Y CombinatorWhy Old Startup Ideas Like Recruiting Now Work With LLMs
Through LLM code evals and gross margin thinking about unit economics; Triplebyte took years to build what AI does in weeks, making recruiting startups viable.
At a glance
WHAT IT’S REALLY ABOUT
AI Unleashes Previously Impossible Startups, From Recruiting To Full-Stack Services
- The hosts explore startup ideas that only became viable with modern LLMs, emphasizing how rapidly improving AI radically reshapes the “idea maze” for founders.
- They highlight recruiting, education, and legal services as prime examples where AI now performs core evaluation and knowledge work that previously required large ops teams and labeled datasets.
- The conversation contrasts old advice about lean validation with a new era where simply working at the frontier of AI often causes you to “bump into” powerful ideas.
- They also discuss moats, platform dynamics, and infra/tooling gaps, arguing that many incumbents and unicorns have barely started to adapt, leaving significant whitespace for new founders.
IDEAS WORTH REMEMBERING
5 ideasPreviously failed categories can now work if AI changes the core constraint.
Recruiting marketplaces and full-stack services struggled due to weak automation and low gross margins; with LLM-based evaluation and agents, you can launch what used to require years of labeled data and large human ops from day one.
AI unlocks much deeper personalization in education, enabling new business models.
True personalized tutors and adaptive tools (e.g., RevisionDojo, Speak, Studyy, Edexia) can approach or match human tutor quality, meaning parents and schools will pay far more than they did for generic learning apps.
As intelligence gets cheaper, freemium consumer AI at massive scale becomes viable.
Model distillation and better hardware are pushing usage costs toward pennies per user; once that threshold is crossed, products can be given away to hundreds of millions and monetized via a paying minority, as OpenAI and Perplexity already do.
Moats will come from brand, UX, integration, and switching costs—not just models.
Despite Gemini’s technical strength, ChatGPT dominates mindshare; similarly, clumsy integrations from Google and Meta show that having great models is not enough—winning requires coherent product design and deep, useful integrations with user data and workflows.
Full-stack and tech-enabled services can now have software-like margins thanks to agents.
Legal, recruiting, and other knowledge-heavy services (e.g., Legora) can automate the bulk of the work with LLM agents, turning what were once people-heavy, low-margin ops businesses into scalable, high-margin software-like companies.
WORDS WORTH SAVING
5 quotesThe idea maze just moved; all of the walls to the idea maze have shifted around.
— Garry Tan
Now actually full stack companies can look like software companies under the hood for the first time.
— Jared (YC partner)
If you're living at the edge of the future and you're exploring the latest technology, you're very likely to just bump into a great startup idea.
— Garry Tan
Triplebyte 2.0s won’t have to hire this huge ops team and have bad gross margins. They'll just have agents that do all the work.
— Jared (YC partner)
My main takeaway from this has been there's never a better time to build.
— Garry Tan
High quality AI-generated summary created from speaker-labeled transcript.
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome