How I use AI as my "Senior Developer"

AI didn’t replace my coding - it upgraded how I work.

These days, I write less code by hand, but I think more about the code that ships.

This is how I use AI in my daily workflow - almost like a senior developer sitting next to me.

---

From "Write Code" To "Plan With AI"

I rarely open my editor and just start typing anymore.

Most tasks start in what I call plan mode.

Here’s what that looks like:

• I describe the feature in plain language
• I explain the context - tech stack, patterns, constraints
• I list edge cases, data shapes, and system boundaries
• I ask AI to propose an approach, not code yet

It’s basically how a good junior works with a senior:

“Here’s the problem, here’s what I’m thinking - what would you do?”

The goal isn’t perfect code.
The goal is a plan I understand and agree with.

Only then do I move to implementation.

---

Letting AI Write The First Draft

Once the plan is clear, I treat AI like a fast, reasonably experienced developer.

Typical requests:

• “Implement this Node service based on the plan above”
• “Generate an Angular service and component for this feature”
• “Create an API endpoint that follows this pattern”
• “Refactor this function without changing behavior”

AI handles:

• Wiring functions, services, components
• Applying patterns I ask for
• Getting 70–90% of the solution done

I don’t trust it blindly.
But it’s an excellent starting point that removes a lot of boring work.

---

My Review Process: Treat AI Like a Mid-Level Dev

I never treat AI output as “copy-paste and move on”.

I review it like a pull request:

Does it match the intent?
Are all requirements covered?

Does it follow conventions?
Naming, structure, patterns, error handling

Is it readable?
Clear names, small functions, no hidden complexity

Is it safe?
Validation, no risky assumptions, no leaks

Sometimes I keep most of it.
Sometimes I throw almost everything away.

The responsibility is always mine.

---

Using AI To Review My Git Diff

One of my favorite habits is using AI as a second reviewer.

Flow:

• Work normally
• Commit changes
• Send the diff to AI

I ask it to:

• Explain what changed
• Spot bugs or code smells
• Suggest improvements

It’s surprisingly good at catching:

• Forgotten logs
• Inconsistent error handling
• Risky patterns

I still review everything myself - but it’s a great extra layer.

---

How I Use AI For Debugging

When something breaks, AI becomes my debugging partner.

I provide:

• Relevant code
• Error message or behavior
• What I expected

Then I ask:

• “What’s likely wrong here?”
• “Walk me through this step by step”
• “Why this specific error?”

AI is great at:

• Spotting small mistakes
• Explaining confusing errors
• Suggesting minimal fixes

But I always verify and understand the fix before applying it.

---

How I Use AI For Refactoring

This is where AI shines.

Typical prompts:

• “Make this function smaller and clearer”
• “Split this module logically”
• “Suggest better naming”
• “Convert this to async/await”

I usually:

• Explain what feels wrong
• Set constraints (don’t break API, don’t touch tests)
• Let AI propose a refactor
• Review and refine

AI gives momentum.
I give direction.

---

Where I Don’t Trust AI

AI is fast, but not always wise.

I’m careful with:

• Security-sensitive code (auth, tokens, payments)
• Complex business logic (pricing, domain rules)
• Architecture decisions (boundaries, scaling, structure)

I might ask for options, but I make the final call.

---

How AI Changed The Skills I Care About

Less focus on:

• Typing speed
• Memorizing APIs

More focus on:

• Clear thinking and communication
• Evaluating solutions
• Designing structure
• Explaining code

AI made judgment more important, not less.

---

Simple Rules I Follow

• If I can’t explain the code, I don’t ship it
• If something feels “clever”, I slow down
• I read every line before using it
• I use AI more for:
  • Refactoring
  • Explaining
  • Reviewing
  • Testing
    than for blindly generating complex features

---

Used like this, AI really does feel like a senior developer:

• Fast
• Experienced
• Full of suggestions

But just like with a real senior,
I decide what to keep, what to change, and what to reject.