The First Thing I Did After Getting Laid Off Was Throw Away My Résumé

In early February, I got laid off. And like a lot of people in that position, my first thought was, “Well, I should probably update my résumé.”

I dug around my docs folder, found my résumé, started reading, and immediately wanted to close it again. My résumé sucked. I hadn’t really touched it since 2018, and it described a version of me that didn’t exist anymore. Trying to edit it felt like trying to renovate a house with a bad foundation.

The obvious thing was to bust out an LLM and ask it to rewrite my résumé. But LLMs are only as good as the sources you give them, and my source was crap. So instead of building from my existing shoddy résumé and then revising it ad nauseam, I decided to spend my effort creating excellent reference material.

Well, I'm unexpectedly unemployed, and it's time to build my resume up and get it out there. Rather than starting with an existing resume that I haven't updated in years and building forward, I want you to interview me about my work history, my accomplishments, what I'm learning, what I'm good at, and help me build a resume that stands out.

Interviewing Instead of Drafting

As a product manager, I’d already discovered that large language models are surprisingly good interviewers. As long as you answer their questions thoroughly, they’ll keep asking questions and digging into interesting threads. Along the way, they’ll often uncover details you’ve forgotten or never thought to mention.

I didn’t prepare at all or treat this as a prompt-engineering exercise, I just rambled like I was talking to another person instead of trying to carefully craft perfect answers.

Claude started by building up context:

Give me the list of roles you've held: company, title, rough dates. One line each. We'll decide together what makes the cut, then go deep on the ones that matter.

What kind of role are you targeting? Senior PM? Director? Are you open to adjacent roles like Product Strategy, Solutions Architecture, or Technical Program Management? Or is it strictly product management?

But very quickly, it switched from chronology to impact and started challenging me to think like a hiring manager. When we got to my most recent role, the questions looked more like this:

You were at Mondoo for 4.5 years. I already know a lot about what you did day-to-day. What I need is the impact framing. Tell me:
1. What was Mondoo's stage when you joined? (headcount, funding, product maturity)
2. What were the biggest things you shipped or drove? Not features — outcomes. Revenue unlocked, customers landed, market position changed.
3. What did you inherit vs. what did you build from scratch? (processes, product strategy, tooling, team practices)
4. If your boss had to describe your single biggest contribution in one sentence, what would he say?

Over the next two and a half hours, Claude would often propose a narrative about my career or suggest a different way of framing an accomplishment. Sometimes I agreed. Sometimes I pushed back because it wasn’t quite true, or because it emphasized something I didn’t actually want to be known for. Pushing back forced me to articulate why I disagreed, and that usually left me with a better answer than I started with.

As we started wrapping up, Claude asked for a link to my GitHub profile. I’d completely forgotten about the bio I’d put there years earlier:

Claude stomped the brakes.

Wait. “Partly responsible for the proliferation of the word DevOps”? Is there a story there? That’s a hell of a claim if it’s legit.

It was legit.

I explained that I’d spent years teaching Chef’s training classes, each of which included a substantial module on what DevOps actually was. I’d delivered that course hundreds of times, spoken at conferences six to twelve times a year, and continued speaking at DevOpsDays events long after leaving Chef.

That’s not a footnote. That’s a positioning statement.

It had a point.

Independent interrogation

At the end of the exercise, I had a lengthy Markdown document that summarized everything Claude had learned about my career. I’d been unemployed for all of four hours.

By then, the weight of what was happening was starting to hit. I had a family to support, a résumé that had already proven itself inadequate, and no idea how long the job search was going to take. Claude had asked good questions, but it was still just one interviewer with one set of instincts. I wanted independent interrogation from different angles.

So I did something a bit ridiculous: I opened a fresh conversation in ChatGPT and started over from scratch. And when that interview finished, I did it again with Gemini.

The questions weren’t identical. ChatGPT fixated on different parts of my career than Claude had, and Gemini went off in its own direction again. I’d wanted three independent interviewers with different instincts, and that’s exactly what I got.

Because each interviewer had discovered things the others had missed, I started passing the transcripts around. Claude got ChatGPT’s notes. ChatGPT got Gemini’s. Gemini got Claude’s.

Each received the same instruction:

I went through a similar interview with another interviewer. Here’s the transcript it produced. Read through it, ask me any follow-up questions you think are worth asking, then fold anything new you learn back into your understanding of my career.

Not one of them said, “Yep, looks good.” Sometimes the second interviewer would notice an inconsistency and ask me to reconcile it. Or maybe it would latch onto a story that the first interviewer had barely touched. Sometimes, it went after word choice directly:

“Catalysed” sounds like someone ran it through Grammarly and a thesaurus. You don’t need clever verbs. You need accurate scope. Let’s strip it back to plain English.

I’d expected things to converge quickly. Instead, this was another round of discovery.

So I kept cycling the transcripts until the follow-up questions dried up and the documents all started saying essentially the same thing, then merged them. At that point, I had what I wanted: a comprehensive, defensible write-up of my career that three models had spent hours interrogating from every angle.

Only then did I ask each of them to generate a résumé. Then I repeated the same convergence cycle until those drafts settled too.

Human feedback without blind deference

Once the résumés themselves converged, I exported the document to Google Docs and shared it with some trusted human reviewers. As their feedback came in, I didn’t just paste the comments into the panel of ‘bots. I made sure to contextualize it first, and that led to really excellent results:

All of this feedback is from Thomas, a one-time customer and long-time friend, who recently retired from his job as an L7 Engineer at a FAANG to work for a startup. I trust him and his expertise, but don't assume his feedback is correct. Evaluate it against the criteria, "What will get Chip hired the fastest." His feelings and mine don't matter, only quickly getting to interviews.

This prompt seemed to kick the agents out of “a human said it, so it must be true” mode. They became much more willing to challenge or reframe the feedback instead of accepting it. In all I had a half-dozen human reviewers, and I made sure the panel knew a little bit about each of them before considering their feedback.

Thomas is an engineer. His feedback reflects how he reads resumes — carefully, linearly, with opinions about craft. Recruiters and hiring managers don't read that way. They scan.

Every prompt contained the same instruction: optimize for getting me hired. On average, each LLM panelist recommended rejecting about 60% of the changes suggested by my human reviewers, and helped me workshop the rest, but of course final judgment was always mine.

Whenever you’re feeding expert feedback into an LLM, tell it who the expert is, why you trust them, and what objective should ultimately win if the two disagree.

What actually mattered

AI has made it so easy to generate résumés, and so easy to screen them, that large parts of the hiring pipeline increasingly feel like bots talking to bots. How do you differentiate and stand out in that kind of environment? By building from better context than everyone else.

I could have spent two days polishing the résumé I’d written years earlier. Instead, I threw it away and spent those two days building the best possible understanding of my own career. Only then did I ask the models to turn that into a résumé.

I also don’t think repeating the exercise with three different models was a gimmick. I wasn’t looking for consensus from three copies of the same brain; I wanted three independent interviewers with different instincts. I chose Claude, ChatGPT, and Gemini on purpose. In practice, each one surfaced things the others had missed. Having them compare notes produced a better understanding of my career than any one of them reached alone.

This process also happened to fit me particularly well. I’m almost fifty years old, I’ve been working since I was sixteen, and I make my living talking to people. I had decades of stories to draw on, and I’m comfortable telling them. If you’re just starting your career, or you hate talking about yourself, your mileage may vary.

Did it work? I can’t prove it. I didn’t establish a control group by sending out the crappy résumé I’d been using a week earlier.

What I do have is this:

Over the next eight weeks, twenty-three opportunities turned into eight interview loops and, ultimately, one accepted offer. At the time I accepted that offer, I was still actively engaged with four other opportunities. 

The résumé wasn’t the only variable, and maybe the old one would have performed just as well. I do know that I came away with a far better understanding of my own career than I started with.

In the end, I think these are the parts of the process that actually mattered:

• Don’t let an existing artifact anchor the exercise.
• Build rich source material before you build from it.
• Seek out independent perspectives instead of settling for the first plausible answer.
• Give your models context about expert reviewers and a clear objective when weighing their feedback.

None of this works if you start from the old document and expect the LLM to rescue it. Instead, you have to understand your career first, and let the résumé fall out of your subject matter expertise. 

Don't outsource the thinking. Outsource the formatting.