Hiring debate: builders vs grinders

A hiring post about software engineers split recruiters and developers over one question: in 2026, should candidates prove skill with shipped apps or with interview drills. (x.com) The post argued that for MongoDB, Express, React, and Node and Next.js roles, a public portfolio now signals more than stacks of certificates or long LeetCode streaks. It pointed specifically to full-stack projects and “agentic” artificial intelligence builds that employers can inspect in public. (x.com) “Agentic” software means an application that does more than answer a prompt once: it can plan steps, call tools, and keep enough state to finish multi-step work. OpenAI’s current developer docs describe agents as applications that plan, call tools, collaborate across specialists, and keep state to complete tasks. (openai.com) That distinction has changed what a “project” looks like in hiring conversations. A candidate can now show a deployed app, a code repository, tool use, and traces of how the system handled real tasks instead of only a résumé line saying they studied a framework. (openai.com) LeetCode still markets itself as a platform to “level up” coding skills and prepare for jobs, and many companies still use algorithm screens in early rounds. Pattern lists such as Sean Prashad’s LeetCode repository remain widely used because they train candidates on recurring interview problem types. (leetcode.com, github.com) Certificates also have a place, but even Microsoft’s community guidance for a MongoDB, Express, React, and Node developer in 2025 framed them as supplements to current-market skills rather than a substitute for proof of work. The hiring argument in this thread turned on that gap between studying and shipping. (learn.microsoft.com) The debate lands as artificial intelligence tools make take-home coding, debugging, and prototyping faster than a year ago. OpenAI’s current documentation now includes dedicated tracks for building agents, coding agents, tool use, orchestration, and evaluation, which gives candidates more ways to build visible software quickly. (openai.com, openai.com) Recruiters and hiring managers are not all moving in the same direction. Some teams still want algorithm interviews because they are standardized and easy to compare across applicants, while others are shifting toward repo reviews, practical assignments, or live build sessions that look more like day-to-day engineering work. (leetcode.com, github.com) For candidates, the practical takeaway from the thread was concrete: a public build with code, deployment, and a clear explanation of what problems it solves is becoming a stronger hiring artifact than another certificate badge alone. The argument was less about ending interview prep than about what counts as evidence when anyone can say they know the stack. (x.com)