AI assistants make standard coding interviews a poor way to tell whether a candidate can build useful AI products.
2
Small companies have to compete with large-company brands, career prestige, and perceived stability as well as compensation.
3
Hiring processes should use realistic workplace simulations to observe how candidates work with AI, handle ambiguity, communicate decisions, and adapt to changing requirements.
Summary
Beth Glenfield argues that AI has broken the technical hiring process. Candidates can use AI assistants during interviews, while AI-customised resumes and coding tools make it harder to tell what someone can do unaided. LeetCode-style puzzles also measure skills that do not match the work of building AI products. Glenfield says the candidates companies need are creative problem solvers and collaborative engineers who can use AI tools, understand business impact, contribute to open source, and make decisions when requirements change. DevDay's proposed alternative is a workplace simulation in which candidates build features for a company's domain alongside AI agents with different personalities and strengths. The process observes collaboration, delegation, communication in pull requests and tickets, mentoring, and adaptation. Glenfield is also direct about the market: small companies cannot interview hundreds of people and absorb bad hires, while candidates compare them with Google and Meta on more than pay.
AI assistants have made coding interviews poor evidence of engineering ability
Glenfield opens by asking whether interviewers and interviewees already use AI in recruitment. She says some services help candidates cheat in interviews, with one service reportedly heading toward $1 million in annual recurring revenue, and claims that AI assistants now appear in one in three interviews. Her conclusion is that companies may be testing who has the best coding assistant rather than who can build the best product. She also points to reported success rates for AI-assisted LeetCode interviews at Google and Meta. Sam Altman's advice, as Glenfield quotes it, is to learn the best AI tools. That changes what an interview result means.
Companies compete for candidates through reputation and stability as well as pay
Glenfield says technical jobs have changed because candidates are weighing brand recognition, career prestige, and perceived stability alongside compensation. Layoffs make security more important. A Series A startup therefore has to compete with Google's AI division for the same person, even when it offers an interesting role. She contrasts smaller companies with Google and Meta, which can interview a hundred candidates, hire five, and offer strong paychecks. Most companies do not have that option. Glenfield says a single bad hire can cost between $20,000 and $60,000 through the hiring process, and small teams cannot afford an engineer who does not know how to ship AI products.
The engineers companies need are not the people optimized for LeetCode scores
The candidates Glenfield wants companies to identify are creative problem solvers and collaborative leaders who can work with AI. They build AI tools, use AI libraries, contribute to open source, and understand how their technical work affects the business. LeetCode does not measure those abilities well because its puzzles are unlike the work many engineers will do on the job. Glenfield wants hiring teams to observe how candidates delegate tasks, handle ambiguity, and respond when requirements change during a sprint. The test should show how someone works in the conditions of the role rather than how well they memorize algorithms.
Realistic simulations can reveal how candidates collaborate with AI
DevDay's proposed process replaces traditional coding interviews with workplace simulations. Candidates work alongside AI agents that have different personalities and roles, including a perfectionist, a pragmatist, a security expert, and a junior developer who needs extensive mentoring. The candidate has to make ordinary tradeoffs while building features for the company's business domain. Glenfield says the assessment should examine collaboration with AI, responses to ambiguity, communication of technical decisions in pull requests, comments, and tickets, mentoring, and adaptation as circumstances change. This creates a setting in which a company can show candidates its engineering culture at the start of the process and observe how they perform in that environment.
AI will change engineering jobs without eliminating the need for engineers
Glenfield cites Marc Benioff's claim that Salesforce saw a 30 percent productivity increase after replacing people with AI, while saying she would like to see the data. She also cites Mark Zuckerberg's prediction that AI will handle mid-level engineering work by the end of the year and a TechCrunch report about entry-level engineering jobs disappearing. Her conclusion is that engineering work will become different rather than simply vanish. The jobs she describes require creativity, collaboration, business judgment, and the ability to work with AI. Hiring therefore has to identify those abilities instead of relying on coding performance alone.
"Jobs that require creativity, collaboration, and the ability to work according to business judgment and not just code."06:10
Who should watch
You are hiring engineers at a startup and cannot absorb a bad hire or compete with Google and Meta on compensation alone.
Your interview process still depends on LeetCode puzzles, even though candidates can use AI assistants during technical interviews.
You want to assess how engineers build AI products, work with AI tools, communicate decisions, and respond when requirements change.
