Evolution of Coding Interview Practices

 ► Coding Interviews in 2008:   - No LeetCode, no online judges.   - Prep = Cracking the Coding Interview, practicing problems from word of mouth.   - Interviews focused on core DSA but weren’t as structured as today.   ► Coding Interviews in 2014:   - LeetCode was new (launched in 2015).   - Companies still relied on books, mock interviews, and past questions.   - Google had already standardized DSA-based interviews, but most companies were still catching up.   ► Coding Interviews in 2020:   - LeetCode became the gold standard for interview prep.   - FAANG interviews shifted towards LeetCode Mediums as the baseline.   - Companies started raising the bar as more candidates got better at structured prep.   ► Coding Interviews in 2024-2025:   - The job market is more competitive than ever, companies have more choices.   - Candidates are grinding LeetCode, so interviewers increase the difficulty to filter people out.   - LeetCode Mediums are still the most frequent, but Hards show up more often.   - System design expectations are creeping into junior roles.  So, are coding interviews harder now?   ► Yes, because companies are optimizing for stronger candidates.   ► Yes, because structured prep is more accessible than ever.   ► Yes, because the job market allows companies to be pickier.  But here’s the good news: If you prep right, you’re already ahead.  Interviews evolve, but the fundamentals don’t change. Master the patterns, practice deliberately, and you’ll be ready.

Most candidates are still preparing for interviews from the last hiring cycle. And that’s becoming a bigger problem than people realize. 👇 My curated newsletter and free AI and job search Blueprint includes 50+ free resources and courses - guides, pdfs and complete books on topics ranging from prompt engineering, AI, AI research, AI agents, complete claude AI learning guide, latest daily jobs not found on LinkedIn. Sharing the link here. 🔗 https://lnkd.in/gtf9KZfa FAANG (2024) → MAG 7 (2025) → MANGOS (2026) → BANANAS (2027?) The acronym doesn’t matter. What matters is what it represents. The center of gravity in tech has shifted from consumer apps and growth metrics to raw infrastructure, foundation models, agentic systems, and autonomous pipelines. Hiring has shifted with it. Interview expectations have shifted with it. Yet most candidates still describe their experience the old way. They talk about what they built. The features they launched. The projects they owned. Those things matter, but interviewers are listening for something deeper. They want to see systems thinking. Impact on efficiency. How you handle AI-scale constraints. Here’s how to upgrade your stories for today’s interviews: 1. Scale vs. Compute Efficiency Old story: “I managed pipelines for millions of daily users.” New story: “I optimized token context windows and reduced GPU inference costs by 30% while maintaining throughput.” Interview tip: Quantify tokens/sec, memory usage, or cost-per-query. Expect questions on hardware-aware optimization. 2. UI Flows vs. Agentic Workflows Old story: “I designed onboarding flows to boost CTR.” New story: “I built guardrails and orchestration for multi-agent systems with error recovery and human-in-the-loop escalation.” Interview tip: Be ready to diagram agent interactions, tool calling, state management, and failure modes. 3. A/B Testing vs. Model Evaluation Old story: “I ran retention experiments on UI variants.” New story: “I created eval frameworks with benchmarks like MMLU and hallucination detectors to improve output accuracy and determinism.” Interview tip: Know metrics such as latency, safety scores, and red-teaming. Walk through continuous evaluation pipelines. 4. Monolithic Services vs. Distributed AI Systems Old story: “I scaled microservices.” New story: “I designed RAG architectures with vector stores, caching, and observability for consistent behavior at scale.” Interview tip: Discuss trade-offs in consistency, freshness, cost, and latency. Prepare for AI-specific system design questions. Before your next interview, review your resume: Are you listing responsibilities… or explaining measurable impact on systems, efficiency, and outcomes? The market didn’t dry up. It evolved. Reframe your past wins around the new stack. How are you adapting your interview strategy right now? Drop your current target layer (inference, agents, training, infra) + one challenge in the comments. Let’s troubleshoot together.

Are LeetCode-style interviews finally on their way out? I'm not sure, but we're seeing the first real cracks in its dominance. Big Tech has been clinging to algorithm interviews for almost two decades. They know it's a game, but it's a game that works for them. Leetcode successfully screens for intelligence, coding ability, and a willingness to grind. It's high precision but low recall - which is a tradeoff that they are fine with. False negatives are acceptable as long as they get consistently high quality hires. But LLMs have to force their hand eventually, right? Maybe. A good friend of mine was telling me just earlier this week that 4 of the last 7 interviews he conducted at Amazon were with candidates provably using AI tools to cheat (don't do this!!). So the status quo certainly isn't sustainable. What we do know is that respected mid-sized companies like Stripe, OpenAI, and Coinbase are already moving away from this model. They're using realistic coding challenges that more closely mirror real work. Build a query engine, not reverse a binary tree. Even smaller startups have pushed further, explicitly allowing AI tools in take-home projects. Their thinking here is straightforward: if engineers use LLMs on the job anyway, evaluate how they use it, don't pretend it doesn't exist. Where the Googles of the world used to set the industry standard for interviewing and everyone would follow, it looks like this time it'll be the reverse. The smaller companies are finding what works, and the big boys will follow suit once they finally reach a breaking point. I know we all welcome a change. Read more about How Tech Coding Assessments Are Splintering in 2025 in our latest newsletter https://lnkd.in/g5MmNkDB

For our recent data scientist hiring cycle, I replaced the Data Structures and Algorithms (DSA)-style Python interview with a code review format based on feedback from candidates in our last cycle, and it worked far better than expected. We still kept 1-2 live coding problems, but focused on data manipulation with numpy and pandas, the tools our team uses. In the previous cycle, 13 candidates cleared my hiring manager round, but only 2 passed the Python coding round, whereas this time, 11 cleared my round and 8 made it to the final stage. In the code review, we gave candidates 3-5 small functions, asked what they did, how they would improve them, and what the expected output was. I learned that abstract live coding questions felt more like a stress test than a real skill check. Code reviews were even better; they naturally led to chats about design, interpersonal feedback, and coding style. Training interviewers is also crucial because programming interviews work best as a collaborative process, not a test, meaning interviewers must be fully present. It’s amazing how much you can learn about a candidate by removing friction from the interview process. Candidate feedback helped me make changes, and the experience was far more positive this time. I’m still learning, but if you’re hiring data scientists, try adding code reviews to your interview structure - it might surprise you. #interviews #datascience #datascientist #codereviews

Technical interviews and hiring have changed dramatically in just a few years. Building our founding engineering team, I’ve been running technical interviews and took the same coding challenges we give candidates. I love watching candidates integrate LLMs into business logic and see how they use AI assistance. Luca Marturana, Kathy Li, and I have been reflecting on what it means to assess engineering talent in the age of AI: 1. We embrace coding agents and IDEs of choice. We’re already in a new programming paradigm -- no reason to pretend otherwise. The question is whether candidates accept AI output blindly or can interpret, debug, and coach it toward the optimal result. 2. Codebase comprehension matters. Can they map the structure quickly and form an intuition for how the system works? 3. Taste in architecture. Strong engineers challenge AI suggestions and think about data models, style, and scalability. 4. Judgment around level of autonomy. When to pair with AI (e.g., system design, prompt engineering to boost accuracy) versus delegating fully (e.g., refactoring, bug fixes). 5. Green flags: A custom personal workflow, .md files for Claude Code, and the ability to deliver 5–10x more features than baseline. I remember grinding through hundreds of Leetcode problems. That’s no longer the game. Today, the real test is whether someone can instruct, judge, and manage coding agents. The best engineers use AI intimately and multiply their productivity. It’s an incredibly exciting time to be an engineer. What have you found most effective in evaluating or demonstrating talent in this new era?

The other day, conducting a couple of technical interviews back-to-back perfectly captured where we are in tech right now. First candidate fired up Cursor, added the right documentation in its context, and started pair-programming with AI. They weren't just writing code; they were guiding the process — generating a PRD, scaffolding the implementation, and then applying their critical judgment to review and iterate on the output. Very much an orchestration role. The second candidate took a more traditional approach. Lots of manual typing, context switching to Google for syntax lookup, and digging through Stack Overflow for debugging errors. Honestly, it felt like watching two different eras of development unfold side-by-side. This experience reinforces why I'm leaning into an AI-first approach for interviews. No LeetCode. Instead, I want to see you at your best: • Bring the tools you actually use (yes, including AI). • Show me how you solve real problems • Demonstrate your thought process and engineering judgment. • Talk through the architecture and the trade-offs you're making. What I'm looking to see is their engineering judgment coupled with their productivity and efficiency. How effectively can they leverage the power of AI while maintaining the architect role and, ultimately, ship at an elite level? Observing someone write code without incorporating AI in 2025 increasingly feels like looking at a snapshot from a bygone era. What are your thoughts? How is this shift playing out in your hiring process or daily work? --- 👉 We're building our team with founding engineers who get this and are excited to define modern engineering practices from the ground up in the enterprise AI space. If that resonates, reach out!

I keep seeing the same pattern: teams building with AI, but interviewing as if nothing has changed. AI is accelerating execution. More importantly, it is shifting what actually differentiates great builders. We are moving beyond pure coding skill toward something broader: the ability to design systems that combine human judgment and machine capability. That shift changes what should matter in hiring: — AI has made first-draft velocity more accessible. Architecture intuition is the differentiator. Can someone design systems that endure as capabilities expand? — As execution speeds up, judgment compounds. When iteration is cheap, taste and prioritization become scarce skills. — The highest-leverage builders know what not to overbuild. They resist hard-coding around constraints that are likely temporary. — As individual output scales, the ability to create alignment determines whether that leverage compounds across a team. The bar is no longer just “Can they code?” It is “Can they design systems and teams that gain leverage as AI capabilities expand?” How are you evolving your hiring bar?

Today, I witnessed my team lead interviewing a Full-Stack Web Developer with 2 years of experience, and I observed some interesting insights. The interview started with basic technical questions, which the candidate answered well. Then, my team lead gave him a coding task: 👉 "Write a function that removes duplicate values from a nested array while preserving its structure (PHP)." The candidate struggled and couldn’t solve it. So, my lead gave him another task: 👉 "Write a function that groups an array of associative arrays based on a specific key." Again, he couldn’t solve it. At this point, he admitted: "I don’t always have syntax and logic rules memorized, as I usually rely on resources like Stack Overflow and ChatGPT. Could you assign me a real-world task instead?" Here,I decided to step in. Since I was working on WooCommerce project, I gave him this challenge: 💡 "Implement a functionality where a logged-in user gets a 15% discount if they place two orders in a single day." We allowed him to use Google, ChatGPT, or any other resources. To our surprise, within 1.5 hour, he understood WooCommerce functionality, analyzed the existing code, and successfully implemented the feature. ✨ We ended up hiring him! ✨ Key Takeaway Many traditional interview techniques overemphasize logical puzzles and syntax-heavy questions, yet modern developers approach problem-solving differently. Rather than memorizing solutions, they adapt, learn, and implement on the spot using the tools and resources available to them. 💡 Instead of testing how well candidates remember syntax, we should focus on how effectively they apply problem-solving skills to practical challenges. What’s your take on this? Let’s discuss in the comments! 👇 #HiringTechTalent #TechCommunity #WhatDoYouThink

Coding interviews are becoming a joke In a conversation with AIM, Pratham Patel, a member of Rocky Enterprise Software Foundation, shared his perspective. “As an interviewee, I can say…that the interview process has become more of a formality than an actual test. It is unfortunate to see that interviewers are more interested in whether the candidate can provide code with flawed reasoning rather than understanding if they truly grasp programming and the art behind it.” Meanwhile, Krishna Vij, vice president of IT staffing at TeamLease Digital, pointed out that AI is pushing companies to rethink their hiring strategies. “How we assess technical talent is evolving and AI-driven tools are accelerating that shift. If a candidate uses AI to clear a coding test, we must ask ourselves: Are we truly measuring their skills or just their ability to leverage technology? This is why companies must rethink their hiring strategies, as standard coding tests alone are no longer sufficient in the current situation.” Vij added that the company is seeing a stronger push towards live coding interviews, project-based assessments, and in-depth problem-solving discussions. The focus now is on critical thinking, adaptability, and real-world application rather than just syntax and speed. Hiring processes must keep evolving because as AI gets smarter, so must the way companies evaluate talent. Similarly, Rahul Veerwal, founder and CEO of GetWork, reinforced this perspective and stressed the need for multi-layered assessments. “At GetWork.ai, we’ve seen firsthand how AI can enhance hiring, but we also recognise its potential for misuse. While AI-assisted cheating poses a challenge, the solution isn’t to abandon coding rounds but to evolve how we assess talent,” he explained. Full story link in the comments.

This video I made 3 years ago just helped a software engineer pass his interview last week. Even with new coding assist tools, the information is still valid and here's why (and some updates...) That's because the way interviews are evaluated hasn't really changed. Here's what most candidates don't realize: Your interviewer isn't just grading your answer. They're grading how you got to it. That's why: • Vague questions are on purpose. They want to see if you'll ask clarifying questions before you start solving. • The simpler solution you can actually finish beats the perfect one you can't. Mention the optimal approach exists, say why you're skipping it, and move on. • Asking "is this right?" hurts you. Coworkers think out loud together. They don't ask for permission. What I'd add today: • Some companies now let you use AI coding tools in interviews. That means reasoning out loud matters even more, because it's the only thing left to evaluate. • If you're grinding random LeetCode problems, switch to NeetCode. Around 25 mediums is still the right target. But the more you practice the more you'll improve. Software engineers passing interviews in 2026 are doing the same thing they were doing in 2023: showing how they think, not just what they know. The tools changed. The evaluation didn't. Clickable link to the full video in the comments. https://lnkd.in/e3Ed3zcA