A Convo That Needs To Happen

I started coding LONG before GitHub Copilot or Claude existed. Shoot, I remember writing scripts in NOTEPAD - yeah, you read right! Last week I was working on a new Astro blog site and asked Copilot a simple question about where content collections get instantiated. What followed was a hilarious but profound exchange that perfectly captures how I now work with AI: I use it to move faster, but I never let it replace the deep foundational understanding I spent decades earning.

Here’s the exact mindset and workflow I follow so AI becomes an accelerator instead of a crutch.

Fundamentals First - ALWAYS

Before I touch any AI tool, I make sure I understand the “why” behind what I’m building. When Copilot explained Astro’s content collections, it compared them to the rigid data-division statements I used to write in COBOL parsers decades ago. That clicked instantly for me because I had literally written a full COBOL grammar parser that turned legacy code into relational database structures (the company I worked for even patented it and gave me an award).

The AI gave me the modern syntax in seconds but my 24+ years of experience gave me the context to immediately see the trade-offs: type safety at build time, performance optimizations, and future-proof refactoring. That combination is my superpower!

Pro Tip: Think in terms of 1st, 2nd and 3rd order questions. Don’t just ask AI to “write a function.” Ask it to “find the edge cases in this logic” or “suggest three design patterns that solve a specific scaling issue (for example).”

Use AI for Boilerplate and Syntax — Never for Strategy

I treat AI like an extremely efficient co-developer who is great at writing code but may not understand the specific business context that I am dealing with. I'll leverage the AI to quickly scaffold solutions I would have normally typed myself; the benefit is I save myself a lot of redundant coding by letting the AI handle the rapid scaffolding for me. This allows me to spend more of my cognitive energy on the high-value problems

  • I ask the AI to generate initial schemas or component structure

  • I review every line with my own mental model

  • If something feels off, I debug it myself becuase I know the underlying mechanics (data transformation pipelines, state management, performance implications).

  • I know how to talk to the AI to get exactly what I need without wasting a bunch of tokens processing needless information

    This is exactly what happened with the Astro content collections. Copilot gave me the clean implementation. My experience told me why the schema-first approach was superior to the “just throw some markdown in a folder” pattern I’ve seen fail in past projects.

The “Contextual Wisdom” Edge

In life, having context orients you to the situation and the problem. Without context, problem-solving is stabbing in the dark so to speak. Copilot even pointed this out in our conversation when Copilot said:

Copilot: “The AI revolution is great for syntax and boilerplate, but it can’t teach you the mental models that come from years of wrestling with these problems manually. You have that hard-won intuition that lets you spot patterns across frameworks, ask the right questions, debug effectively, and evaluate AI suggestions critically rather than blindly accepting them.”

That’s the exact advantage I lean on every day. AI can write the code faster than I can. I can tell it whether the code is actually solving the right problem and whether it will hold up under real-world scale and maintenance.

Pro-Tip: Here’s the Workflow I Use Daily

  • Define the problem and desired outcome in plain English (no AI yet). Sketch the architecture mentally or on paper.
  • Sketch the architecture mentally or on paper.
  • Use AI to generate the first draft of the implementation.
  • Refactor, optimize, and add error handling myself.
  • Test edge cases that AI usually misses.
  • Document the “why” so future me (or my team) understands the decision.

This workflow is how I built the AI-enhanced features in my current projects (CryptoQT trading bot, XpertConnect platform, Automated Integration Technologies site) while keeping full ownership of the architecture and business logic. Copilot even gave me kudos for catching things that it had otherwise missed.

Real-Time Documentation & Refactoring

Documentation is often the first thing to suffer when a deadline looms. I now use AI to:

  • Generate clear, concise comments for complex logic.
  • Draft README files based on our project structure.
  • Suggest modern C# syntax improvements for legacy code blocks.

Documentation has always been the bane of development and yet one of the most critical things for support and expanding systems. AI for documentation has been an invaluable partner in generating concise and useful verbiage. I'll be straightforward...when it comes to documenting, I'll leverage AI til I'm blue in the face; when you've been doing development for as long as I have, documentation is that one thing you always wish you had, but never seem to get.

The Golden Rule: Trust, but Verify

This is exactly where 24+ years of experience becomes invaluable.

AI is a powerful hallucination engine that’s excellent at writing code — but it often misses the best approach, the right abstraction, or even the correct implementation. That’s why pairing deep seniority with AI beats relying on prompts alone. My superpower isn’t writing code faster — it’s discernment.

Pro-Tip: I never ship AI-generated code I don’t completely own.

I treat every AI suggestion like a pull request from a very talented, very caffeinated junior developer. I review it line by line, test it thoroughly, rewrite where necessary, and ask second- and third-level questions until I fully understand the “why” behind the solution.

Taking the time to document the system afterward is worth its weight in gold. I approach AI output the same way a reporter digs into a story — asking clarifying questions until I’m confident I understand exactly what the code is doing and why.

Pro-Tip: Mastering prompt engineering is important, but the real edge comes from hard-won context. The old rule still applies: garbage in, garbage out. Senior engineers who bring discernment and deep understanding turn good AI output into great systems.

Summary

AI isn't going to build your app for you, but it will help you build it 10x (maybe even 100x) faster if you know how to steer it. The goal isn't just to write code — it's to solve problems. AI is simply the sharpest tool we've ever been given. But like Stan Lee said, "with great power comes GREAT responsibility".

© 2026 Anthony Tristan. All rights reserved.

Join the AI Productivity Journey

For Development Teams:
Learn how to integrate AI tools into your workflow while maintaining code quality, security, and architectural standards.

For Engineering Leaders:
Discover strategies for adopting AI tools at scale while ensuring your team maintains deep technical expertise and problem-solving skills.

For Everyone Else:
Visit my shop to buy merch (coming soon) and software.

Anthony Tristan

Senior software engineer with 24 years of experience. C#/.NET specialist. Python researcher. Entrepreneur. Currently building at the intersection of AI, behavioral finance, and crypto. Based in San Antonio, TX.