Lately, AI has been all the rage. With ChatGPT and DeepSeek assisting in writing code, many are worried: "Are programmers going to become obsolete?"

I think it's not that simple.

Because what programmers do goes way beyond just writing code - coding might actually be the easiest part. The real challenge lies in dealing with people.

1. Client requirements are like your girlfriend saying "whatever" 

When the client says: "I want an e-commerce website, keep it simple."

A rookie programmer would respond: "Got it! Let's just use a Shopify template!"

A seasoned programmer would ask: "How simple is simple? Do you need flash sales? How should membership points work? What's the return process? Which payment gateway to integrate? Do you want a distribution system? How detailed should the data analytics be?..."

Client (suddenly silent): "Uh... I haven't thought that far yet."

You see, the real requirements are often hidden in what's left unsaid. AI can generate code based on existing data, but it can't slowly uncover what the client truly wants through conversation like humans can.

2. "This feature should be simple, I don't care how you implement it"

The most terrifying phrase in this world isn't "your code has bugs," but "this feature should be simple."

Clients think "adding a login feature" is just a few lines of code, but in reality:

- Phone number login or email login?

- SMS verification? Which provider to use?

- How to handle forgotten passwords?

- Do we need fraud prevention?

- What about user agreements and privacy policies?

AI can write the code for a login feature, but it can't determine which details the client actually cares about and where compromises can be made. Programmers have to constantly balance technology, cost, and user experience - something AI truly can't do.

3. Programmers are essentially "emotionally stable customer support"

When the system crashes and the client is panicking: "All our orders are stuck! Fix it now!"

AI can spit out logs and solutions, but what the client needs at this moment isn't just technical fixes, but emotional reassurance:

- "Don't worry, we're on it. No data will be lost."

- "We've seen this issue before. It'll be fixed within 30 minutes."

- "We can add monitoring later to prevent this from happening again."

Technical problems can be solved, but human anxiety and trust require human maintenance.

4. Operations are like being a doctor - experience beats textbooks every time

AI can handle known errors following standard procedures, but real-world problems never follow the script:

- "Server CPU spikes, but logs show nothing?" (Maybe some outsourced team secretly ran a script)

- "Users complain the page loads slowly, but monitoring shows everything's normal?" (Could be a CDN issue in some region)

- "Database suddenly crashes, but no one touched it?" (Maybe the disk is full, or the team next door messed up)

True ops veterans rely on experience and intuition, like an old-school doctor diagnosing by pulse. No matter how advanced AI gets, it can't replace this almost "mystical" troubleshooting ability.

5. Programmers also have to be "translators" between tech and business

When the product manager says: "We need a smart recommendation system to boost user engagement!"

The boss understands: "So like Douyin, where you keep scrolling addictively, right?"

The tech team understands: "We'll need machine learning, user profiling, A/B testing..."

Then finance asks: "How much will this cost?"

Programmers have to find the balance between the boss's expectations, product feasibility, technical costs, and business value. AI can't handle this because it doesn't understand concepts like "the budget's tight, let's make do for now."

AI is a tool, programmers are the ones wielding it

So don't worry about AI taking your job. Truly excellent programmers will always be dealing with people, not just code.