sethserver.com Subscribe
A magnifying glass hovers over stacked books (in purple and white) against a coral pink background, revealing a close-up view of a green circuit board or microchip through the lens.

How to Hire When AI Makes Every Resume Look Perfect | Screening Real Engineers in 2026

By Seth Black • Updated: September 17, 2026

Startups · 4 min read

Like this kind of writing? Get one email a week with notes on startups, AI, and the occasional strong opinion about Python: subscribe to the newsletter.

You can feel the AI resume wave the second you open your inbox. Everything sounds like it was written by ChatGPT. Perfect grammar. Zero personality. Somehow everyone has expert-level experience with the exact obscure tool you listed... including the one you added because you thought nobody would have it.

This isn't a moral panic about people using tools. I don't care if someone used an LLM to fix commas. I care if the resume is a costume.

Resumes used to filter out bad candidates. Now they filter out honest ones.

A while back, someone cheated our hiring process at a Series A startup I was at. They hoped we wouldn't notice long enough for them to learn Python. That was the last time I treated a resume + a quick quiz like it was a truth machine.

Coding quizzes are easy to game. Resumes are even easier.

If you still treat the resume as the "truth layer," you're going to hire people who are great at sounding right. Which is different from being right.

When everything reads like a TOS agreement

Most humans write like humans. Even smart ones. Especially smart ones.

AI-generated resumes tend to:
- over-explain obvious things
- stack buzzwords with no story
- avoid specifics that could be checked

A simple screen that works: ask for a quick, informal writing sample during the first touch.

Example prompt:

"In 5-6 sentences, describe your ideal work environment and how you like to get feedback."

If they respond with something about "synergistic collaboration in a fast-paced environment," you just saved yourself three rounds of interviews. This is the linguistic uncanny valley of AI writing—grammatically perfect but eerily lifeless.

The too-perfect skill match

LLMs are great at mirroring job descriptions. You'll see resumes that hit every keyword like they were generated by a search engine.

Real engineers have opinions. They've been burned. They have scar tissue from the last framework that promised everything and delivered a memory leak.

They say stuff like:
- "I used X, but I wouldn't pick it again for Y."
- "I shipped Z, and the part that hurt was A."
- "I'm good at B, still learning C."

If every bullet point is a flawless match, I start asking for friction. Tell me about the time it broke. Tell me what you hated. Tell me what you'd do differently.

Cross-checking "AI companies" (and other fantasies)

I once worked at a SaaS company that claimed it used AI to automate boring, error-prone financial tasks. A few weeks in, support/services resisted showing me anything. I was literally told to "come to our office and watch us work" by a team in South Asia.

Code review revealed the "AI" was three people in Bangalore copy-pasting from Excel. Which, to be fair, is technically automation.

That experience made me allergic to vague claims. Same with candidates. When your AI startup isn't actually an AI startup, the same red flags appear: vague technical descriptions, resistance to showing actual systems, and claims that can't be verified.

So I cross-check in a pretty normal, slightly paranoid way:
- I'll check LinkedIn to see if the story holds up.
- I'll look for any writing they've done - even a Medium post about debugging Postgres.
- If they're applying for an engineering role, I want to see code somewhere.

No repo isn't an auto-reject. Plenty of people work in private codebases. But if you claim you built systems and there's zero artifact trail anywhere, I'm going to ask for a concrete walkthrough.

Replace quizzes with conversation + real work

My best hiring signal now is a collaborative, realistic exercise.

Not a trick puzzle. Not "invert a binary tree" while I watch you sweat and pretend this has anything to do with the job.

More like:
- "Here's a messy CSV export. Map it into a clean schema."
- "Here's a small Python service. Add one feature and one test."
- "Here's a support ticket. Explain how you'd debug it."

Then we do it together for 45 minutes.

I'm watching for:
- how they think out loud
- whether they ask good questions
- whether they can admit confusion without melting down
- whether they can reason about constraints

AI can write a resume. It can even write code. It can't fake a coherent engineering brain in a live, messy situation for long. This is why not wasting time in programmer interviews means focusing on collaborative work instead of abstract puzzles.

Small signals beat one giant signal

At a hardware startup, I watched a 50-line Python script outperform a vendor's ML platform that cost us $40k/year. Same lesson applies to hiring.

Stop trying to solve hiring with one giant signal (resume, quiz, pedigree). Use a few small signals that are harder to fake:
- spontaneous writing
- a live walkthrough of past work
- a short paired task
- a discussion about tradeoffs

I'm not looking for someone who interviews well. I'm looking for someone who can debug production at 2am without having an existential crisis.

-Sethers

Share this post
Newsletter

One email, once a week.

Notes on databases, systems, and the occasional strong opinion about Python. No spam, unsubscribe anytime.

Seth Black
Written by

Seth Black

Engineer and founder based in Texas. Writes about databases, AI, and running things in production. Embeds in small teams as lead engineer.

More from Startups

View all →
Startups

What Actually Goes in a Consulting Contract

Sep 17, 2026
Startups

Productizing a Service: From Hours to Licenses

Jul 26, 2026
Startups

The MVP That Is Not a Product

Jul 11, 2026