What we learned from AI/ML YC applications

“You have Fortune 500 pilots, not Fortune 500 revenue, and if OpenAI builds onboarding primitives next quarter, your wedge evaporates.”

8.8/10·Sharp team, exceptional velocity, real enterprise pull, but you need one pilot to convert to six-figure ARR in 60 days or the story stalls.

“Two active users and zero revenue after eight months is not traction, it is friends and family testing at best.”

6.2/10·Strong builders with a smart insight, but pre-traction and no GTM plan in a PowerPoint-dominated world.

“V1 broke on real data, you have zero customers outside your dad, and a $100K to $700K ACV range means you have never actually priced this.”

4.9/10·Legitimately smart technical builder, but 8 weeks too early, come back when V2 works and someone not related to you will pay for it.

“The data flywheel only works if optimization patterns generalize across architectures, one tf32 example on Qwen doesn't prove this isn't just custom tuning disguised as software.”

8.5/10·Sharp technical team with credible early proof, but needs to show the database scales across 10-20 model families before May 16th, or this becomes a linear services play.

“Zero revenue and unproven GTM after pivoting just one month ago, needs paid conversion fast.”

7.8/10·Strong team with real customer pull, but must convert an LOI to paid in 30 days to prove this works.

“Built a working prototype in one week but has zero users, zero revenue, and no proof anyone will pay $20/month when ChatGPT is free.”

5.8/10·Sharp insight and fast execution, but solo founder with no users yet in a space where OpenAI could crush you overnight.