Stop Building Portfolios. Start Managing Repos.

Why commit history is the only resume that matters in the AI age

I’ve spent a decade in the engine room of talent marketplaces. From 2012 to 2016, I sold Careers 2.0 at Stack Overflow to hiring managers obsessed with finding “10x developers.” Later, at Bolster, I helped businesses hire from a network of thousands of fractional executives.

I’ve seen a zillion ways these marketplaces fail, but today there is a new, lethal variable: Velocity. On paper, “candidate matching” is easy. In practice, it’s a graveyard of good intentions because the half-life of a skill has collapsed.

This creates a “Grim Market” paradox: Everyone needs AI skills, but nobody can get hired because the “skills” are moving faster than the “job descriptions.”

As a recent New York Times report noted, young people aren't just facing a tough job market; they are facing a world where the “Degree Moat” is evaporating.

“For those who were employed, more than 40 percent held jobs that do not typically require college degrees, the highest level since 2020.”

I’ve been a huge proponent of project-based learning for the past decade and have launched and spearheaded over a dozen initiatives with high school and college students that center around helping kids get portfolio projects.

It used to be good enough to show a “point in time” artifact as evidence that you know your stuff in any singular domain. But I’m afraid now that point in time artifacts aren’t good enough.

At the rate things are moving, the project where you knocked it out of the part six months ago almost doesn’t matter. What I care more about is how quickly you can adapt and pick up the next thing.

I’ve learned the value of this from a very personal experience.

Last year, I spent 6 months exclusively building MuseKat, a mobile app that helped kids experience art at museums through customized audio guides told by Miko the Meerkat. This January, I led a classroom of entry-level CS students at UNC Charlotte through an AI no-code builder challenge to build an app that solved the same problem. What happened was sobering:

While it’s still marginal interesting for me to share about my own experiences in successfully deploying an app or two to iOS and Android devices, it’s not as useful as me being able to demonstrate that I’m continually using the latest AI tools to consistently solve my own problems.

In that way, the apps I built last year don’t matter at all. But my current stack sure as hell does.

Today I run Build First, my AI learning lab, through about several key repos.

• There’s my marketing site (buildfirst.ai) which I built first (last summer, with Replit)

• There’s my web app (buildfirst.app), which is a custom tool built in Claude Code that I reskin for each new corporate training as a way to teach non-technical people how to think like engineers

• There’s my personal operating system, which is a permissioned instance of my GTM dashboard that runs through Claude Code and a series of interconnected calls to things I care about (specifically: my emails, my call transcripts, and my calendar)

Each new problem that I come across starts with a fresh repo. I start with a new problem, I ask my AI, “What are the best tools today for me to solve that problem?” And then I figure out how to build that thing.

What I notice in looking back on my own building arc is the increasing complexity and modularity of how I build. The time stamps of the initial commits in each codebase reflect back clearly to me about how I was thinking based on the “in vogue” AI builder’s mentality at the time.

In other words, I’m able to see my own cognition develop and mature.

This is now the skill I look for when I speak with potential human collaborators.

I don’t care where you worked before. I don’t care where you went to school. I don’t care what you built once, a very long time ago. I care about what you are building today, as it relates to what you built yesterday, and how you are thinking about what to build next.

Stop sending resumes; start sharing repos. If you’re a modern orchestrator whose velocity has outpaced your job description, let’s build something together. Drop me a note at bethany@buildfirst.ai.

I’ve spent a decade in the engine room of talent marketplaces. From 2012 to 2016, I sold Careers 2.0 at Stack Overflow to hiring managers obsessed with finding “10x developers.” Later, at Bolster, I helped businesses hire from a network of thousands of fractional executives.

I’ve seen a zillion ways these marketplaces fail, but today there is a new, lethal variable: Velocity. On paper, “candidate matching” is easy. In practice, it’s a graveyard of good intentions because the half-life of a skill has collapsed.

This creates a “Grim Market” paradox: Everyone needs AI skills, but nobody can get hired because the “skills” are moving faster than the “job descriptions.”

As a recent New York Times report noted, young people aren't just facing a tough job market; they are facing a world where the “Degree Moat” is evaporating.

“For those who were employed, more than 40 percent held jobs that do not typically require college degrees, the highest level since 2020.”

I’ve been a huge proponent of project-based learning for the past decade and have launched and spearheaded over a dozen initiatives with high school and college students that center around helping kids get portfolio projects.

It used to be good enough to show a “point in time” artifact as evidence that you know your stuff in any singular domain. But I’m afraid now that point in time artifacts aren’t good enough.

At the rate things are moving, the project where you knocked it out of the part six months ago almost doesn’t matter. What I care more about is how quickly you can adapt and pick up the next thing.

I’ve learned the value of this from a very personal experience.

Last year, I spent 6 months exclusively building MuseKat, a mobile app that helped kids experience art at museums through customized audio guides told by Miko the Meerkat. This January, I led a classroom of entry-level CS students at UNC Charlotte through an AI no-code builder challenge to build an app that solved the same problem. What happened was sobering:

While it’s still marginal interesting for me to share about my own experiences in successfully deploying an app or two to iOS and Android devices, it’s not as useful as me being able to demonstrate that I’m continually using the latest AI tools to consistently solve my own problems.

In that way, the apps I built last year don’t matter at all. But my current stack sure as hell does.

Today I run Build First, my AI learning lab, through about several key repos.

• There’s my marketing site (buildfirst.ai) which I built first (last summer, with Replit)

• There’s my web app (buildfirst.app), which is a custom tool built in Claude Code that I reskin for each new corporate training as a way to teach non-technical people how to think like engineers

• There’s my personal operating system, which is a permissioned instance of my GTM dashboard that runs through Claude Code and a series of interconnected calls to things I care about (specifically: my emails, my call transcripts, and my calendar)

Each new problem that I come across starts with a fresh repo. I start with a new problem, I ask my AI, “What are the best tools today for me to solve that problem?” And then I figure out how to build that thing.

What I notice in looking back on my own building arc is the increasing complexity and modularity of how I build. The time stamps of the initial commits in each codebase reflect back clearly to me about how I was thinking based on the “in vogue” AI builder’s mentality at the time.

In other words, I’m able to see my own cognition develop and mature.

This is now the skill I look for when I speak with potential human collaborators.

I don’t care where you worked before. I don’t care where you went to school. I don’t care what you built once, a very long time ago. I care about what you are building today, as it relates to what you built yesterday, and how you are thinking about what to build next.

Stop sending resumes; start sharing repos. If you’re a modern orchestrator whose velocity has outpaced your job description, let’s build something together. Drop me a note at bethany@buildfirst.ai.

>600 subscribers

>600 subscribers