From Old Instincts to New Insights: React Native, Enterprise Dev & AI at Scale

I recently found myself revisiting one of my earliest dev lessons: don’t cling too tightly to the first frameworks you learn. In a recent LinkedIn post, I shared how re-learning that lesson led me back into React Native. What followed was an incredible exchange of ideas that reshaped how I think about systems, architecture, and AI in modern development.

This article reflects what I learned from the conversation, not just what I initially believed. These are field notes from people working in real systems, with real scale, and real constraints.

What AI Accelerates vs. What It Exposes

"Connecting all the parts together is the real challenge. People have been googling syntax and patterns since the dawn of time – it's just faster now."
-- Todd Shevlin, Director of Platform Technologies, CypherTax

Todd said what most of us feel in practice: AI helps write code, but it doesn't understand systems. The friction lives in integration, not syntax.

React Native might generate a component for you, but getting that component to work across services, platforms, auth layers, and legacy backend systems still requires real judgment.

Bullet Points: Where AI struggles

• Environment configuration
• Data validation across layers
• State management in production
• Understanding system intent, not just code output

Database Design Is Still an Art (and a Risk)

"A well-designed database takes a lot of thought and is somewhere between art and science."
-- Todd Shevlin

This stuck with me. Database decisions often get buried under the speed of delivery. But unlike code, they’re not cheap to fix. Schema changes create ripple effects. Deletion becomes scary. Cleanup becomes a business risk.

Key Takeaway: The cost of poor data modeling shows up late, but hits hard.

Platforms Over Gatekeepers

"You just need a unified API layer and an enterprise template, and then everyone builds on the same backend... Gatekeepers kill software. Now AI kills gatekeepers."
-- Sam Wu

Sam brought a bold perspective: if your architecture is solid, you can let go of control. The API layer becomes the control plane. Governance becomes embedded. Gatekeeping shifts from people to protocols.

When this works:

• Backend APIs enforce consistency
• Contracts are clear and versioned
• Internal tooling makes the right thing easy

I loved this: It’s not about removing discipline, it’s about shifting where discipline lives.

Legacy Code and the Reality Check on AI

"AI is impressive with new development, but throw it at millions of lines of code with siloed processes and undocumented business logic and suddenly that 'just prompt it better' advice falls apart."
-- Andrey Martinkov

Andrey said what a lot of people miss: the real problem isn’t writing new code. It’s understanding old systems. The kind of systems where knowledge lives in heads, not docs. Where logic is implicit.

This is where AI stalls:

• Legacy systems
• Undocumented flows
• Business logic buried across services

A Moment of Humor (and Truth)

"So we’ve got a few years left after all :)"
-- Rob Levin, Senior Software Engineer, MasterCard

Rob’s comment was short and warm, and also right. AI helps. But the judgment, decision-making, and experience behind architecture still belong to people.

My Takeaways

What This Conversation Taught Me

• AI makes execution faster, not easier
• Architecture still decides outcomes
• Data modeling is foundational, not optional
• Speed without structure is just acceleration toward entropy
• Good systems amplify AI’s value; bad ones magnify its risk

AI removes human gatekeepers only if architecture replaces them.

Thank You to the Contributors

This piece was shaped by the insights and generosity of:

• Todd Shevlin
• Sam Wu
• Andrey Martinkov
• Rob Levin

Thank you for pushing my thinking further. If you have your own thoughts on React Native, enterprise dev, or AI workflows, I’d love to hear them. This conversation is still open.

‍