Salvatore Sanfilippo (antirez), creator of Redis and one of the world’s most respected leaders in the field of system engineering, recently published a thoughtful analysis of AI’s impact on programming. His conclusion? AI hasn’t killed programming, it’s transformed it from code writing to system thinking. At MosierData, we’ve been in the development business for 20 years and have built everything from foster care management systems to logistics platforms handling $10M+ annually. We’re not threatened by AI coding tools. In fact, we’re excited about them. But there’s a critical gap between generating code and deploying production software that serves Central Florida businesses, non-profits, and organizations reliably for years. This gap involves security architecture, system integration, data integrity, compliance requirements, and liability protection that AI tools alone simply cannot address.
The Honest Reality: AI “Vibe” Coding is not System Engineering
Let’s start with what antirez got exactly right: AI tools like Claude Code, ChatGPT, and Cursor can now generate functional code in minutes that would have taken days or weeks to write manually. In his article, he describes building a UTF-8 library, debugging Redis test failures, creating a pure C inference engine, and implementing complex Redis Streams changes. All with AI assistance, all dramatically faster than traditional coding.
This is real. This is happening. And anyone telling you otherwise is either uninformed or selling you fear.
But Here’s What He Also Said (That Gets Less Attention) Buried in antirez’s analysis is a critical insight: “It is now a lot more interesting to understand what to do, and how to do it.”
That sentence is everything.
The Three Layers AI Can’t Handle Alone
1. System Architecture (“What to Build”) AI can generate code when you describe what you want. But who decides what to build? At MosierData, when a Central Florida company or non-profit comes to us with spreadsheet chaos and manual processes, they don’t need code, they need someone to understand their workflows, identify bottlenecks, design data models, and architect a system that will scale with their mission for the next decade.
Example: Heartland for Children needed an inter-county foster care case management system. The challenge wasn’t writing code, it was understanding how multiple counties coordinate care, how privacy regulations apply, how caseworkers need to share information securely, and how to design a system that protects vulnerable children while empowering staff. That’s system engineering, not code generation.
2. Integration Reality (“How It Fits”) Real businesses don’t operate in isolation. Your custom software needs to integrate with QuickBooks, Salesforce, your existing database, your email system, your payment processor. It needs to handle edge cases, manage transactions, validate data, log errors, and degrade gracefully when third-party services fail.
AI can write API calls. It cannot architect a fault-tolerant integration layer that handles the messy reality of production environments. We’ve been doing this for 20 years, we know where things break, and we design accordingly.
3. Security & Liability (“What Keeps You Protected”) Here’s the uncomfortable truth: We regularly see business owners deploy AI-generated code into production who:
- Are not professional programmers
- Don’t understand application security fundamentals
- Have no knowledge of OWASP Top 10 vulnerabilities
- Lack cyber liability insurance appropriate for software development
- Don’t grasp their legal exposure when a data breach occurs
You’re insured for your line of business. You’re likely not insured for software development liability. When AI generates code with SQL injection vulnerabilities, cross-site scripting holes, or inadequate authentication, and your customer data gets compromised, who’s liable? Spoiler alert: It’s you.
How We Actually Use AI (The Pragmatic Approach)
At MosierData, we’re pro-AI efficiency. Here’s our actual workflow:
Phase 1: Discovery & Architecture (Human-Led) We map your workflows, identify bottlenecks, design data models, and create system architecture. This is where years of experience matters. We know what works in production environments across Central Florida organizations.
Phase 2: Accelerated Development (AI-Assisted) We use AI tools to generate boilerplate code, implement standard patterns, and accelerate routine development tasks. This is where we gain efficiency, building in days instead of weeks.
Phase 3: Integration & Security Hardening (Human-Led) We architect integrations, implement security best practices, conduct penetration testing, and ensure code meets enterprise standards. AI can’t do this as it requires expertise and judgment.
Phase 4: Testing & Deployment (Human-Led) We test edge cases, handle data migration, train your team, and provide ongoing support. This is where system engineering expertise protects your investment.
The Result: We deliver production-ready software faster than traditional development, but with the reliability and security that comes from experienced, professional, system engineering experience and oversight.
What This Means for Central Florida Organizations
If you’re considering custom software:
✓ Great idea to explore AI-assisted development for efficiency
✗ Terrible idea to deploy AI-generated code without professional review
✓ Leverage AI tools for prototyping and rapid iteration
✗ Skip security audits, penetration testing, and compliance reviews
✓ Partner with developers who use AI to accelerate delivery
✗ Assume AI eliminates the need for system engineering expertise
We’re hosting free 30-minute Clarity Calls for Central Florida organizations exploring custom software. We’ll be honest about whether custom development is right for you, how AI tools fit into the process, and what a pragmatic system engineering approach looks like. No pressure, no sales pitch. Just a conversation about what your organization actually needs.
Book your Clarity Call: https://mosierdata.com/clarity/
