The interview is in a week and you cannot learn everything
With seven days left, most preparation advice is useless because it assumes you have a month. Here is what actually moves the outcome in a week, in the order that matters, and what to deliberately skip.
You have seven days. Somewhere there is a list of two hundred questions you are supposed to know, and reading it makes things worse rather than better.
The instinct is to start at the top and grind. It is the wrong move, and not for motivational reasons: in a week you cannot close a knowledge gap that took two years to open, so any plan built on closing it fails. What you can do is change how much of what you already know actually comes out under pressure — and for most candidates that gap is larger than the knowledge one.
I ran final-round interviews for ten years. The people who did badly were rarely the people who knew least.
Day 1 — find out what will actually be asked
Not "revise everything". Read the job posting like evidence.
The technologies listed first, and the ones repeated in different words, are what the team is doing right now. Something mentioned once at the bottom is aspirational. If the posting says "event-driven architecture" twice and "Kubernetes" once, the interview will spend twenty minutes on messaging and two on the cluster.
Look at the company: product or agency, size, what they sell. A product company asks how you would maintain something for three years. An agency asks how quickly you get to a working result. The same question — "how do you approach technical debt" — has different right answers in the two.
Write down the ten topics you now expect. Ten, not thirty. That list is your week.
Time: one hour. Skipping it makes the other six days generic.
Days 2–3 — sort those ten into three piles
Honestly, out loud, and this is where people flinch.
Know it and can explain it. Say it aloud once to confirm — some things live comfortably in your head and fall apart in your mouth. Then leave them alone.
Know it but explain it badly. This is your entire week's work. These are questions where you have the knowledge and lose the point in the delivery: you start in the middle, you cannot pick an example, you say "basically" nine times. Fixing one of these takes twenty minutes and gains more than a day of new material.
Do not know it. Now decide, per item, whether it is central to the role. Central and unknown: learn the shape, not the depth — what it is for, what it replaces, what its main trade-off is. That is enough to have a conversation and admit the edge honestly. Not central: skip it, on purpose, and stop paying attention to it.
The skipping is the part people cannot do, and it is where the week is usually lost.
Days 4–5 — say it out loud, under something like pressure
This is the highest-return activity available to you and almost nobody does it.
In your head your answers are excellent. Your head fills gaps, skips transitions and forgives false starts. The mouth does none of that. The first time you say an answer aloud you discover it takes ninety seconds and does not have an ending.
Three ways, in descending order of usefulness:
A person who will interrupt you. A friend in the industry who asks follow-ups is the best simulation. Rare, so use it on your weakest pile.
A tool that asks and reacts. Practising against generated questions for your role and level, out loud, is the closest thing to the real rhythm you can get alone. That is what we built — you can run a session from the questions for your specialisation, and the first three are free.
Recording yourself. Free, unpleasant, effective. Ask yourself a question, answer it into your phone, listen back. You will hear the filler words and the missing endings immediately and want to stop after two. Do five.
Two hours across two days is enough to change how you sound.
Day 6 — prepare your four stories and your questions
Four stories. Most behavioural questions are four or five questions in different clothes: something hard you solved, something that went wrong, a disagreement, something you are proud of. Four pieces of real work, ninety seconds each — problem, what you did, what changed — cover almost all of it. Do not write scripts; scripts sound like scripts.
Your questions for them. Not "what is the team culture like" — that gets a rehearsed answer. Ask what the last three months looked like. Ask what breaks most often. Ask why the position is open. Those answers tell you whether you want the job, which is the half of the interview candidates forget they are conducting.
Day 7 — do almost nothing
Reread your ten topics and your four stories. One hour, no more.
Cramming on the last day trades sleep for material you will not recall, and tiredness costs more in an interview than any single gap. Sleeping badly makes you slow at exactly the thing being measured: thinking out loud while somebody watches.
What to skip on purpose
Two hundred questions. You will remember the first thirty badly and none of the rest.
Things that are not in the posting. Hard algorithm puzzles for a role that will never ask them; Kubernetes internals for a job whose posting mentions Kubernetes once.
Rewriting your CV. They already read it. That week is gone.
The last-day cram. See above.
What actually changes the outcome
Interviews are not lost on missing knowledge nearly as often as candidates believe. They are lost on structure — starting in the middle, no example, running out of road halfway through an explanation — and on the moment somebody guesses instead of saying "I do not know".
Both are fixable in a week. Neither requires new material.
Say your answers out loud. That is the whole plan, and everything above is just deciding which answers.