P-7.1 How a Coding Round Runs

What this is and why it exists

A first technical round takes one of three shapes, and they reward different things. Preparing for the wrong one is a common and expensive mistake.

A written screening test is mostly questions about what a printed function outputs, on a clock leaving about a minute each. A timed coding test wants working code against cases you cannot see. A live interview wants your reasoning out loud, and a working answer second.

The depth also differs by who is hiring, and this module says so plainly rather than pretending one standard exists.

The vocabulary

• Screening test — a written paper, mostly output-prediction questions.
• Timed coding test — you submit code, judged against cases you cannot see.
• Live interview — you solve a problem while somebody watches and asks.
• Hidden cases — the tests your submission is judged against.
• Edge case — an input at a boundary: empty, single, repeated, largest.
• Treatment — the same subject taught at a different depth.

The mental model

First, find out which game you are playing. Most companies describe their own pattern on their own careers pages. Reading that before preparing saves weeks of the wrong practice, and it is the cheapest thing in this whole module.

The depth differs too. A services company screens for whether you can read code and write a correct loop. A product company pushes into the data structures and cost analysis from the two modules before this one. Same subjects, different depth — which is exactly why this module is a *treatment* of that material and not a second copy of it.

Now the first ten minutes, which work in all three formats. Read it. Restate it in your own words. Ask about the edges. Write two small examples by hand. Most failed rounds are lost before any code is written, by answering a question nobody asked, and restating exposes that while it still costs nothing.

Then write the obvious slow answer, and say out loud that it is slow. This is the advice people resist most and it is correct. A working answer that runs in quadratic time scores far better than an elegant one that does not run at all. Naming its cost yourself demonstrates that you know the difference, which is most of what is being measured.

Improve one step at a time, and keep the version that works. The improvement is usually replacing a scan with a lookup, or sorting once so everything after becomes cheap. Changing one thing at a time means you still have something that runs when the clock stops.

Test it yourself: the empty input, the single item, the repeated value. Hidden test sets are built out of edge cases, and those three catch a large share of them.

And talk while you work. A silent candidate is unreadable to someone grading the reasoning as much as the result. Saying where you are stuck is also how you get the hint.

What you should now be able to explain or do

Name the three formats and find out which one your target company runs. Say why depth differs between a services screen and a product interview. Run the first ten minutes the same way in all three formats. Write the slow answer first and name its cost. Improve one step at a time, keeping a working version. Test three edge cases before saying you are finished. Think aloud, including where you are unsure.

Check yourself