AI Development Guidelines

Practical guidance for using AI during Software Development. These are a work in progress as I learn how to use AI in daily development work.

Do’s

• Refer to your company’s policy
• Think about the system
• You are responsible for creating a maintainable, working system
• Think - pay attention to each step
  • “Stop Babysitting, Start Leading” ~ https://dev.to/datadog-frontend-dev/ai-coding-is-boring-and-what-to-do-about-it-5bji?context=digest
  • Consider using Spec Kit
• Understand what you need to build
• Write specific prompts
• Review generated code at each step
• Keep learning and improving
• Find pragmatic uses of AI
  • https://itrevolution.com/articles/youre-head-chef-not-a-line-cook/
    • One take, leaning towards “vibe” coding
  • https://articles.pragdave.me/p/the-sorcerers-apprentice
    • Keep your “genie” (Kent Beck’s name for it) / spell under control
    • “Never get an AI to write code that you couldn’t have written yourself.”
    • “AI is a tool that amplifies your ability to code. It does not replace your understanding of what needs to be done, and it won’t take the blame if what it produces is wrong. You have to know the basics. You have to know design, and techniques, and performance, and the thousand other things, both large and subtle, that make you a developer.”
• Strive for efficiency
• Get an Business or Enterprise Copilot license for organization work
  • Addresses privacy and legal concerns (we still have to trust Microsoft)
• Use AI to help create automated tests and increase confidence and coverage
• Stop, build, and run tests after changes
• Use specific and clarifying prompts
• Use Copilot to help you learn
• Use Copilot for syntax and common approaches to save time
• Use Copilot to help update projects
• Use Copilot to identify
• Experiment with the tools
  • Share what you learned and experienced
  • Provide feedback to your teammates
  • Ask why we are doing things this way
  • Identify constraints and help eliminate them
• Consider using Spec Kit for Spec Driven Development to guide development

Don’ts

• Paste specific code, IP, or company secrets into ChatGPT and others
• Let AI do all your work and thinking
• Don’t assume generated code is correct — verify, test, and review
• Blindly accept changes
• Rapidly create technical debt with AI
  • GenAI is the new Offshoring | Blog - Ardalis
• Avoid using unvetted AI output in production without review
• Don’t rely on AI for legal, security, or privacy decisions without human approval

IT Revolution Guidelines

I added this on February 6, 2026 after reading “No Vibe Coding While I’m On Call”: What Happens When AI Writes Your Production Code

The story looks very interesting. This article has a link to a pdf you can download along with lessons for each section.

I quote from the end of the article:

The Path Forward: From Chaos to Discipline

After a string of incidents and the refrain “No vibe coding while I’m on call,” the team shifts from speed-at-all-costs to disciplined AI use. The goal isn’t to abandon AI, but to add guardrails.

Pre-Flight Checklists for AI-Assisted Development

• Break work into small, verifiable tasks with explicit operational requirements
• Add architectural oversight so AI suggestions align with resilience patterns
• Use AI to critique its own changes before accepting them
• Require security, scalability, and error-handling constraints up front

The AI Development Framework

• Co-author detailed specs with AI, including operational requirements
• Maintain prompt libraries that preserve system architecture
• Add automated checks to flag violations of architectural principles
• Run real-time, cross-functional reviews (product, security, QA, ops)

The Tools

• Fast, local automated tests that run in seconds
• Observability built into every feature from day one
• Canary releases with automatic rollback based on real metrics
• Operations data made available to AI agents for analysis

By the end, the on-call engineer can sleep again. The lesson: AI doesn’t replace judgment - it amplifies it.

Source: https://itrevolution.com/articles/no-vibe-coding-while-im-on-call-what-happens-when-ai-writes-your-production-code/

The bottom line: In DevOps, “You build it, you run it.” But what happens when GenAI is doing most of the building? You need to run it even more carefully, with even better observability, even faster feedback loops, and even more disciplined engineering practices.

The “vibe coding” chaos can give way to structured, disciplined, AI-assisted engineering. But only if you build the guardrails before you need them—not after your on-call engineer has been paged at 2 a.m. for the fifth time in a week.

Instruction Resources

• https://github.com/github/awesome-copilot
• https://codemag.com/event/CP-07-2025
• https://tidyfirst.substack.com/p/augmented-coding-beyond-the-vibes?publication_id=256838&post_id=166781850&isFreemail=true&r=84e0h&triedRedirect=true
• https://www.totaltypescript.com/cursor-rules-for-better-ai-development

Other Interesting Links

• Programming Deflation by Kent Beck
• Has This Report EXPOSED THE TRUTH About AI Assisted Software Development? - From Modern Software Engineering on the DORA 2025 Report
  • The fundamentals of Programming
    • A Precise Specification of What We Want
    • Verification that we Got What We Want
    • The ability to make progress in small steps
      • Source Control
    • The DORA Model reflects this
  • small change, verify it’s doing what it should be doing