You forget LeetCode problems because you solve each one once and move on. What keeps a problem is spaced repetition: re-solving it from a blank editor several times, each time just as it starts to slip, with the gaps growing as it sticks. Five or six spaced re-solves over two months do more for your recall in a live round than the same hours spent grinding new problems. Firecode's learning engine, built on decades of memory science, runs that schedule for you, so every problem you solve comes back at the right moment.
Trying to run that schedule yourself with a spreadsheet or a browser extension turns you into a full-time bookkeeper of your own memory. Every missed day piles up, and by week four the tracking breaks down right when your early problems start fading. On Firecode, you never log a date or pick what is due: you open the app, solve today's session, and walk into the onsite with every pattern ready.
Why solved problems disappear
You see LRU Cache in a live interview, know you solved it a month ago, and stare at the blank editor unable to remember whether the hash map stores nodes or values. The interviewer is waiting, the clock is running, and a problem you already paid the time to learn is gone.
In 1885 Hermann Ebbinghaus measured how quickly memory fades and found that most of the loss happens in the first hours and days. A coding problem means more than a nonsense syllable, so it holds on a little longer, but the curve is the same. A medium you solved in week one and never touched again is mostly gone by week six.
The way most engineers practice makes the drop even steeper.
First, solving a problem with the pattern already in your head feels like knowing it. When you finish three sliding window problems in a row, the fourth goes fast because you already know what tool to reach for. A real interview gives you no category header. You have to read a prompt cold and recognize the underlying structure on your own.
Second, reading your old code tricks you into thinking you still have it. Roediger and Karpicke showed in 2006 that students who tested themselves remembered more a week later than students who reread their notes, even though the rereaders walked away feeling more confident. Skimming your accepted submission the night before an onsite is rereading. Rebuilding the solution from an empty editor is the test that makes it stick.
Timing matters just as much as testing. A review does the most work when it arrives right as the memory starts to fray, with the gap growing wider each time you succeed. Pull the solution out of your own head from a blank editor, right before it slips, and you stop blanking when the clock is running. Firecode's learning engine times that whole cycle for you.
What a real review looks like (and why DIY fails)
Most arguments online about whether spaced repetition works for coding interviews come from engineers reviewing the wrong way. Rereading old code, or memorizing a solution line by line, only trains you to recite. The moment an interviewer changes a constraint or asks for an in-place pass, recited code falls apart.
A review that holds up under pressure means opening the problem statement with a blank editor, stating the approach in a sentence or two, and writing code that passes the test suite. When you try to run this on your own, four traps usually kill it:
- Reviewing by reading instead of typing. If your fingers did not write the code and pass the tests from scratch, you did not review the problem.
- Generous self-grading. Telling yourself "I basically had the idea" after peeking at a solution hides the exact edge case that will trip you up in a live round.
- Reviewing in topic order. When all your sliding window problems come back on the same Tuesday, the list itself tells you the pattern before you read the prompt.
- Skipping due reviews to chase a higher solved count. In an interview, 300 problems solved once lose every time to 120 you can write cold.
| Route | Who decides what is due | What a review is | What you keep up every day |
|---|---|---|---|
| Spreadsheet | You, sorting date columns by hand | Whatever you force yourself to do | Logging, grading, sorting, rescheduling |
| Browser extensions for LeetCode | A third-party extension | A manual re-solve on the same page | Rating every solve and picking all new problems yourself |
| Firecode | The learning engine | The real problem in a blank editor, tests run | Nothing. You open it and solve |
The spreadsheet dies first: the first busy week at work leaves a backlog of overdue rows so long that most people abandon the sheet, right when their early problems need it. Browser extensions bolt a reminder onto a problem archive, but you still pick every new problem yourself and hope your mix matches a real loop. If you are fitting prep around a full-time job, see how to run interview prep on a busy schedule in one daily block.
One problem, followed for eight weeks
Take Longest Substring Without Repeating Characters, a classic sliding window medium.
The first time you see it, you need a nudge to realize the left pointer can jump directly past the duplicate index instead of stepping one by one, and the whole solve takes 35 minutes. A little while later, the problem shows up in your session again. You remember the hash map of last-seen indices, fumble the off-by-one check when shrinking the window, and finish in 14 minutes. The next time it comes back, you write it cleanly in eight.
Weeks later, it appears mixed in between a graph traversal and a heap question. You blank on whether to store counts or indices, fail a test case, and have to work through the pointer logic again. It is back in front of you soon after, while the mistake is still fresh enough to fix. The next two times you see it, the code flows out in five minutes.
By the second month, you are not remembering code anymore. When you hit Longest Substring with At Most K Distinct Characters for the first time, you spot it in thirty seconds as the exact same window with a different shrink condition. Across eight weeks, keeping that pattern razor-sharp took about an hour in total, and you never once had to look at a calendar. Firecode's learning engine decided every one of those returns.
The engine that does it for you
LeetCode is a massive archive of problems, and NeetCode and AlgoExpert give you curated lists and video explanations, which makes them useful for browsing a topic for the first time. None of them brings a solved problem back when you are about to forget it, or adapts your daily session to how well you actually performed.
Firecode was built around that missing piece. Its learning engine runs on Firecode's own algorithm and machine learning, grounded in decades of memory science, and it learns from every solve you make. Solve Merge Intervals in September and it is still in your fingers at the November onsite, without you ever opening a calendar. That is the part members talk about most: the work they put in stays put.
When the engine brings a problem back, you face the real problem in a blank editor, in any of 14 languages, with the tests waiting and no topic header to tip you off, just like a live round. Reviews arrive mixed in with new problems, so you practice spotting the pattern cold every day. If you stall on a recursive base case or a pointer update, Firebot gives you the smallest nudge that gets you moving, the way a good interviewer would, so the solve is still yours. On many problems, a drawn step-by-step walkthrough of the pointers or the call stack makes the idea click for good. And the engine keeps the balance for you, so older patterns stay sharp without reviews burying your new work. If you are mapping out a longer runway, see how this fits into a 12-week FAANG plan.
Firecode was built by former Meta, X and LinkedIn engineers who designed interview loops, ran thousands of interviews, and watched strong candidates blank on problems they had solved a month earlier. Since then, 50,000+ engineers have used its 1,500+ real interview questions to get ready for top tech loops. The median offer comes after 173 problems, at about 22 minutes a day.
Start your 14-day free trial on Firecode today. Within the first couple of weeks, a problem that stumped you on day one will come back and you will write it cold. That is how you walk into the onsite ready, and out with the offer.
Frequently asked questions
Does spaced repetition work for coding interviews?
Yes, provided each review means writing a working solution from a blank editor and passing the tests. Re-solving at widening gaps trains you to spot the underlying pattern cold and handle variations under time pressure, whereas rereading old code only builds false familiarity.
How often should I review LeetCode problems?
A problem needs to come back soon after you first solve it, then at progressively wider gaps each time you pass it cleanly, and right away whenever you stumble. Because the right gap depends on how hard you found the problem and how close your interview is, Firecode's learning engine times every review for you automatically so you never have to calculate dates.
Why do I keep forgetting LeetCode solutions I've already solved?
Solving a problem once only proves you could work it out in the moment, often with the topic name already giving away the approach. Without spaced retrieval practice over the following weeks, the forgetting curve wipes out the details and leaves you blanking when the same question appears in a live round.
Should I memorize LeetCode solutions?
No. Memorized lines of code fall apart the moment an interviewer changes a constraint or asks a follow-up. Rehearse the problem-solving steps by rebuilding the code from scratch across several spaced sessions until the pattern and its edge cases feel automatic.
How many problems should I re-solve each day?
You want enough reviews to keep your past problems sharp while still leaving room for new patterns, shifting toward more mixed reviews as your interview date gets close. Managing that balance by hand quickly turns into a backlog, which is why Firecode's learning engine builds your daily session for you and keeps practice to about 22 minutes a day.
Does LeetCode have built-in spaced repetition?
No, LeetCode is a problem library and does not schedule reviews or resurface questions you solved weeks ago. Firecode has spaced repetition built into its core: its learning engine tracks every solve and brings each problem back just before you would forget it.
Is there an app for spaced repetition on LeetCode problems?
Firecode is built specifically for this. Its learning engine brings back every problem you solve at the right moment in a blank editor, and pairs it with new questions from a catalogue of 1,500+ real interview questions. The median offer comes after 173 problems, at about 22 minutes a day.