AI Leverage

Comparing Level 1 and Level 2

Level 1

Level 2

Description

At this stage, you're learning to work alongside AI tools without depending on them. You use AI to explore code, draft solutions, and make sense of unfamiliar patterns—but the core skill you're building is reading AI output critically, even when you're still learning the patterns yourself. Nobody feels fluent yet, and that's normal. What matters most is curiosity: questioning what AI gives you rather than accepting it at face value, and pursuing real understanding even when shortcuts are available.

Description

At this stage, AI is a genuine productivity multiplier—but speed makes your judgment more consequential. You use AI fluently for drafting, testing, and exploring approaches, and you've started recognizing that "works" and "right" are different questions. You review AI output with the same rigor you'd apply to any pull request, because you understand that vague thinking plus fast tools produces polished-but-wrong results. The discipline isn't in using AI. It's in thinking clearly about what it produces.

Key Behaviors

  • •
    Uses AI tools to explore unfamiliar code and generate first drafts
  • •
    Reviews AI-generated output before committing or sharing it
  • •
    Asks whether AI suggestions match the project's patterns and conventions
  • •
    Seeks to understand the code AI produces, not just use it
  • •
    Treats AI output as a starting point, not a finished product

Key Behaviors

  • •
    Uses AI to accelerate routine work—boilerplate, tests, documentation drafts
  • •
    Reviews AI output with the same rigor applied to any pull request
  • •
    Iterates on AI suggestions rather than accepting the first response
  • •
    Recognizes AI failure modes—hallucinated APIs, cargo-culted patterns, plausible-but-wrong logic
  • •
    Knows which tasks AI accelerates and which it complicates

Common Struggles

  • Accepts AI output that "works" without asking whether it's the right approach
  • Misses subtle errors or cargo-culted patterns in generated code
  • Skips building understanding because AI provides quick answers
  • Has difficulty distinguishing good AI suggestions from plausible-but-wrong ones

Common Struggles

  • Trusts AI output too quickly when under time pressure
  • Produces more code than can be carefully reviewed
  • Confuses AI fluency with engineering judgment
  • Misses context-specific issues that AI doesn't know about

Success Indicators

  • Read AI-generated code critically before using it
  • Build real understanding of the patterns AI suggests, not just surface familiarity
  • Ask questions about AI output the same way you'd ask questions about any code you didn't write
  • Pursue learning deliberately, even when AI makes it tempting to skip ahead

Success Indicators

  • Use AI as a productivity multiplier without sacrificing code quality
  • Review every piece of AI output with the same care as human-written code
  • Develop instinct for when AI is likely to be wrong or incomplete
  • Bring clear thinking to prompts and iteration—not just speed

Mindset Shift

From:

"AI gives me the answer."

To:

"AI gives me a starting point that I need to understand and evaluate."

Mindset Shift

From:

"AI helps me code faster."

To:

"I direct AI toward good outcomes, and I know which context I'm operating in."

Questions to Ask Yourself

  • Do I understand why this code works, or just that it works?
  • Would I be able to explain this approach to someone else?
  • Am I building real knowledge, or just getting things done faster?

Questions to Ask Yourself

  • Did I review this output thoroughly, or am I trusting it because it looks right?
  • Is this the right approach, or just an approach that works?
  • Am I moving faster with better results, or just faster?

Build These Habits

  • 1
    Read AI-generated code line by line before using it
  • 2
    Try writing code yourself first, then compare with AI suggestions
  • 3
    When AI suggests something unfamiliar, look it up independently

Build These Habits

  • 1
    Review AI-generated code as rigorously as you review pull requests
  • 2
    Iterate on prompts to get better results rather than accepting the first output
  • 3
    Keep a mental list of where AI tends to be wrong in your codebase

Seek Feedback

  • "Does this AI-generated code follow our patterns?"
  • "What did I miss in reviewing this output?"
  • "Am I relying on AI too much for things I should be learning?"

Seek Feedback

  • "Did this AI-assisted code introduce any issues I missed?"
  • "Am I reviewing AI output carefully enough?"
  • "Where should I be using AI more—or less—effectively?"

Signals You're Ready to Level Up

  • You catch errors in AI-generated code before they reach review
  • You can explain why AI suggestions work, not just that they do
  • You use AI to learn faster without skipping the learning itself

Signals You're Ready to Level Up

  • Your AI-assisted work is indistinguishable in quality from your manual work
  • You catch AI's subtle mistakes before they reach production
  • You help others understand when and how to use AI effectively

Focus Summary

  • Verify before you trust
  • Understand before you use
  • Stay curious

At this stage, curiosity matters more than command of any specific tool. The engineers who grow fastest are the ones who stay curious about what AI gets right, what it gets wrong, and why.

Focus Summary

  • Review everything
  • Think before you prompt
  • "Works" is not enough

At this stage, speed is the easy part. The real skill is maintaining the discipline to think clearly when the tools make it tempting to skip ahead.