstracted away from the programmer by ML agents. This includes addressing other issues related to codebase maintenance, such as regression testing (ensuring that new code changes do not introduce errors into existing functionalities), complex code refactoring (improving the internal structure of existing code without altering its external behaviour) and migrations (transitioning codebases to new platforms, languages, or frameworks).
In the long term, they’ll use AI to empower enterprises to innovate and release new products and features at the speed of startups.
If you’re looking to automate codebase maintenance, you can reach out to the Recurse ML team directly here.
If you’re a founder at the very earliest stages of building your company, I’d love to connect and hear the vision. You can find me on LinkedIn or pitch the wider team at Playfair here.
You can follow the Playfair team on LinkedIn, Twitter, Forbes, Vimeo and here on Medium.
