How Software Developers Should Be Building With AI

How Software Developers Should Be Building With AI

Developers are special. Because software application developers have a particularly condensed way of reasoning, rationalising and even eating (pardon the pizza joke), they play a highly valued role in the modern age of information management and data science. While the rise of artificial intelligence has been variously described as a sort of death knell for programmers (the suggestion being that we won’t need them anymore), there is an arguably more likely reality which sees AI tools support, augment and extend developer functions.

The question then is, how should software developers be building with AI?

Tough Toolset

Despite the AI hype cycles that fuel media and marketing, there’s no visible sign of automation fatigue just yet; that might be because we’re still working out how to apply AI at ground level on the developer “command line”, even if coders are using increasingly abstracted low-code or no-code tools. The advantages have been described as a force multiplier for developers. Why? Because AI coding functions can handle jobs like refactoring code (debugging and cleaning up existing code without changing its actual function to make it more efficient), generating new code, translating code from one programming language to another and more.

In this space, AI-powered tools like Cursor.ai are getting rid of developers’ dirty chores and gaining a lot of traction. The technology is essentially a software code editor with features such as AI code autocompletion and it offers an AI chat service for programmers to ask questions of it.

“In this new world of AI-fuelled development, we believe it’s important to embrace change. The AI empowerment era has started, but importantly, it’s intersecting with Kubernetes [cloud orchestration technology] as a de facto hybrid multi-cloud infrastructure control plane and the standard for application delivery,” said Randy Bias, vice president & VP of open source strategy and technology at hybrid multi-cloud company Mirantis.

Bias suggests that combining AI-powered tools like Cursor.AI for developers and GPTscript (a software tool that lets developers use natural language as simple sentences using normal human syntax) with AI-enabled Kubernetes management tools, the industry can delivery time for new applications dramatically.

Eliminating Historical Toil

“We will also be able to ramp up feature velocity, decrease deployment times, remove risk during upgrades and generally remove much of the historical toil that has reduced software engineer effectiveness. Why do the dirty work yourself when you can have the computers do it for you? We believe deeply on the intersection of these technologies and what they enable for the modern developer,” said Bias.

Continuing this thought process then, we can suggest that as AI transforms software development, companies face a challenge in terms of how to they should reliably integrate large language models (and the generative AI functions they deliver) into their applications. This isn’t a question whether a business “installs some AI”, it’s a more fundamental question related to how well it is architected in.

“Developers are racing to customize generic LLMs with domain-specific content and real-time updates. This make-or-break challenge demands two crucial components: sophisticated retrieval augmented generation pipelines for customization and robust conversational application programming interfaces for interaction,” proposed Mark Fussell, co-found and CEO of Diagrid, the company behind the open source Dapr (distributed application runtime) project that aims to simplify the development of cloud-native applications.

Fussell says that RAG pipelines (the whole creation and management of retrieval augmented generation functions in the AI universe) require orchestration across multiple stages: data ingestion, processing, embedding generation and vector storage. Without proper orchestration, teams risk pipeline failures, inconsistent updates and maintenance nightmares.

“The conversational API layer is equally critical, managing everything from stateful interactions to security controls. It’s a shield against vendor lock-in and a gateway to reliable AI interactions,” said Fussell. “Together, these components form the backbone of production-ready AI applications. In today’s AI-driven world, the difference between success and failure often comes down to this fundamental architecture.”

Methodology Of The Month

There’s a lot to take in here, primarily because technology platforms are still evolving so rapidly and – in many cases – we’re on the intersection of truly cloud-native deployments now dovetailing with AI intelligence and automation. We also have factors like DevOps (for unified developer-operational harmony), FinOps (for cloud cost control) and DevSecOps (the sec here being security) all vying for attention as methodology of the month.

So of course, some white noise is inevitable.

In his role as chief evangelist, Cloud Foundry Foundation – the body behind an industry-standard open source cloud application platform – Ram Iyengar reminds us that compute execution now varies between edge, cloud, fog, bare metal, IoT, quantum and others. Further, he notes that computing workloads can be run as containers, unikernels, WASM binaries, on virtual machines and in other immutable ways – all in a world where exploitation of software supply chain vulnerabilities is at a record high.

“In such a noisy and distracting universe, developers need ways to help them publish an app irrespective of the target,” asserts Iyengar, in positive terms. “There isn’t an easy way to compartmentalize an application to run on a single stack. Those with that luxury are considered lucky. Automation is becoming ever more important. More aspects of development are being controlled through automation frameworks. Rather than task a developer, teams are finding ways to build platforms which encapsulate more of these frameworks. Security and compliance are easier dealt with using intelligent tools. Write once, run anywhere has a considerably new meaning. For the next decade, the secret sauce is the developer platform.”

Consumption, Coordination, Creation

While it’s still tough to define exactly how software developers will be using AI in the near and immediate future at every level, any suggestion that the role itself will be completely automated within the current decade seems a little premature.

The current technology industry focus is more concerned with how developers will be able to consume and coordinate AI-powered services in the creation process. We probably need to get those baby steps right first before we ship all our programmers off to the farm.

In basic terms, AI programming tools can fix a lot, but they need to be affixed first.

link

Leave a Reply

Your email address will not be published. Required fields are marked *