How Misused AI Is Creating More Tech Confusion Than It Solves 🕸️

Tutorials & Methods
, ,
1

image
image539Ă—303 57.5 KB

How Misused AI Is Creating More Tech Confusion Than It Solves :spider_web:

:brain: Garbage in, Garbage out isn’t just for AI developers anymore.
Now, every user of AI models faces the consequences of misapplied or misunderstood results—often without realizing they’re part of the problem. This deep dive explores a common but overlooked pitfall in modern software engineering: relying on AI-generated code without comprehending the underlying systems.

:spider_web: The Real-Time Problem

A developer tried to implement a real-time data update system using .NET Core 9, Angular 19, and MS SQL Server 2022. The goal was simple: detect database changes (like INSERT) and push updates to the frontend. While traditional polling worked in the past, it was inefficient for bandwidth and performance in a production environment.

The search for a better solution led to AI. With a prompt to Perplexity AI, the developer asked for a detailed setup using SignalR. AI responded with complete steps, required libraries, and code samples. Sounds like magic—until reality intervened.

:warning: Where It All Went Wrong

Despite following the AI instructions and integrating SignalR into QuickApp (a boilerplate project by ebMonney), the system returned a baffling 404 Not Found error. Here’s what was missed:

  • Angular routing hijacked the SignalR /hub endpoint.
  • The fix required editing proxy.conf.js to exclude /hub from Angular’s route handling.
  • Even then, WebSocket connections (wss) must be explicitly enabled via ws: true in the proxy config.

Only after understanding the proxy layer and manually correcting mismatched routes (changing /api/hub back to /hub) did the app finally behave as expected.

:magnifying_glass_tilted_right: AI Isn’t a Shortcut to Understanding

The mistake wasn’t technical—it was conceptual. The developer expected AI to output a turnkey solution. But without understanding the environment (e.g., Angular routing, proxy layers, WebSocket behavior), even perfect code fails.

And when the developer posted their issue and fix on Stack Overflow, it got downvoted for “lack of debugging details.” Ironically, the real debugging insight wasn’t in the logs—it was in knowing the AI-generated steps didn’t account for project-specific setup.

:chart_decreasing: AI Summaries vs. Full Context

This scenario mirrors what teachers face when students submit AI-generated homework. Bullet-point summaries from tools like Perplexity lack the nuance of full content. As Mortimer Adler emphasized in How to Read a Book, digests are poor proxies for understanding.

:robot: AI’s Future Isn’t Just Code Generation

We’re building AI like LLMs and chatbots not just for productivity—but for companionship, assistance, and emotional support. That dream fuels aggressive adoption, even by users who don’t fully understand how AI works. But with that ease comes a danger:

We expect AGI performance from narrow AI models.
We apply AI solutions blindly, then blame the tool when our ignorance shows.

:puzzle_piece: Key Takeaways for AI Users and Developers

  • Don’t copy-paste AI code without understanding.
  • Always verify routing, proxy, and environment-specific logic.
  • Use AI as an assistant, not a crutch.

:pushpin: To avoid misfires in your own stack, review your build configurations, understand your proxies, and treat AI outputs as starting points—not gospel.

ENJOY & HAPPY LEARNING! :heart:

Appreciate the share, Don’t be cheap!

7 Likes