CodeSignal.

Who they are, right now.

A candidate in Lagos, a recruiter in Manhattan, and a hiring manager in Bengaluru are all looking at the same screen at the same time. The screen is a CodeSignal assessment - 70 minutes, four problems, a timer that does not blink. By the end of the hour the candidate will have produced a Coding Score, a three-digit number that travels through Greenhouse, Workday and a dozen ATSs to a recruiter who has never met them. That is CodeSignal in 2026: not a recruiting tool, exactly, but a referee.

The company sits inside a tower at 201 California Street, in the part of San Francisco that still believes in elevators and badge readers. There are roughly 140 employees scattered across time zones - it has been remote-first since long before that meant anything - and an AI tutor named Cosmo doing roughly the work of a few dozen more. Cosmo is the company's mascot, its product, and, increasingly, its growth lever. He looks like a tiny astronaut. He grades how you talk to your coworkers.

The problem they saw.

Hiring engineers, for most of the last decade, has been a parlor game dressed up as a science. A resume gets you a phone screen. A phone screen gets you a whiteboard. A whiteboard gets you, on a good day, a job; on a bad day, a story you tell at parties. Each step filters more for vibe than for skill. Studies have suggested that traditional interviews predict performance about as well as a coin toss with anxiety.

The founders found this absurd, and a little personal. Tigran Sloyan and Aram Shatakhtsyan grew up in 1990s Armenia, a country with rolling blackouts and a remarkable density of math olympiad medalists. They met as teenagers at international competitions, the kind where you solve combinatorics problems while jet-lagged. Tigran left for MIT. Aram stayed in Yerevan for informatics. Neither of them had a Silicon Valley resume. Both of them could code.

When they started talking about hiring, the question was simple: why does the world prefer to read about a developer when it could just watch one work? The answer, mostly, was inertia. Resumes scale. Watching people code does not. Someone needed to make it scale.

The founders' bet.

In 2015 the two of them, plus co-founder Felix Desroches, launched a thing called CodeFights. It was, frankly, a game. Developers logged on for head-to-head coding duels, won points, climbed leaderboards. It looked like an arcade and it ran like one. But under the hood, every keystroke was being measured. Every problem was being normalized for difficulty. CodeFights was a Trojan horse and the horse was full of regression analysis.

By 2018 the company had quietly rebranded to CodeSignal and the leaderboards had become Coding Scores. The scale was 300 to 850 - the deliberate mimicry of FICO was the joke and also the point. If credit could be a number, why not coding? Enterprises agreed. The product walked from gamers to recruiters in about eighteen months, which is a faster pivot than most companies survive.

The product.

CodeSignal today is two products that look like one. The first, Hire, is the assessment engine that the Fortune 500 has quietly standardized on: pre-screens, technical screens, certified evaluations, an interview IDE that supports more languages than most universities teach. It is the part that grades the candidate.

The second, Learn, is the part that teaches them. Launched in February 2024, it is an AI-tutored skills platform where lessons are conversational, feedback is instant, and a small cartoon astronaut named Cosmo will, if you ask, explain pointers as if you were five. In August 2025 Cosmo became its own iOS app: more than three hundred courses on generative AI, coding, marketing, finance and leadership, all designed for the cracks between meetings.

Then there is Conversation Practice, launched in 2024 and expanded in 2025 with full voice AI. It runs you through workplace conversations - delivering feedback, defusing a conflict, listening for the actual thing your report is asking about - and grades how you do. The early users were sales managers and engineering directors. The early surprise was how many of them needed the practice.

The proof.

The customer list reads like a software conference badge wall. Meta. Netflix. Capital One. Robinhood. Zoom. Brex. Databricks. Instacart. Upwork. Dropbox. Roughly speaking, if you have submitted a coding challenge to a tech company in the last three years, there is a non-trivial chance CodeSignal saw your screen first.

The funding has tracked the customers. A 2015 seed round from Felicis Ventures. A 2017 Series A from Menlo. A pandemic-era Series B. Then, in September 2021, a $50 million Series C led by Index Ventures - a round that took the total to $87.5 million and put the company on the kind of valuation trajectory that doesn't get printed in press releases. They have not raised since, which in the current market is less a confession than a flex.

There are integrations - Greenhouse, Workday - and there is a research arm, which publishes the University Coding Score Rankings, the closest thing the industry has to a sortable transcript of computer science programs. The rankings cause arguments. That, presumably, is the point.

The mission.

Officially the mission is to "discover and develop the skills that will shape the future." Unofficially - and the unofficial version is the one Tigran says in interviews - it is to replace credentialism with measurement. To make pedigree count for less and signal count for more. To do for hiring what the SAT tried and mostly failed to do: surface talent that the existing system was structurally missing.

It is not a small claim, and CodeSignal is not the only company making it. HackerRank, Karat, Codility and CoderPad all sit in the same room, and Pluralsight and DataCamp share a wall with Learn. What separates CodeSignal, in the company's telling, is that it built its assessment science before it built its AI - meaning the simulations are calibrated by humans and graded by models, not the other way around. Critics will say that distinction matters less than CodeSignal thinks. Customers, so far, disagree by signing checks.

Why it matters tomorrow.

The next decade of work will not be kind to the resume. Generative AI is collapsing the half-life of skills - the language you learn in 2026 may be deprecated by 2028 - and companies are already moving from hiring credentials to hiring capabilities. CodeSignal's bet is that capability needs a measurement layer, and that the measurement layer needs to teach as well as test. Hire and Learn, in other words, are not two products. They are the same product, on either side of an HR transaction.

Whether Cosmo turns into the Duolingo of job skills is an open question. Mobile learning has chewed up bigger names. But CodeSignal has a quietly enviable thing on its side: it already runs the world's most-watched technical assessments, which means it knows, with rare precision, what skills companies actually pay for. That is a dataset most ed-tech companies would lie about having.

Back to the screen.

The candidate in Lagos finishes the last problem with thirty-two seconds on the clock. The recruiter in Manhattan opens an email and sees the Coding Score before she sees the name. The hiring manager in Bengaluru notices that the candidate paused twice on the third problem, asked Cosmo a clarifying question, and recovered. None of these three people have met. None of them ever will, until the loop closes with a Zoom and a yes.

What CodeSignal sells, in the end, is a small, specific kind of fairness: the chance that the person on the other side of the screen will be evaluated on what they can actually do. It is a narrow promise. It is also, demonstrably, a billion-dollar one.