The problem with how we hire right now

The current hiring stack goes: resume → portfolio → phone screen → panel → offer. Every step is human-bottlenecked, proxy-heavy, and designed around filtering for the wrong things.

A resume tells you where someone worked and what titles they held. A portfolio tells you what got shipped. A phone screen tells you whether someone can answer questions they've been practicing for weeks.

None of it tells you how someone thinks under pressure. None of it tells you how they'd handle your specific problem. None of it tells you where they break down.

We've been using the same hiring system for decades. Meanwhile the work has changed completely. And now the tools have too.

What a skills file actually is

A few months ago I built a page on my portfolio site called a skills file.

Not a resume. Not a case study. A deployment spec.

The concept borrows from how AI systems describe themselves — a structured document that tells you not just what someone does, but when to deploy them, what conditions they operate best in, and critically, where they fail. Most resumes are pure upside. A skills file includes the constraints.

Mine covers: operating context, proof of work with real metrics, how I think, my position on AI, my toolkit, active side projects, and a section I called "where I break down" — a direct description of the environments where I'm a bad fit.

That last section is the one people react to most. It feels counterintuitive to put your failure modes in a hiring document. But that's exactly what makes the whole document trustworthy.

The experiment

I took my skills file and reformatted it as an AI system prompt. The idea: what if an employer could drop this into their AI and interview me directly — before we'd ever spoken?

The system prompt instructs the AI to answer as me. First person. No hedging. Back claims with real examples. Push back when challenged. Don't perform — think out loud.

Then I ran it. I played the interviewer. The AI played me.

"You say you get hired into broken design orgs and make them the advantage. What does broken actually look like when you walk in the door — and how do you know within two weeks whether it's fixable?"
"Walk me through the mechanics of your platform consolidation work. Not the vision — how you decided what stayed, what got cut, and how you handled the product teams whose baby just got killed."
"You've lost people during long platform work. Be honest about that."
"Give me a situation where telling an exec the truth cost you something."
"What's the actual gap between where you are and where you need to be to do the VP job well?"

That last one is the one I want to focus on.

The answer that surprised me

The AI — running as me, on my own skills file — gave this answer to the gap question:

"The gap is board-level storytelling. I can tell the story in the room — to a CEO, to a CPO, to a peer exec. What I haven't done is own the design narrative at the board level. Quarterly business reviews, investor updates, the moments where design has to justify its existence in the language of a capital allocation decision. I've been in those rooms. I haven't been the one accountable for what comes out of them. That's real exposure I don't have yet. And I know it."

I didn't write that answer. I wrote the inputs — my experience, my proof of work, my constraints. The AI synthesized it and named a gap I'd been circling but hadn't articulated that cleanly.

That's not a parlor trick. That's a signal.

What the scorecard looked like

After the interview I asked the AI to score the session as the interviewer.

Interview scorecard
8.2 / 10
Would advance to next round.
Strategic thinking
9/10
Self-awareness
9/10
Business fluency
8/10
Executive presence
8/10
Authenticity
8/10
Team leadership
7/10
"The strategic and structural answers were sharp. The people answers were one story short of being as sharp."

That's accurate. I know it's accurate. And I now know exactly where to focus before my next real interview.

What hiring could actually look like

Here's the version of this that's coming whether we're ready or not.

A candidate submits a skills file instead of — or alongside — a resume. The employer's AI ingests it and runs a structured interview session. The transcript is the deliverable. The hiring manager reviews the transcript, not the resume.

What this surfaces that a resume can't:

The objection I'd raise myself

You could game this. You could write a skills file that's pure performance — aspirational rather than accurate — and the AI would produce an impressive transcript that doesn't reflect how you'd actually operate.

True. But that's also true of every resume ever written.

The difference is that a skills file-powered interview can be stress-tested in real time. The AI can push on the gaps. It can ask for specifics. It can notice when answers stay abstract when they should be concrete. A static resume can't do any of that.

The other objection: this removes the human element from hiring. Also true — at the screening layer. But the human element in screening is mostly theater anyway. We're pattern-matching against proxies. A transcript of how someone actually thinks is more human signal, not less.

What I'm doing with this

The skills file is live on my portfolio site. The system prompt version is something I'm offering directly to hiring managers and search firms — drop it in, run the interview, see what comes back.

This isn't a gimmick. It's a bet that the next generation of hiring doesn't look like a PDF and a phone call.

The resume tells you where I've been. The skills file tells you how to deploy me. The transcript tells you whether I'm actually what I say I am.

That's the whole job.