Y Combinator38 min
2024: The Year the GPT Wrapper Myth Proved Wrong
Jared Harge, Diana, Harj, Lily Yang, Cheng Cheng, Suk Peng, Yiu, Francesc Campoy Flores, Priyanka Vergadia, Anan, Mark Mandelbaum, Melanie Warrick, Mark Mirchandani, Gary Miles, Leslie Kendrick Magnuson, Harjit
Webinar, Interview, Fireside Chat
- Startups can generate tens of millions of dollars within 24 months starting from zero with only $2 million to $5 million in capital.
- Market consensus has shifted away from value accruing solely to foundation model companies, with non-OpenAI applications like Perplexity and Harvey emerging as breakout successes.
- Open-source models such as Llama are expected to prevent monopoly pricing by allowing competitors access to the same foundational technology.
- Competitive advantage is increasingly defined by product quality, sales execution, user feedback loops, and zero churn rather than model ownership.
- Model routers are establishing themselves as viable entry points for new stacks, evolving into multi-model orchestration architectures that combine fast models for simple tasks with larger models for complex ones.
- Companies like Camphor, Cursor, and Variant are optimizing performance and aesthetics through specific model routing, codebase-wide understanding, and post-training workflows on open-source models.
- Enterprise proof-of-concepts from 2023 have translated into real revenue in 2024, with YC batch startups reaching $1 million ARR faster than previous years.
- Summer and fall batches achieved a 10% weekly aggregate growth rate, representing 3x growth over the course of YC, a rate previously unseen on average.
- The timeframe for startups to reach $100 million in annual revenue is trending downward, while the number of potential $100 billion companies is projected to increase 10x every decade over the last 20 years, potentially reaching 1,500 per year.
- Vertical AI is expected to enable over 1,500 companies to thrive by bypassing traditional enterprise sales cycle constraints through strong, domain-specific value propositions.
- AI agents have matured to perform thousands of tickets daily with reliability suitable for enterprise deployment, shifting the industry focus from chat-like interactions to agentic, multi-step tasks.
- Regulatory uncertainty remains regarding future restrictions or registration requirements for advanced AI models, though recent regulatory outcomes have favored startups.
- Risks exist that AI platforms could evolve into Win32-like monopolies where owners access all stats and launch competing applications.
- The AI sector remains highly volatile with rapid changes making any relief in 2024 potentially premature.
- Success stories like Scale AI demonstrate the potential for pivoting data labeling strategies from self-driving cars to RLHF for LLMs to build companies valued at $100 billion.
- Founders who delay finding a specific idea until a batch to explore future opportunities, such as vertical AI back offices, have found significant success.
- Voice AI is anticipated to support hundreds of tailored vertical applications across sectors like language learning and banking rather than a single winner-take-all market.
- Robotics startups are increasing, with specific hardware targets set for 2025 shipments, though the field remains early and hardware costs are high.
- AR hardware faces significant constraints, with the Apple Vision Pro facing challenges in identifying a killer application beyond media consumption.
- The explosion of AI coding tools in 2024 allows individual productivity to rise, shifting hiring needs toward engineers capable of evaluating AI output.
- Programming interviews are projected to shift toward measuring absolute output via AI tools, raising the bar for productivity and changing traditional whiteboarding methods.
- Amazon's internal LLM applications for back-office processes may be released to the public, potentially creating new startup stacks analogous to AWS.
- In-person events are returning with a ratio of approximately 10 investors per company at YC demo day, signaling a "homecoming" for Silicon Valley.
- The "remote forever" era is ending, with late-stage startups prioritizing employee return to the office.
- San Francisco's recovery is expected to take 10 to 20 years, though recent elections indicate a positive trajectory necessary to maintain the city as a beacon for top talent.