Why Learning to Code Matters More Now That AI Can Write It

Why I put the alert in the checkout code

The test would tell me whether a browser could reach checkout. If analytics counted those test visits, cart additions, and checkout starts alongside activity from real shoppers, the reports could make it look like more people left before buying.

I knew where the checkout code already handled errors, so I put the alert there. A failure that reaches that part of the code can trigger an email, and a cooldown limits repeats. A successful checkout stays quiet. If the email service goes down, the checkout can still handle the original error.

As AI gets faster at giving me a first draft, I get more value from knowing how the system works. I can tell whether a suggestion fits the job before I spend time building it.

What I count as time saved

I’m happy to use AI for a first pass once I know what the change needs to do. For the checkout alert, I’d still test a successful checkout, a failure, and a repeat failure during the cooldown. I’d also check what the shopper sees if email delivery goes down. I count that work when I decide whether AI saved me time.

Researchers studying GitHub activity estimated a bigger increase in code changes than in versions released to users after developers started using AI tools.

Follow one action through the code

When I’m learning a codebase, I start with an action I can repeat. In a test cart, I’d change an item’s quantity from one to two and follow what happens. Where does the quantity change? Which part calculates the subtotal? What updates the screen?

I’d ask AI to point me to the files involved, then open them and check its explanation against what happens in the browser. If the subtotal is wrong, I can trace where the value first differs from what I expected.

Give AI a rule you can test

Say a store offers free shipping when the cart subtotal after discounts is $50 or more. Before I ask AI to change the banner, I’d write down what it should show. At $49.99, the shopper has one cent left to qualify. At $50, the banner should say free shipping is available.

After AI suggests a change, I’d try both amounts, then apply a discount that drops the subtotal below $50. I’d compare the banner with the shipping options at checkout. If they disagree, I’d check which subtotal each one uses.

Conclusion

I want to be able to come back to an AI-assisted feature a week later and still explain how it works. After an AI-assisted change, I’d leave a short note about where the rule lives and what I tested. If the shipping threshold changes next week, I’ll know where to make the update and which checks to run again.