How to hire engineers for strong judgment and taste

Estimated reading time: 6 minutes

Key takeaways:

• Judgment and taste are now the most important engineering hiring signals. AI handles execution; developers must evaluate what agents produce and make product decisions faster than ever.
• The best companies are ripping up their hiring processes. Open-ended take-home projects and watching candidates’ AI prompt chains reveal how engineers think, not what they know.
• The biggest red flag: candidates who can’t explain their decisions. Curiosity and nuanced thinking matter more than the answer itself.

---

Every software developer knows there is no objectively right way to build a system. There are always multiple solutions that are technically correct and would work, but at the same time, not all paths are created equal. Developers need to constantly think through tradeoffs. 

This hasn’t changed now that AI agents are doing much of the technical execution. In fact, developers must now also evaluate the decisions agents make.

For example, generative models are trained to add new code – not so much to simplify and delete code. Developers need to bring strong judgment to ensure the agents don’t write overly complex code that will just create bloat and problems down the line. 

“All these things were always important, and are now even more important,” says Andrew Hsu, co-founder and CTO at AI language tutoring company Speak. 

What’s more, in addition to the responsibility for agents’ code, developers are also taking more ownership of products from-end-to-end.

So what does judgment and taste look like in today’s engineering landscape? How can engineering leaders evaluate candidates for these traits while hiring? Some say it’s all about finding ways to dig deeper into how a candidate thinks. 

Increased ownership demands strong taste

Now that the lag from “here’s a spec” to “now write the code” has gone from weeks to hours, you can’t wait for decisions anymore, says Hsu. He believes engineers are more effective if they have the strong product sense and technical judgement to make more decisions themselves. 

Luke Behnke, VP of product at Superhuman, who is building out the company’s new team of Forward Deployed Engineers (FDEs), echoes this sentiment, adding that FDEs are the best role for demonstrating this.

FDEs embed directly into customer organizations to develop and deploy customized technical solutions for the customer’s unique needs, so they have to be exceptionally thoughtful about their decisions and able to convey their reasoning to customers. “Our FDEs are on calls with customers constantly. They’re getting that feedback loop faster than anybody,” he says. 

FDEs are the hot new job in engineering, with postings for the role soaring more than 800% between January and September 2025, according to analysis by Indeed. Combined with the new hyper-fast speed of development and general shift toward engineers owning product outcomes from end-to-end, the widespread embracement of the FDE model is putting an increasing number of engineers in the driver’s seat. 

“[FDEs] are also tasked with doing whatever it takes, even outside of the bounds of what a traditional product team would do, to make sure that the customer is successful,” says Behnke. “So that is another reason why taste matters so much. They are going to be making a lot of decisions that we’re going to have to live with for the next few years for that customer.”

Interviewing for judgement 

Peter O’Connor, director of platform engineering at Stack Overflow, says he’s been thinking about how to redesign the interview process to better surface candidates’ thinking and ability to explain their decisions. 

“I think the interview as a we have to do the thing here to prove something from your brain directly, as memorization, is just not as useful. I need to understand thinking,” he says. 

The company has historically done pair programming and evaluated on how the candidate worked, not what they came up with. If the engineer would just go ahead and do the coding without asking questions, even if it was perfect, they didn’t pass, he O’Connor explains.

To meet the new demands of engineering, he wants to take it a step further. He’s interested in take-home assignments, but he’s wrestling with the pros and cons of the approach.

On one hand, he likes the idea of the candidate working through something on their own, bringing it back, and then having a conversation about it with them.

On the other, he worries such assignments could create inequities between candidates who have the luxury of time and those who don’t.

Additionally, cutting edge AI tools are expensive, and he doesn’t want to see anyone priced out of being competitive. Potential solutions to the latter could include providing temporary access to the team’s AI tools or covering a one-month subscription to a platform they need.

Hsu says Speak ripped up its entire process for evaluating engineer candidates, now leaning into take-home test projects. Specifically, they task candidates to build a light product spec with agents, intentionally giving an open-ended prompt so they can see where the engineers go with it and the decisions they make along the way.

“There are many ways of interpreting it. The candidates now have to basically set things up however they want and make the tradeoffs,” he says, adding that it felt very obvious they had to change the process “because the job itself had fundamentally changed.”

Behnke is looking to zero in on product sense, searching for engineers through the lens of “who would win the hackathon with the most interesting customer solution no one else would’ve thought of?” The team at Superhuman is not actually hosting hackathons to find such candidates, but they are challenging candidates to think of product ideas from scratch. 

This starts with a hypothetical, for example, “come up with the best AI toaster.” Behnke says this gets them to think through everything from limitations to the customer interface. He evaluates based on how well they can come up with a solution in a reasonable time frame, and there’s no wrong answer. 

“What’s important to me is, why did they do that? Does it demonstrate sound product thinking [around the goal]?” he says. 

For the technical part of the interview, he gives them another fictitious problem for them to code with AI and watches their prompt chain carefully. He finds this to be the best way to see how their thought process works and get a sense of whether they can hone in with the AI on the right kind of problems.

“There is still so much technical judgment that has to come into play in terms of, is this the most efficient way to do it?” he says. “So our technical interview also does take into account the question of, if you’re looking at some code that you wrote with AI or that AI wrote for you, are you able to assess the strengths and weaknesses of that code?”

Green and red flags 

Overall, O’Connor thinks hiring at this moment is about taking a step back and revisiting some of the fundamentals. Engineering leaders will need to listen for certain cues, he says. For him, he’s looking for curiosity, deep questioning, and nuanced thinking – and it’s more about the way they give an answer than the answer itself.

The biggest red flags are if an interview is one-sided or if the candidate can’t explain their thinking. Hsu puts it plainly: if an engineer can’t explain why they made a design decision over an equally valid alternative, “that’s obviously a pretty huge red flag.”

Importantly, evaluation for taste and judgment requires the interviewer to bring strong judgement as well. While interviewers used to be able to focus more on the code, now they need to have deeper conversations, which means coming in with strong communication skills.

The other key, according to O’Conner, is knowing what you’re looking for in the first place. This is vital to asking the right question to make sure you’re actually filling your gaps. At Stack Overflow, he’s not trying to build a team of high-functional, “explain everything” engineers; he’s trying to build a well-rounded baseball team. 

“I need a shortstop and I need a catcher,” he says. “I don’t need 80 pitchers.”

Get articles like this in your inbox

Choose your LeadDev newsletters to subscribe to.