When you do not have an interview scheduled, the right structure is two modes, not one. Run a low-intensity base layer you can sustain for months without burning out, then switch to a high-intensity spike plan the moment a real loop lands on the calendar. The mistake we see most often in mock interviews is people trying to sprint indefinitely: they cram 60 problems in two weeks with no interview in sight, forget most of it, and are exhausted when an actual recruiter emails. Prep without a deadline is an endurance problem first and a sprint problem second.
This is a real constraint, not a hypothetical. Job searches stretch out. As of mid-2026, the median unemployment spell in the US sat around 11 to 12 weeks, while the average was closer to 23 to 24 weeks because a long tail of searches drags the mean up. Long-term unemployment (27 weeks or more) accounted for about 27 percent of all unemployed people. Translation: plan for months of uncertainty, then be ready to compress hard on short notice.
Why "no deadline" is its own problem
An open-ended timeline creates three specific failure modes.
You forget faster than you learn. Massed cramming produces short-term recall that decays quickly. Spaced practice, revisiting material across days and weeks, reliably outperforms cramming for long-term retention. A meta-analytic review of the spacing effect across hundreds of experiments found that distributed practice produces substantially better retention than the same study time packed together. If you do not know when the test is, spaced practice is not optional, it is the only approach that survives an unknown gap.
You over-index on hard, rare problems. With unlimited runway and no pressure, people gravitate to obscure dynamic programming puzzles because they feel productive. Meanwhile they never practice explaining a solution out loud or handling a follow-up. The base layer should protect against this.
You burn out before the interview arrives. A 4-hour-a-day pace is impossible to hold for three months. When the loop finally lands, you have nothing left in the tank.
Build a base layer you can sustain
The base layer is the pace you could keep up indefinitely while working or job hunting. The goal is not speed. It is keeping fundamentals warm and slowly widening coverage so a future spike has a strong foundation.
A realistic base looks like this:
- 3 to 5 problems per week, not per day. Quality over volume.
- Rotate patterns, do not marathon one. One week lean into two pointers, the next binary search, the next hash table work. Breadth keeps everything from decaying at once.
- Re-solve, do not just read. Revisit a problem you did two weeks ago from a blank editor. This is the spacing effect in action.
- One mock every two weeks. Enough to keep verbal muscles alive without heavy prep load.
Cover the patterns that appear across almost every company: arrays and hashing, dynamic programming, trees, graphs, and heaps. Our concepts index shows question counts per topic, which is a decent proxy for where to spend base-layer time. Do not chase a target problem count. When people say they did "200 problems," that number is anecdotal and does not track how well you explain code under pressure, which is what actually gets scored.
Keep a one-line log per problem: the pattern, the trick you missed, and the date. That log is what makes your future spike fast.
The spike plan: trigger it when a date lands
The instant a recruiter confirms an onsite or a phone screen, switch modes. Now you have a deadline and a target company, so prep becomes specific and intense. Two templates depending on lead time.
The 2-week spike (short notice)
| Days | Focus | Volume | Mocks |
|---|---|---|---|
| 1 to 3 | Company-tag the loop; drill top 2 patterns for that company | 3 problems/day | 0 |
| 4 to 7 | Medium and hard problems in tagged patterns; time every solve | 3 to 4/day | 2 this week |
| 8 to 11 | Mixed sets, no pattern labels, simulate cold problems | 2 to 3/day | 2 |
| 12 to 13 | Behavioral stories, resume walkthrough, re-solve your misses | 2/day | 1 |
| 14 | Light warm-up, one easy problem, sleep | 1 | 0 |
The 3-week spike (comfortable lead)
| Week | Focus | Volume | Mocks |
|---|---|---|---|
| Week 1 | Rebuild breadth across the company's top 4 patterns | 3/day | 2 |
| Week 2 | Depth: mediums and hards, timed, explain out loud every time | 4/day | 3 |
| Week 3 | Full-loop simulation, behavioral prep, taper the last two days | 2 to 3/day | 3 early, then taper |
Mock cadence matters more than raw problem count in the spike. We suggest 2 mocks the first week and ramping to 3 the week of the interview. If most of your prep is silent solving, the first time you talk through code should not be with a real interviewer. For getting mock reps right, see our guide on practicing mock interviews with AI.
Company-tagging workflow
The spike is where a company breakdown pays off. Pull the company page (for example, our company breakdowns) and build a tag list of the patterns and problems that recur there, then drill in frequency order.
Take Line as a worked example. Its question bank leans on problems like these, in rough order of how often they show up:
- LRU Cache (mid-senior): design a fixed-capacity cache with O(1) get and put. This tests hash map plus doubly linked list fluency and appears frequently across many companies, so it is high value regardless of your target.
- Sliding Window Maximum (staff): return the max of each size-k window as it slides. A monotonic deque problem that also sharpens your array and deque instincts.
- Minimum Cost to Cut a Stick (staff): interval dynamic programming where cut order changes total cost. A good stress test of whether your DP framework holds up on a non-obvious state definition.
Notice these three span three different patterns: design, sliding window, and interval DP. That is deliberate. Tag by pattern, drill the highest-frequency pattern first (design and LRU Cache here), then work down. You can drill these directly on our practice page filtered by company or concept.
Track progress and taper
Two numbers tell you if the plan is working. First, cold-solve rate: on unlabeled problems, can you name the pattern and reach a working solution in the time limit? Second, explain quality: in mocks, are you narrating a plan before coding or coding in silence? A rising cold-solve rate with clean narration is the signal you are ready.
Taper the last 48 hours before any loop. Re-solve two or three problems you have already nailed, review your behavioral stories, and stop learning new material. Cramming a novel hard problem the night before an interview raises anxiety and buys almost nothing. For the final morning, our piece on warming up on interview day covers the routine.
When the loop ends, do not drop to zero. Return to the base layer. Most searches involve multiple loops, and the person who stayed warm between them starts the next spike halfway up the hill.
FAQ
How many problems should I do per week with no interview scheduled?
Three to five, done well, with periodic re-solves of older problems. The base layer is about retention and breadth, not volume. Save high daily counts for the spike phase when a date is confirmed.
Should I still do mock interviews if I have no interview lined up?
Yes, but sparingly, roughly one every two weeks. Explaining your reasoning out loud is a separate skill from solving, and it decays if you never practice it. In a spike you ramp to two or three mocks per week.
How far ahead should I start preparing if I do not know my timeline?
Start the base layer now and treat it as ongoing. Because searches commonly run several months, and long-term unemployment made up about a quarter of the jobless in 2026, a sustainable low-intensity habit beats waiting for a trigger and scrambling.
What if I get an interview with only a few days of notice?
Compress the 2-week spike: company-tag immediately, drill the top two patterns for that company, do one mock, and taper the last day. Your base-layer log of past mistakes is what makes this fast.
Is it better to go deep on one pattern or broad across many?
Broad in the base layer, deep in the spike. With no deadline, rotating patterns keeps everything from decaying at once. Once you know the target company, go deep on the two or three patterns their breakdown shows most often.
Sources
- Employment Situation Summary (August 2026), U.S. Bureau of Labor Statistics
- Median Weeks Unemployed (LNU03008276), FRED, Federal Reserve Bank of St. Louis
- Average Weeks Unemployed (UEMPMEAN), FRED, Federal Reserve Bank of St. Louis
- Spacing Effects in Learning: A Temporal Ridgeline of Optimal Retention (meta-analytic review), Cepeda et al.