Key Takeaways
- The AI economy stabilized in 2025, enabling a clearer playbook for AI-native companies.
- Anthropic surpassed OpenAI as the preferred LLM for Y Combinator applicants, signaling a market shift.
- The commoditization of AI models is shifting opportunities towards the application layer for startups.
- Despite AI advancements, rising customer expectations continue to drive significant hiring needs in startups.
Deep Dive
- In 2025, Anthropic achieved 52% adoption among YC applicants, surpassing OpenAI as the preferred LLM.
- Anthropic's rise is linked to its strong performance in coding and agent tasks.
- Gemini is also gaining traction, reaching 23% adoption due to its reasoning capabilities and real-time information access.
- Startups are arbitraging multiple AI models, with Series B companies building orchestration layers for dynamic swapping.
- One strategy involves using Gemini for context engineering and OpenAI for execution, based on specific task evaluations.
- The trend suggests a commoditization of AI models, creating significant opportunities at the application layer.
- Large tech companies are making significant capital expenditures on AI infrastructure, resembling an 'installation' phase.
- This buildout faces challenges including a lack of power generation and suitable land for data centers.
- Regulatory hurdles and environmental concerns in regions like California are hindering terrestrial development.
- There is increased interest in developing AI model companies, including specialized models for edge devices or specific languages.
- A YC startup achieved better healthcare AI benchmarks than OpenAI using an 8 billion parameter model with a curated dataset.
- The availability of open-source models and fine-tuning capabilities enables domain-specific AI to potentially outperform larger general models.
- The 'vibe coding' category is growing, with companies like Replit and Emergence noted for their advancements.
- Google's 'Anti-Gravity' demo highlights potential in AI video generation tools.
- Skepticism remains regarding the current usability of AI for shipping 100% production code.
- Companies initially achieved significant ARR, such as $1 million, with minimal hiring, often just founders.
- Post-Series A, these companies began hiring actual teams, indicating a shift from founder-led growth.
- Gamma achieved $100 million ARR with only 50 employees, demonstrating high revenue per employee in the AI space.