Less Competition Lives in the Non-Obvious Problems

01 /The Concept

After surveying more than 500 YC founders, Outship confirmed what many suspected: the majority are not just using AI coding agents – they're using them aggressively. Which means they also need to hire engineers who can do the same.

YC made this explicit in February, adding a new item to its application: a recording of a founder's AI coding session – specifically one they're proud of.

As YC's president explained, the point isn't to evaluate prompting skill. It's to assess whether founders have developed a systematic approach to getting AI to build what their users actually need.

This kind of evaluation won't stay unique to YC. Any company hiring engineers will soon want to test the same thing. But how?

That's exactly what Outship is building – a platform for assessing how well engineering candidates work with AI tools.

The platform works in interviews or as a take-home exercise. A company uploads a code repository, writes a task description, and sends the candidate a link to a personal workspace.

The workspace is a virtual machine with the source code pre-loaded and AI tools pre-configured – ready to use from a browser, or connectable to a local development environment.

The critical feature is that Outship captures not just the result, but the process. Every prompt and every manual edit is recorded. The platform analyzes how the candidate approached the problem and how effectively they used AI to solve it – because that's the actual thing being evaluated.

Outship went through YC's Winter 2025 batch, but only published its platform listing on the YC site yesterday.

02 /Why It Matters

The natural assumption when AI coding tools appeared was that demand for human engineers would drop sharply. That hasn't happened. If anything, it's growing. Why?

There's a "10x productivity trap" at work. Because AI makes it easier and faster to write software, far more software gets written – both user-facing and internal. But every piece of software still needs human oversight, which means more engineers, not fewer.

Then there's the technical debt problem. AI generates code quickly, but that code often isn't production-ready – it can be insecure, hard to scale, or poorly structured. Companies need engineers to refactor it. And the more code AI generates, the more cleanup is required.

At the same time, the profile of the engineer that companies actually need has shifted. The old model relied heavily on "coders" – people who translated specs from senior engineers into working code. AI can now do most of that. What companies need instead are engineers who can supervise and refactor AI-generated output, working at a senior level or above.

And to direct AI effectively – telling it what to build and verifying that the output fits the broader system – an engineer needs to think at the architecture level. They need to understand how each piece of generated code fits into the larger structure before it's ever written.

In short, basic coders aren't needed anymore. Hiring based on a resume that lists programming languages no longer makes sense.

What companies need to evaluate is: can this candidate work as a senior engineer or systems architect, and can they use AI as a force multiplier for that work? That's what Outship helps assess.

The same dynamic is playing out beyond software. Humans can't compete with AI at AI's own game – trying to match its speed and volume is a losing strategy. The real play is learning to direct AI well enough to push past its limitations. Or to use AI to push past your own.

Startup Hupside ([related review](/review/kak-ne-stat-odnoj-iz-95-neudach)) calls this "original intelligence" – the capacity that allows someone to collaborate with AI at maximum effectiveness, rather than just running prompts. Not everyone has it. Hupside's platform measures it, helping companies identify who on their team can thrive in an AI-native workflow. The company raised $1.7M in seed funding last September.

03 /Opportunities

AI is landing two distinct challenges on every company simultaneously.

The obvious one: automate business processes using the new generation of AI tools. There are already hundreds of platforms built for this.

The less obvious one – and where far fewer solutions exist – is rethinking hiring criteria and evaluation processes entirely. Companies need to identify and hire people who can use AI effectively, not just people who can code. The competencies required are different, and the ways to assess them are completely different too.

This is an open and growing market: platforms for evaluating how well candidates perform with AI – role by role.

The "role by role" part matters. Assessment for engineers looks completely different from assessment for analysts, marketers, or operations people. Each requires a specialized platform. Which type of role could you build an evaluation platform for?