The Problem

Eve spent a summer doing manual QA with spreadsheets. She knew exactly what she was trying to escape.

Before she was a product designer at Y Combinator, Eve did manual QA work the old way, flows checked by hand, every result tracked in a massive spreadsheet. She describes it simply: "It was dreadful."

So when she joined YC's applications team and found herself inheriting a version of that same process, she understood the problem personally. Every batch, before the deadline, the team would host a bug bash, pulling in engineers from other teams across YC, asking them to create fake applications, and hoping they'd stumble across anything broken. It was informal, incomplete, and entirely dependent on how many people showed up.

"It's probably one of the most mission-critical pieces of software we have at YC. All the companies that we accept for a given batch start with an application."

The stakes make this harder than it sounds. The application form has dozens of conditional sections, different questions depending on your company type, your co-founder count, your answers to earlier sections. That branching logic generates hundreds of unique applicant paths. With just two engineers maintaining the portal and four batches a year, there was no realistic way to cover all of it manually before every deadline.

And there was a particular irony to all of this. YC's core advice to every company they fund is to ship as fast as you can. They weren't following it themselves, because QA was the bottleneck.

The Solution

The integration took two weeks. Coverage went from wherever it was to 90% of the application flow.

Eve describes it as: seamless. Within the first two weeks of onboarding, Spur had covered more than 90% of the application flow. No lengthy setup, engineering sprint or rewriting existing tests.

The workflow that made it click for Eve was record-to-test. Instead of writing test scripts herself, she records a video of the flow end to end, uploads it, and Spur's agent creates the test automatically, generating every single step she went through in the video. No code and no scripting knowledge required.

"Instead of writing the test myself, I can just record a video of the flow end to end and then upload that video and the agent will create the test for me. It's just going to auto-generate every single step."

Once the tests were running, Spur launched thousands of agents each batch, simulating real founders navigating every conditional branch, field state,  edge case a human tester might not think to try. And when something broke, Eve found out immediately via Slack, not from a support ticket.

The Crucial Moment

It’s 9pm on a Sunday in peak season and a critical bug hits. Fixed before anyone noticed.

This is the story Eve keeps coming back to. It was a Sunday night, around 9pm, in the final stretch before a batch deadline, the single highest-traffic window of the entire year, when thousands of founders submit in a single day.

Spur flagged a critical bug in the application flow. The team knew about it immediately. They fixed it that night. By the time the submission surge arrived, the portal was clean.

Without Spur, that bug would have surfaced the way bugs always used to surface, when someone encountered it. At 9pm on application Sunday, that someone would have been a founder trying to submit.

"It's really day and night now that at YC we're using Spur. It is so much easier. I never thought that testing could be made as simple as that."