The more machines can do, the more you matter.

My friends say I push them too hard.

For as long as I can remember, I’ve been the one asking the uncomfortable question: does your work feel like play to you? And if not, what would? I’ve spent countless evenings helping friends hunt for that answer, and once they found it, helping them chase the jobs where it lived.

That second part is where things kept breaking. The work was right, the person was ready, and the process still failed them. I watched offers go to whoever performed best in an interview instead of whoever was best at the work.

Then I spent three years at LinkedIn, watching up close how people look for jobs and how companies decide whom to hire. Both sides lean on thin evidence: a resume that lists what you did, a rehearsed answer in an interview. Good people get missed constantly, and neither side can see it happening.

Two problems, one root. Most people never discover the work they’d love, because trying things is expensive. And the people who do know are invisible to the employers looking for exactly them.

In the last few years, something genuinely new happened: intelligence became a commodity. Give AI almost any task, from code to analysis to writing to design, and competent output comes back in minutes, at a cost falling toward the price of electricity.

This is not a story about tools. When execution becomes abundant, everything downstream of it changes: what work is, what skill means, and what could possibly count as proof that a person is good.

The old proofs broke first, because every proof we had was an artifact. A document, a project, a confident answer. Artifacts are exactly what AI makes cheap.

Every hiring signal we use was honest when it started.

The resume began as a verifiable record. When writing one took effort and lying was risky, it meant something. Then it became a genre to optimize, then keyword bait for screening software, and now AI writes hundreds of them a minute. The portfolio proved you could build, until a polished project became a weekend of prompting. The interview tried to test thinking in real time, but the questions converged on a standard set, candidates studied the set, and it began rewarding rehearsal over ability.

AI didn’t start this decay. It finished it.

A polished artifact now proves exactly one thing: that something, somewhere, was able to produce it.

Not that you understood the problem. Not that you’d make the right call under pressure. Not that you should be trusted with what ships.

Both sides did the rational thing, and the system broke. Candidates apply with AI, and recruiters now see three hundred applications per role, triple what it was a few years ago. Employers screen with AI, and seventy percent of hiring managers trust it to decide faster, while only eight percent of candidates believe it decides fairly. Everyone escalates. Nobody is seen. Hiring retreats to the oldest signal there is: who you already know. What remains is a lottery with a guest list.

When intelligence is cheap, value moves to what machines cannot supply. I think there are three things.

1. 01 / 03

   Judgment.

   When you can point intelligence in any direction, the direction is the work. What do you choose to build, and why? Do you reason from first principles? Do you have depth and taste in your domain, the kind that notices what’s missing and not just what’s wrong? Give a hundred people the same tools and the outcomes now diverge wildly. Judgment is the variable.

2. 02 / 03

   Accountability.

   AI output is probabilistic. It can be brilliant and wrong in the same paragraph. Someone has to verify it, and someone has to sign. You sign off on AI’s work; AI will never sign off on yours. A machine can assist a decision. It cannot take responsibility for the outcome.

3. 03 / 03

   Agency.

   The willingness to act was always underrated. Now it’s multiplied: one person who moves, who asks, builds, tests, and ships, carries leverage that used to require a team. The gap between people who act and people who wait has never compounded faster.

One honest question: judgment has always been built by doing junior work under supervision. If AI absorbs the junior work, where does the next generation’s judgment come from? It has to be built somewhere safe to try, to fail, and to be seen.

Imagine finding your work the way you actually learn anything: by doing it. Step into a role for ninety minutes. Feel its problems, its tradeoffs, its rhythm, and know rather than guess whether it lights you up. Then do it again, and get visibly better.

Imagine your proof of ability is work someone witnessed: real decisions, under real constraints, owned by you and carried from employer to employer.

Imagine companies coming to you, because they saw how you work and want exactly that. And when one passes, you don’t get silence. You learn why, and what to work on next.

Work you love, found by doing. Ability, proven by being seen.

We’re starting where the collapse is sharpest: hiring senior engineers.

Caliber is a flight simulator for real work. A candidate steps into a fictional company with a running product, an existing codebase, and real documentation, and takes on an ambiguous problem, the kind that never fits in an interview. Ninety minutes. AI agents at their command. A release decision at the end.

We watch the whole loop. Did they understand the problem before solving it? Plan around the biggest risks? Direct AI, and catch it when it was confidently wrong? Keep checking after everything looked done? Did they know what was safe to ship, and say honestly what wasn’t?

The employer gets a report in which every conclusion cites the moment it happened. Trained humans judge the decisions. Software makes sure nothing is lost or distorted.

Today every company runs its own gauntlet, and candidates repeat themselves endlessly. That’s the part we intend to delete.

One deep, role-specific simulation, taken once, owned by you. Employers don’t see a score; they apply their own rubric to your observed work, because what Google needs from a senior engineer is not what a ten-person startup needs. You might be loved by one and passed over by the other. That tells you something real, and Caliber shows you your weak spots and how to work on them.

We call it a Demo Day for talent: everyone presents the same kind of evidence, and companies approach the people whose strengths match what they need.

The resume told employers what you did. Witnessed work shows them how you decide.

We don’t yet know the perfect way to measure judgment. Nobody does. So Caliber starts with controlled situations, evidence with citations, and human review, and will not produce an automated score until real hiring outcomes prove it deserves one. We would rather show the moments and let people judge than hide behind a confident number.

Witnessed work also won’t be the only signal that matters. Trust built over time, the people who vouch for you, the consistency you show across years: these will matter more, not less. We’re building one signal that is honest, and building it carefully.

If you’re an engineer who is tired of being invisible on paper: come get seen doing your best work. If you’re a company that wants evidence instead of vibes: come design a pilot with us. And if this is the problem you want to spend a decade on, so do we.