In the AI Era, "People Who Can Explain" Beat "People Who Can Write"

I’ve been reading that oral exams are quietly making a comeback at US universities lately.

The reason is simple.

AI can just do the homework for you.

Take-home reports don’t really tell you much anymore

From what I’m reading in the news, it’s mostly essay-style assignments that are causing the problem.

Throw it into ChatGPT and you get pretty high-quality writing in a few minutes,

so professors are ending up in a situation where

“looking at the submission alone doesn’t tell you anything about the student’s actual ability.”

And from here it’s just my guess, but

programming assignments are probably going through the same thing, right?

• Sample code
• Algorithm explanations
• Test code

These are all stuff AI cranks out no problem.

So they make students explain it on the spot

What’s apparently increasing on the university side is in-person oral exams.

If it were a programming assignment, it would probably look something like throwing questions like:

• Why did you go with this design?
• What other options were there?
• What’s the weak point of this code?
• Which part of the AI’s answer did you doubt?

at the student in real time.

If you just copy-pasted code, you’d probably stall after a few follow-up questions.

On the flip side, if you actually understand it, that comes through in conversation pretty fast.

This is probably coming to engineering interviews too

OK, from here it’s pure speculation on my part.

But I have a feeling engineering interviews will drift in this direction too.

Even if someone shows you code on GitHub, we’re already in an era where you can’t really tell

“did they write this, or was it AI?”

So instead, questions like:

• What’s the worst incident you ran into?
• What was the root cause?
• Why did you pick that tech stack?
• What did you explicitly decide NOT to use?

feel like they’d reveal way more about whether someone actually did the work.

(Again, just a “isn’t this where things are heading?” kind of thought.)

A random thought I had

There used to be a vibe of “if you can write code, you’re good.”

But these days, AI will spit out something decent for pretty much anyone.

So lately, what feels more interesting on the ground are conversations about:

• Why did you think that way
• Where did you get suspicious
• What part gave you a weird gut feeling

I think it’s fine to just use AI

By the way, I’m not saying “don’t use AI.”

I use it. It’s useful.

It’s just that there’s a real gap between

nodding along at whatever it spits out,

and being able to pause and go “wait, is this actually right?”

Lazy conclusion

Writing skill + speaking skill.

These two are starting to feel like a set.

Maybe the AI era is, weirdly, one where “conversation” becomes the clearest window into someone’s real skill.

Just something I’ve been thinking about lately.