Daksh Gupta, Soohoon Choi and Vaishant Kameswaran entered Georgia Tech’s CREATE-X startup program with an idea called Tabnam: an AI shopping assistant. It’s not the company any of them run today. Through the program’s customer-discovery process — the unglamorous, repetitive work of actually talking to potential users before building more product — the team found that the technology they’d built had far stronger traction applied to a completely different problem: helping engineering teams understand and review large, unfamiliar codebases. They rebuilt around that instead, and called it Greptile.
The pivot happened because they listened to a problem they weren’t looking for
Startup pivots get told, after the fact, as a moment of insight. What actually happened at Greptile was less dramatic and more disciplined: customer conversations that were supposed to validate the shopping assistant kept surfacing a different, more specific pain point in software engineering — the time and context lost when a developer has to understand code they didn’t write, at a scale no single engineer can hold in their head. Recognizing that the signal in those conversations mattered more than the plan they’d walked in with, and being willing to set aside a product they’d already built to build a different one, is the actual skill underneath the pivot — a considerably harder discipline than the idea itself.
From a class project to a $180 million valuation
Greptile joined Y Combinator’s Winter 2024 cohort, and the traction that followed was fast even by startup standards: the company raised a $25 million Series A led by Benchmark, bringing its total funding to $30 million and valuing the AI code-review company at $180 million — a striking outcome for a team that started as computer science undergraduates working through a university startup accelerator with an entirely different product.
The lesson underneath Greptile’s story isn’t ‘have a good idea.’ It’s ‘be willing to notice that your good idea is the wrong one, using evidence you weren’t looking for, and rebuild around what the evidence actually says’ — a considerably less comfortable skill than conviction, and a considerably more useful one.
Why this matters beyond one company’s funding round
Gupta and Choi graduated with computer science degrees in 2023, meaning the distance between undergraduate coursework and a $180 million company was measured in months, not years — a timeline that says less about overnight luck and more about a specific, repeatable process: build something, talk to real users honestly, and treat a pivot as a discipline rather than a failure. It’s a considerably more instructive story for other student founders than the shorthand version — ‘AI startup raises funding’ — usually conveys.