Software in the Age of AI – O’Reilly

In 2025 AI reshaped how groups assume, construct, and ship software program. We’re now at some extent the place “AI coding assistants have shortly moved from novelty to necessity [with] as much as 90% of software program engineers us[ing] some type of AI for coding,” Addy Osmani writes. That’s a really completely different world to the one we have been in 12 months in the past. As we look forward to 2026, listed below are three key traits we’ve seen driving change and the way we predict builders and designers can put together for what’s forward.

Evolving Coding Workflows

New AI instruments modified coding workflows in 2025, enabling builders to put in writing and work with code quicker than ever earlier than. This doesn’t imply AI is changing builders. It’s opening up new frontiers to be explored and abilities to be mastered, one thing we explored at our first AI Codecon in Might.

AI instruments within the IDE and on the command line have revived the debate in regards to the IDE’s future, echoing previous arguments (e.g., VS Code versus Vim). It’s extra helpful to deal with the instruments’ objective. As Kent Beck and Tim O’Reilly mentioned in November, builders are in the end accountable for the code their chosen AI instrument produces. We all know that LLMs “actively reward current prime tier software program engineering practices” and “amplify existing expertise,” as Simon Willison has identified. And a superb coder will “factor in” questions that AI doesn’t. Does it actually matter which instrument is used?

The crucial transferable talent for working with any of those instruments is knowing find out how to talk successfully with the underlying mannequin. AI instruments generate higher code in the event that they’re given all of the related background on a mission. Managing what the AI is aware of about your mission (context engineering) and speaking it (prompt engineering) are going to be key to doing good work.

The core skills for working effectively with code won’t change within the face of AI. Understanding code assessment, design patterns, debugging, testing, and documentation and making use of these to the work you do with AI instruments would be the differential.

The Rise of Agentic AI

With the rise of brokers and Mannequin Context Protocol (MCP) within the second half of 2025, builders gained the flexibility to make use of AI not simply as a pair programmer however as a whole crew of builders. The audio system at our Coding for the Agentic World reside AI Codecon occasion in September 2025 explored new tools, workflows, and hacks which can be shaping this rising self-discipline of agentic AI.

Software program engineers aren’t simply working with single coding brokers. They’re constructing and deploying their very own customized brokers, usually inside advanced setups involving multi-agent scenarios, teams of coding agents, and agent swarms. This shift from conducting AI to orchestrating AI elevates the significance of actually understanding how good software program is constructed and maintained.

We all know that AI generates better code with context, and that is additionally true of agents. As with coding workflows, this implies understanding context engineering is crucial. Nevertheless, the differential for senior engineers in 2026 will probably be how nicely they apply intermediate abilities reminiscent of product considering, superior testing, system design, and structure to their work with agentic programs.

AI and Software program Structure

We started 2025 with our January Superstream, Software Architecture in the Age of AI, the place speaker Rebecca Parsons explored the architectural implications of AI, dryly noting that “given the tempo of change, this may very well be outdated by Friday.” By the point of our Superstream in August, issues had solidified a bit extra and our audio system have been capable of share AI-based patterns and antipatterns and clarify how they intersect with software program structure. Our December 9 occasion will have a look at enterprise structure and the way architects can navigate the affect of AI on programs, processes, and governance. (Registration continues to be open—save your seat.) As these occasions present, AI has progressed from being one thing architects might need to contemplate to one thing that’s now important to their work.

We’re seeing profitable AI-enhanced architectures utilizing event-driven models, enabling AI brokers to behave on incoming triggers relatively than mounted prompts. This implies it’s extra necessary than ever to know event-driven structure ideas and trade-offs. In 2026, subjects that align with evolving architectures (evolutionary architectures, health capabilities) may also grow to be extra necessary as architects look to search out methods to modernize current programs for AI with out derailing them. AI-native architectures may also deliver new concerns and patterns for system design subsequent 12 months, as will the development towards agentic AI.

As was the case for his or her engineer coworkers, architects nonetheless must know the fundamentals: when so as to add an agent or a microservice, find out how to take into account value, find out how to outline boundaries, and find out how to act on the information they have already got. As Thomas Betts, Sarah Wells, Eran Stiller, and Daniel Bryant word on InfoQ, “additionally they need to understand how an AI component pertains to different components of their system: What are the inputs and outputs? How can they measure efficiency, scalability, value, and different cross-functional necessities?”

Firms will proceed to decentralize duties throughout completely different capabilities this 12 months, and AI brings new units of trade-offs to be thought-about. It’s true that regulated industries stay understandably cautious of granting entry to their programs. They’re rolling out AI extra fastidiously with larger guardrails and governance, however they are nonetheless rolling it out. So there’s by no means been a greater time to know the foundations of software program structure. It is going to put together you for the complexity on the horizon.

Sturdy Foundations Matter

AI has modified the way in which software program is constructed, nevertheless it hasn’t modified what makes good software program. As we enter 2026, an important developer and structure abilities gained’t be outlined by the instrument you realize. They’ll be outlined by how successfully you apply judgment, talk intent, and deal with complexity when working with (and typically towards) clever assistants and brokers. AI rewards sturdy engineering; it doesn’t change it. It’s an thrilling time to be concerned.

Be a part of us on the Software program Structure Superstream on December 9 to learn to higher navigate the affect of AI on programs, processes, and governance. Over 4 hours, host Neal Ford and our lineup of consultants together with Metro Financial institution’s Anjali Jain and Philip O’Shaughnessy, Vercel’s Dom Sipowicz, Intel’s Brian Rogers, Microsoft’s Ron Abellera, and Equal Specialists’ Lewis Crawford will share their hard-won insights about constructing adaptive, AI-ready architectures that assist steady innovation, guarantee governance and safety, and align seamlessly with enterprise targets.

O’Reilly members can register here. Not a member? Sign up for a 10-day free trial earlier than the occasion to attend—and discover all the opposite sources on O’Reilly.