1 pointby ilyaforfun8 hours ago2 comments
  • ilyaforfun8 hours ago
    I was playing around with gstack and thinking about what to build from scratch using the whole process.

    Since I plan to apply to YC as a solo founder, I decided to create something YC-related. That's how this YC interview simulator came about.

    So I generated 9 personas (YC partners, current and legacy), cloned their voices with eleven labs and tried to recreate the environment of an interview.

    Each partner has a unique style and questions, which adapt to what you answer. Also, there are different emotions you can see on partner's avatar based on how you answer.

    At the end, there is a scoring, feedback and some additional learning tools you might read. Also it remembers you previous sessions and you can see the progress.

    Free to use until my credits run out. No login.

    I found it helpful for the application itself too. Or just chatting about ideas with different partners.

    No affiliation to YC. Just a fun side project

  • Bondig8 hours ago
    How did you train the personas, areas of interest etc.. Interesting tho
    • ilyaforfun7 hours ago
      No training in the ML sense. Each partner's persona is built from public sources: their essays, podcast appearances, interviews, YC talks, and things they're known to emphasize. Pretty much prompt engineering based on deep ai research.

      The question bank is hand-curated from real YC interview questions that have been written about publicly over the years. The scoring rubric tries to reflect what those partners have said they look for.

      The main limitation is obvious: it's a simulation of their public persona, not the actual person.