In an exclusive interview with VMblog, Francisco Martin, CEO of Stride, unveiled the company’s groundbreaking product, Stride Conductor – a first-of-its-kind AI solution that leverages large language models (LLMs) to accelerate software development processes. Founded in 2014 by tech trailblazer Debbie Madden, Stride has swiftly risen to become an industry leader in generative AI (GenAI) and custom software development.
Stride Conductor introduces a novel framework where multiple LLM agents collaborate within development environments, enabling semi-autonomous building, enhancing, and testing of software. This innovative approach redefines efficiency and creativity in coding, allowing tasks that typically take developers 4-8 hours to be completed in mere seconds – a game-changer for both new and existing codebases.
According to Martin, Stride Conductor addresses the key bottleneck in today’s software development workflows: the creation of working, well-tested code. By harnessing the power of LLMs to write code and mimic patterns in existing codebases and documentation, Stride Conductor can surgically operate on codebases, translating non-technical requirements into code changes that can be reviewed and approved by human gatekeepers, seamlessly integrating with existing code review and deployment processes.
VMblog: Who is Stride and what is the Stride Conductor?
Francisco Martin: Stride was founded in 2014 by Debbie Madden, a trailblazer in the New York tech community. Headquartered in Manhattan, we also have offices in Seattle, Chicago and the SF Bay Area. Stride stands at the front of technology and innovation. Under Madden’s leadership, we’ve grown the organization into an industry leader specializing in GenAI and custom software development. We create highly tailored software solutions and GenAI agents for enterprise workflows. We continue to grow quickly in headcount and footprint.
Stride has created a first of its kind product for leveraging LLMs to accelerate software development. It’s called, “Stride Conductor.” Stride Conductor utilizes a novel framework where multiple LLM agents collaborate on code in your development environment. Moreover, Stride Conductor enables semi-autonomous building, enhancing and testing of your software. This novel framework and methodology redefines notions of efficiency and creativity in software development. For example, 1 point stories that ordinarily take a developer 4-8 hours to code, now can be written in seconds. This applies to both new and existing codebases.
VMblog: What does Stride Conductor actually do and deliver? Why this, why now?
Martin: Stride Conductor addresses the key bottleneck in today’s software development workflows: the creation of working, well-tested code. Most software pipelines are bottlenecked by developer capacity, and tools like Copilot only marginally improve that constraint. The ability of LLMs to write code and to mimic patterns in existing code and documentation allows Stride Conductor to surgically operate on codebases, translating non-technical requirements and tasks into code changes which can be reviewed and approved by human gatekeepers, just as they do in today’s code review and deployment processes.
LLMs are going to reinvent many industries, but none so significantly as software development. For many companies, the most expensive thing they do is write and maintain code. Entire business processes are designed around how fast development teams can ship code. Agile was a revolution because it brought development cycles down from years to weeks, and CI/CD was a revolution because it shrank deployment cycles from days to minutes. LLMs are going to deliver the same kind of acceleration around going from ideas to code, and making large scale changes which are difficult or impossible for human teams to execute. Companies need a tool that can be customized for their specific needs and environment, and which will help them build better software with AI, not just create larger volumes of bad code. Stride Conductor is that tool, and delivery teams will see immediate benefit in applying it to their workflows.
VMblog: What are the market conditions for Conductor and what’s the competitive field like?
Martin: Gartner’s 2024 outlook highlights the significance of democratized Generative AI (GenAI), which is becoming more accessible due to advancements in cloud computing and open-source initiatives. This is driving significant interest and innovation in technologies such as autonomous code generation and multi-agent collaboration. Although there is strong interest in solutions like GitHub Copilot which focus on developer assistance, we see more and more awareness and demand for solutions such as Stride Conductor, which are more deeply integrated into software workflows and don’t depend entirely on human orchestration. This part of the GenAI market is relatively young but it is maturing quickly, and we see new companies getting into this market almost daily.
VMblog: On the customer side of the equation, what is the business need behind Stride Conductor? And how does it meet those needs?
Martin: Businesses spend a tremendous amount of money building and maintaining their software, and we’ve never met a team which was completely satisfied with their own velocity – you can always move faster, or do more. Creating additional capacity in the delivery process via AI acceleration opens up huge new possibilities, either for efficiency or new product/feature development. So many teams are held back by technical debt that costs them real money, either in terms of expense to run bloated systems or inability to take advantage of new tools and approaches. And expanding the capacity to create means that instead of spending time prioritizing your build pipeline, you can build quickly and focus on curating the best ideas and features. Stride Conductor removes a key business bottleneck by expanding throughput in software delivery teams.
VMblog: There are a lot of objections in the market about autonomous code generation in regard to the role of the software developer role in the future. What is your take on that and how does Stride Conductor differ from the competition on that?
Martin: We are software developers, first and foremost. But the developer role is about much more than writing code. As several of our customers have told us directly, “If you think your primary job is writing code, you’re missing the point.” Software development is about framing and delivering solutions to customer problems, with high quality and manageable cost. We think AI tools improve our ability to deliver this kind of value to our customers.
That said, it’s also our job as technology leaders to chart a future where humans are empowered to define their role in relation to AI. The most interesting and satisfying software work should still be done by humans in concert with AI, whereas AI can autonomously take on the challenges that are tedious or unpleasant. Every software team has work they don’t like doing but is essential to success, and which is formulaic and predictable enough to require minimal oversight. This kind of work is perfect for semi-autonomous AI tools like Stride Conductor, and we encourage teams to focus on improving the quality of life for their developers, rather than thoughtlessly replacing humans with AI, which also carries tremendous business risk.
VMblog: How are you dealing with issues such as AI hallucinations and AI code quality?
Martin: Stride Conductor uses your existing codebase, preferred code patterns and documentation to extend and enhance your existing code. Hallucinations most commonly occur when LLMs aren’t given sufficient context of what a “good” response looks like, so by providing extensive code context we can minimize their occurrence.
Stride practices TDD (Test Driven Development) – which in an LLM context is very useful for verifying code quality. Our multi-agent approach allows for self-correction as agents break down requirements into tests and code, and we also provide several opportunities for human verification throughout the process. But generally, we find that providing good context up front and robust tests minimizes issues with generated code.
VMblog: These are challenging times on a number of fronts. Why would you decide to launch a new GenAI product to market now amidst such turbulence and also heavy established competition?
Martin: We beg to differ – there is very little established competition. We are not model-builders, and do not intend to compete with anyone on a better LLM for coding (although we do think we can innovate on providing code context to LLMs). The majority of AI-powered development tools are human assistants with very little autonomy. This is largely because an interactive assistant can be generically applied to most software dev workflows, whereas true automation is much more difficult and customized. We see a real opportunity in leveraging LLMs based on our experience as custom software developers, who are familiar with a wide range of client environments and workflows (including security conscious and/or regulated industries). Put simply, there are many companies that wouldn’t touch an off the shelf tool, but would benefit tremendously from a thoughtful partnership with Stride that leveraged our experience and our tools.
VMblog: What are Stride’s plans for the future as a company? There’s a lot of serious competition in this space, what makes you confident you’ll be able to prevail with Stride Conductor, and what comes next?
Martin: From Day 1 in the GenAI space, Stride has distinguished itself by doing the real work of applying AI to our customer problems. We’ve built and deployed LLMs inside secure clouds, put together POCs for redefined workflows, and leveraged our custom software experience to reimagine how code is built and shipped. Many of our competitors are invested in demonstrating various capabilities but Stride has gone above and beyond, focusing on real world applications. We are positioned to ride the coming wave of adoption. With a combination of Stride Conductor and Stride’s deep understanding of software development, we help drive real transformation in how it is built, and stay ahead of the curve by focusing on the tasks our customers need done.
##
