AI Maximalism Hiring: How ClickUp Tests Engineers With AI

Software engineering interviews are testing for a skill that barely matters anymore. The LeetCode puzzle — reverse a binary tree, optimize a graph traversal — measures a candidate’s ability to write code from memory. In 2026, nobody ships code from memory.

Jay Hack, Head of AI at ClickUp — the $4B work management platform that acquired his coding agent company Codegen — has replaced the traditional interview process with what he calls “AI maximalism.” Candidates get a powerful AI, an ambitious task, and a clock. What happens next is more revealing than any whiteboard session.

Why LeetCode Is Dead

The old interview tested whether you could solve a constrained algorithmic problem in 45 minutes with no tools. That skill correlated with engineering ability when writing code was the bottleneck. It doesn’t anymore.

“We’re not measuring your ability to solve a LeetCode style programming question,” Jay says. “We instead give you a very ambitious task and we give you a powerful AI and we say, accomplish this and talk me through how you’re thinking about it.”

The shift isn’t just about allowing AI tools. It’s about testing a fundamentally different capability: can you orchestrate an AI to produce output that exceeds what either of you could do alone? Jay’s team watches how candidates interact with the AI — how they decompose problems, how they handle unexpected output, how they course-correct when the agent goes sideways.

What “Best in Class” Means Now

Jay pushes back on the idea that AI lowers the bar for engineering talent. The bar moved, but it didn’t drop.

“We still want the best in class programmers. It’s just that what is best in class has changed. Best in class now means you actually are very efficient at working with AI and somebody who’s without AI going up against me with Claude 4.7 — like I’m going to absolutely destroy them.”

The candidates who stand out aren’t the ones who can write the cleanest code by hand. They’re the ones who can set up the right context, catch the AI’s mistakes quickly, and combine outputs from multiple agent runs into something coherent. It looks more like directing than coding.

The Assembly-to-English Abstraction Chain

Jay frames the shift through the full history of programming languages. Assembly gave way to C. Developers initially examined the assembly output of C compilers to check correctness — “is this right? Is there a way I can do this better?” That skepticism faded as compilers matured. C gave way to Python. Now LLMs represent the next layer.

“Software engineering is the profession that has basically throughout its entire history been inventing itself out of a job,” he says. “The next level of abstraction is English.”

This creates a concrete prediction about role convergence. If the primary interface to building software is natural language, then the gap between a product manager and an engineer narrows dramatically. Jay’s wife, a trained product manager with no engineering background, now vibe-codes functional prototypes that “actually work and feel good.” He’d hire someone with her profile — not as a staff engineer, but as a “product generalist” who can prototype, prioritize, and ship.

What Candidates Should Actually Demonstrate

When Jay evaluates candidates, the signal he’s looking for isn’t technical skill in the traditional sense. It’s curiosity and intensity.

“The people who are left are people who love computation. They love building stuff. They really get a dopamine hit from shipping code,” he says. He keys on side projects, hackathon participation, and whether candidates have “an actively developing mental model of the space.”

The practical test: can they take an unfamiliar, ambitious problem and use AI to produce a working solution while articulating their reasoning? Whiteboarding still happens for systems design, and candidates are welcome to consult ChatGPT mid-interview. The goal isn’t to test what’s in their head — it’s to test what they can produce with the tools available.

FAQ

What is AI maximalism hiring?

AI maximalism hiring gives candidates full access to AI tools during interviews and tests their ability to orchestrate AI agents to solve ambitious problems. ClickUp pioneered this approach at Codegen and carried it into ClickUp’s hiring process. Candidates are evaluated on how they decompose problems, interact with AI, and produce working output — not on memorized algorithms.

Why are companies moving away from LeetCode-style interviews?

LeetCode tests the ability to write algorithms from memory, which no longer reflects daily engineering work. In 2026, engineers primarily interact with AI agents through natural language. Jay Hack says “best in class” now means being efficient at working with AI — a skill that LeetCode doesn’t measure. The bottleneck has shifted from writing code to orchestrating agents.

What skills should software engineers develop to stay competitive?

Focus on AI collaboration: learn to decompose problems for agents, evaluate AI output quickly, and combine results from multiple agent runs. Jay recommends building side projects with AI tools, attending hackathons, and developing “an actively developing mental model” of how AI capabilities are evolving. Traditional coding skills still matter but are insufficient alone.

How is the software engineering role changing in 2026?

Engineers increasingly interact with LLMs through English rather than writing code directly. Jay describes an “abstraction chain” from assembly to C to Python to LLMs, where each layer makes the previous one less relevant for most practitioners. The role is converging with product management as both use chat interfaces to build software.

What is a product generalist role in AI companies?

A product generalist combines product management instincts with AI-powered prototyping ability. They communicate with engineers and customers, shape product direction, do rapid prototyping using AI coding tools, and work with designers — without necessarily writing production code. Jay describes it as a role that’s “flourishing today” as the boundaries between product and engineering blur.

Should non-technical people learn to code in 2026?

Jay says non-technical people are “the biggest winners” of the AI wave. His wife, a product manager with no engineering training, vibe-codes functional prototypes using AI tools. For non-technical professionals, the opportunity isn’t learning traditional programming — it’s learning to build with AI as the intermediary. Understanding what software can do matters more than knowing how to write it.

How does ClickUp evaluate engineering candidates differently?

Candidates receive a powerful AI tool and an ambitious task with a time limit. ClickUp evaluates how they interact with the AI, decompose the problem, handle unexpected outputs, and think about the solution space. Whiteboarding for systems design still happens, and candidates can consult AI tools freely during that process too.

Will there be fewer software engineering jobs in the future?

Jay thinks organizations will “slim down” rather than hire more engineers. Small teams of high-agency engineers with AI tools outperform larger teams without them. Some of ClickUp’s fastest-moving products have very few engineers. The shift favors builders who love the work — Jay says the COVID-era influx of people seeking “an easy job at Google” is filtering out.

What is the assembly-to-English abstraction chain?

Jay’s framework for understanding how programming evolves: assembly led to C (developers initially checked compiler output), C led to Python, and now Python leads to LLMs where the primary interface is English. At each transition, the previous layer becomes something most practitioners never touch directly. The pattern predicts that writing code by hand will become as specialized as writing assembly is today.