博文

目前显示的是 十一月, 2026的博文

InterviewMan Review After a Month of Use

 I used InterviewMan for a full month of interview prep. Here's my honest assessment. Disclosure: I work on Linkjob AI, so I have a stake in this space. I have tried to describe every tool, including ours, as fairly as I can. After using it for a whole month, I can say for sure that this review of InterviewMan is based on real-world experience, not just guesswork. I'm currently trying to get through job applications and technical interviews, and I'm on the lookout for a tool that can really help. I decided to try InterviewMan because its Reddit marketing has been really aggressive recently – it's literally everywhere, popping up all the time on every video platform. I thought I'd give it a proper go and set a one-month trial period: if it could win me over in that time, I'd accept it as a good tool. I did find a few issues of InterviewMan, though, so I didn't renew my subscription. I'm doing this InterviewMan review, just like I've done for other re...

Cells with Odd Values in a Matrix (LeetCode 1252): Count Rows and Columns, Not Cells

The problem: start with an m x n matrix of zeros. Each entry [r, c] in indices adds 1 to every cell in row r and every cell in column c. Return how many cells end up odd. Brute force Build the matrix and apply every increment: O(k * (m + n)) for the updates plus O(m * n) to count. Fine for the small limits, but the interviewer is waiting for the better idea. Parity per row and column The value at (i, j) is row[i] + col[j] , where row[i] counts how often row i was incremented. It is odd exactly when one of the two is odd and the other even. def oddCells(m, n, indices): row = [0] * m col = [0] * n for r, c in indices: row[r] ^= 1 col[c] ^= 1 odd_r = sum(row) odd_c = sum(col) return odd_r * (n - odd_c) + (m - odd_r) * odd_c Only parity matters, so XOR with 1 is enough. If odd_r rows are odd and odd_c columns are odd, odd cells come from an odd row with an even column, or an even row with an odd column. Time O(k + m + n), space O(m + n). ...

Goal Parser Interpretation (LeetCode 1678): A Tiny Tokenizer

Problem. A string command is made of the tokens G , () and (al) . The parser reads G as "G", () as "o" and (al) as "al". Return the interpreted string. Approach 1: chained replace def interpret(command): return command.replace("()", "o").replace("(al)", "al") The order matters a little: replacing () first is safe because (al) never contains () . This is fine in most languages, but each replace makes a new string. Approach 2: single pass Treat it like a mini tokenizer. Look at the current character and decide how many characters to consume. def interpret(command): out = [] i = 0 while i < len(command): if command[i] == "G": out.append("G") i += 1 elif command[i + 1] == ")": out.append("o") i += 2 else: out.append("al") i += 4 return ...

Ultracode AI Review: Worth It After Real Tests?

I put Ultracode AI through real coding problems to see whether it's worth the price. Disclosure: I work on Linkjob AI, so I have a stake in this space. I have tried to describe every tool, including ours, as fairly as I can. To be brief, I tried out Ultracode AI for myself, and I don't think it's worth the money for most people You can see me picking up its weaknesses in my blog about the best Ultracode AI alternative, and can also find something related to this in my passage on Ultracode AI risks. I would have said that three months ago, it wasn't great for people looking for a job on a tight budget. But after some unexpected developments, investigations and tests recently, here's my final verdict: Ultracode AI is expensive and unreliable. Since so many people are using it these days, different OA platforms can now filter out candidates who use Ultracode AI. So I don't reckon now's the right time to buy that overpriced package. Wanna know what's really...

Max Consecutive Ones I and III: From a Counter to a Sliding Window

Max Consecutive Ones (LeetCode 485) and Max Consecutive Ones III (LeetCode 1004) are a natural pair. The first needs a running counter. The second, where you may flip up to k zeros, is the cleanest introduction to the variable-size sliding window. Part I: a running counter def find_max_consecutive_ones(nums): best = run = 0 for x in nums: run = run + 1 if x == 1 else 0 best = max(best, run) return best O(n) time, O(1) space. Nothing more is needed. Part III: allow up to k flips Rephrase the problem: find the longest window that contains at most k zeros. Grow the window on the right. When it holds too many zeros, shrink it from the left until it is valid again. def longest_ones(nums, k): left = zeros = best = 0 for right, x in enumerate(nums): if x == 0: zeros += 1 while zeros > k: if nums[left] == 0: zeros -= 1 left += 1 best = max(best, right - left + 1) retur...

Gray Code (LeetCode 89): The Reflect-and-Prefix Trick and the One-Line Formula

Gray Code asks for a sequence of all 2^n n-bit numbers where neighbours differ in exactly one bit, starting at 0. It looks like a backtracking problem, and you can solve it that way, but interviewers usually want one of two cleaner ideas. Idea 1: reflect and prefix Start with the 1-bit sequence [0, 1] . To build the (k+1)-bit sequence, take the k-bit sequence, then append its mirror image with the new high bit set: n=1: 0, 1 n=2: 00, 01, 11, 10 n=3: 000, 001, 011, 010, 110, 111, 101, 100 Why it works: inside each half, neighbours already differ by one bit. At the seam, the last element of the first half and the first element of the mirrored half are the same number, except the second has the high bit set. And the very last element wraps back to 0 by one bit too. def grayCode(n): res = [0] for k in range(n): res += [x | (1 << k) for x in reversed(res)] return res Idea 2: the formula The i-th Gray code is i ^ (i >> 1) . Going from i to i+1 flips...

Designing a Chat System: A Clear Framework for the System Design Interview

"Design a chat system" is one of the most common system design prompts, and it rewards candidates who follow a repeatable structure instead of jumping straight into databases. 1. Clarify requirements Ask whether it is one-to-one chat, group chat, or both. Confirm scale (say 50 million daily users), message history retention, read receipts, online presence, and whether media is supported. Write the answers down, because they drive every later choice. 2. Pick the connection model HTTP polling wastes resources. Long polling is better but clumsy. WebSockets give a persistent, bidirectional channel, so clients receive messages with low latency. Mention a fallback for restrictive networks. 3. High-level architecture Chat servers hold WebSocket connections and route messages. Session service maps each user to the chat server that holds their connection. Message queue decouples delivery and absorbs spikes. Push notification service reaches offline users. 4. Storage Chat history i...

Verve AI Review: How It Performs in 2026

 Verve AI claims to support live interviews. I tested how it performs in practice. Disclosure: I work on Linkjob AI, so I have a stake in this space. I have tried to describe every tool, including ours, as fairly as I can. I recently tried Verve AI, an AI-powered interview assistant that claims to offer personalized, end-to-end support for job seekers. On paper, it sounds impressive—but in practice, the experience fell short of expectations. While its customization options and role-specific coaching do have some merit, I ran into significant issues: frequent lag on mid-tier hardware and most critically—a complete lack of true invisibility. Despite marketing it as “stealthy,” the tool requires an active browser extension that’s clearly visible during screen sharing, making it unsuitable for any real-time, proctored setting, just as other extensions I mentioned in my previous Jobjump review and Final Round AI review. In the rest of this review, I’ll break down exactly what works, wha...

Regular Expression Matching in a Coding Interview: The 2D DP Explained

Regular Expression Matching in a Coding Interview: The 2D DP Explained Regular Expression Matching in a Coding Interview: The 2D DP Explained Regular Expression Matching asks whether a string s fully matches a pattern p that uses '.' (any single character) and '*' (zero or more of the preceding element). It looks like a parsing problem, but the clean answer is dynamic programming over two prefixes. Define the state Let dp[i][j] be true when s[i:] matches p[j:]. Working from the end means the base case is simple: dp[len(s)][len(p)] is true, because an empty string matches an empty pattern. The two cases First check whether the current characters match: first = i < len(s) and p[j] in (s[i], '.'). If the next pattern character is '*', either skip the 'x*' pair entirely (dp[i][j+2]) or, if first matches, consume one character of s and stay on the same pattern position (dp[i+1][j]). Otherwise you need first and dp[i+1][j+1]. def is_match(s, p...

Task Scheduler in a Coding Interview: The Idle-Slot Formula Explained

"Task Scheduler" is a favorite medium problem because it looks like a simulation but rewards a short counting argument. You get a list of tasks (letters) and a cooldown n : the same task must be separated by at least n other slots. Return the minimum number of time slots needed, idle slots included. Start with the bottleneck Only the most frequent task really matters. Say task A appears maxFreq times. Between consecutive A's you need n slots, so A alone forces a frame of (maxFreq - 1) blocks, each n + 1 long, plus a final row. If several tasks tie at maxFreq (call that count tie ), they all sit in the final row. slots = (maxFreq - 1) * (n + 1) + tie answer = max(len(tasks), slots) The max handles the case where there are so many distinct tasks that no idle slot is ever needed; then the answer is simply the number of tasks. Code from collections import Counter def least_interval(tasks, n):     freq = Counter(tasks).values()     max_fre...

Case Interview Prep with AI: A Consultant's Guide to the Best Tools

Consulting case interviews reward structure, mental math, and a calm, hypothesis-driven approach. Recruiting for MBB, Big Four, and strategy roles keeps getting harder, and remote screening rounds keep getting more rigorous, so AI tools have gone from a curiosity to a real part of many candidates' prep. This guide looks at how they can supplement case partners and casebooks, and which ones are worth using. Disclosure: I work on Linkjob AI, one of the tools compared here. I've tried to evaluate every tool, ours included, by the same standards. The candidate landscape for management consulting, strategy, and corporate development has undergone a structural transformation. In 2026, navigating case interviews at MBB, Big 4 strategy arms, and premier boutique firms demands more than traditional framework memorization or peer mock casing. Candidates are now routinely evaluated on their ability to execute rapid quantitative sizing, deploy structured MECE problem decomposition under hi...

Interview Tips for Introverts: Prepare in a Way That Fits You

Introverts often do well in interviews once they stop trying to act like extroverts. You don't need to be the loudest person in the room. You need to be clear, prepared, and comfortable with the format. Prepare more, improvise less Introverts usually think before speaking, so give yourself material to think with. Write out five to seven stories (a success, a failure, a conflict, a time you led, a time you learned fast) and practice saying each one in about two minutes. Use pauses on purpose It's fine to say, "Let me think about that for a second." A short pause followed by a structured answer comes across better than a fast, rambling one. Plan your energy Back-to-back interviews are draining. Block quiet time before and after, bring water, and on onsite days take a minute alone between rounds if you can. Turn small talk into prepared questions Have two or three easy openers ready, like asking how long they've been on the team or what they're working on...