Building Modulewise

Software engineering is at a crossroads. AI isn’t just another tool. It’s forcing us to rethink how we build systems.

Many organizations are starting down the evolutionary path: adding language models alongside databases, prompts alongside queries, and chatbots in their UIs. But the bigger opportunity follows a more revolutionary path: AI agents driving dynamic behavior from the middle out.

That path creates new opportunities and new risks. But it is untraversable in the traditional model of applications, with their siloed and hardwired views and queries. Even fine-grained microservices aren’t flexible enough, because they are still fundamentally assembled stackwise.

The Wedge

Agentic middleware will reach its full potential only when it splits apart presentation and data access layers, freeing them to be used in adaptive context-driven interactions: business data and generated content combined on-demand to support user experiences that are exploratory rather than predetermined.

But this only works if that split reveals the well-defined boundaries of a modular system.

The Essence

Modular systems are adaptive by design. They’re composable and observable because they’re built on explicit interface contracts. They support event-driven and asynchronous interactions. They favor horizontal integration over vertical stacks. But unfortunately, modular systems remain uncommon compared to the status quo of stackwise applications interacting through request/reply networking.

With the inclusion of AI in its varying degrees of autonomy and its powerful yet nondeterministic behaviors, the most successful organizations will be those who do build modular systems. No longer just a nice-to-have, modularity is essential to seizing the opportunities and managing the risks that come with agentic integration.

The Shift

Building modulewise requires thinking modulewise. Like any paradigm shift, it first requires a mindset shift.

Thinking modulewise transforms collaboration:

• Software teams should collaborate throughout the full lifecycle, not just at handoffs.
• Systems must be observable and securable at every boundary, not just the surface.
• AI collaborators amplify both the opportunities and the risks, making zero-trust and least-privilege principles more important than ever.

Thinking modulewise drives efficiency:

• Components are the primary artifacts, aligned to each contributing engineer’s expertise.
• Reusable components avoid the duplication of undifferentiated code.
• Swappable components enable continuous composition in dynamic environments.

Thinking modulewise means prioritizing:

• Systems over Applications
• Interfaces over Infrastructure
• Components over Containers

What’s Next

Thank you for reading this introduction to the Modulewise vision. The next post will begin to explore what “Composable Integration for Intelligent Systems” means in practice. In the meantime, you can find Modulewise on LinkedIn and X.