Y Combinator Calls It: AI Applications Feel Like Lying

A founder in Valletta spends three evenings crafting their Y Combinator application. They know their market. They've lived the problem. They have numbers that don't lie. Then they submit — alongside ten thousand other applications, a growing number of which were written in forty minutes by a language model that has read every successful YC application ever published.

The partners at Y Combinator can tell. They've said so directly: it feels like being lied to.

This is the uncomfortable tension sitting at the heart of the startup world right now. Y Combinator — the accelerator that launched Reddit, DoorDash and Airbnb and has backed more than 5,000 companies since 2005 — is not anti-AI. These are people who fund AI companies for breakfast. What they're flagging is something more specific: the polished absence of a real person. Applications that answer every question correctly and reveal nothing true.

Here is the mechanism worth understanding. Investors at the seed stage are not primarily buying a product. They are buying a person's ability to navigate uncertainty — to make decisions in conditions where the map doesn't exist yet. An application that reads like a template tells them nothing about that capacity. It tells them the founder is good at prompting. That is not the same skill.

The harder lesson is this: AI is now a filter, not an advantage. Every mediocre application has been upgraded. Which means differentiation has moved entirely to the things a model cannot replicate — specific customer conversations, uncomfortable market truths, the precise reason this particular founder cannot leave this problem alone. Specificity is now the only currency that clears the bar.

My call: founders who use AI to polish genuine thinking will be fine. Founders who use AI to substitute for genuine thinking will be screened out faster than ever, because the gap between fluent and true has never been more visible to people who read applications for a living. The condition under which I'm wrong: if volume overwhelms human review entirely and algorithms replace partner eyes. That day may come. It hasn't yet.

For anyone in Malta building something and considering an international accelerator application — the advice is the same as it has always been. Know why you, why this, why now. Write that. Then, if you want, ask the machine to clean the grammar.

The application that gets funded is the one that could only have been written by you.