How has AI changed the recruiting startup landscape?

Previously, recruiting startups like Triple Bite had to invest heavily in building manual evaluation systems. Now, AI like LLMs can handle code evaluation and other assessments from day one, making recruitment marketplaces like Meror more viable.

Can AI truly personalize education?

Yes, AI is enabling hyper-personalization in education, moving beyond generic learning platforms. Companies are building tools that act as personalized tutors, tailoring the learning journey to individual student needs, similar to how Duolingo aims to adapt language learning.

Why are full-stack startups making a comeback with AI?

In the past, full-stack startups struggled with operational complexity and low gross margins. Now, AI agents can automate many operational tasks, allowing these companies to function more like software businesses with better margins, potentially becoming major players in their fields.

What lessons can be learned from the MLOps boom?

The history of MLOps shows that the timing of technology is crucial. Companies building tools for technologies that weren't quite ready, like Replicate and Ollama, often struggle initially but can see explosive growth when the underlying technology matures and adoption takes off.

What is the new startup advice in the AI era?

The old advice of 'sell before you build' and extensive customer discovery is becoming outdated. In the AI era, it's more effective to follow your curiosity, explore interesting new technologies, and build based on what becomes possible, as great ideas are likely to emerge from this exploration.

Why are big tech companies like Google struggling to leverage their AI?

Large tech companies often suffer from internal organizational silos, competing teams (e.g., DeepMind vs. GCP for Gemini), and a tendency to protect existing revenue streams. This can lead to fragmented product experiences and slower adoption of new AI capabilities.

Key Moments

Startup Ideas You Can Now Build With AI

Y Combinator

Science & Technology4 min read41 min video

May 16, 2025|284,930 views|4,871|187

YC Y Combinator

Save to Pod

Key Moments

On this page

TL;DR

AI unlocks new startup opportunities, especially in recruiting, education, and full-stack services.

Key Insights

AI, particularly LLMs, enables previously unviable business models, like advanced recruiting platforms and personalized tutoring.

The cost of AI intelligence is decreasing, bringing consumer AI applications closer to reality.

Full-stack companies, which combine software with operational services, are becoming more feasible with AI agents handling operational tasks.

Following curiosity and working with emerging technologies is a more effective startup strategy in the AI era than traditional customer discovery.

Infrastructure and tooling for deploying and managing AI agents represent a significant area for new startups.

AI advancements highlight the need for platform neutrality to foster competition and innovation, similar to past internet battles.

THE RESURGENCE OF OLD IDEAS THANKS TO AI

The conversation highlights how advancements in AI, especially Large Language Models (LLMs), are making previously unfeasible startup ideas viable. This includes sectors like recruiting, where AI can now handle sophisticated evaluations that once required significant manual effort and data collection. Companies like Meror are cited as examples, leveraging AI from day one to evaluate candidates, a stark contrast to earlier recruiting startups like TripleBite that spent years building proprietary evaluation systems.

AI'S TRANSFORMATIVE IMPACT ON RECRUITING AND EVALUATION

The recruiting industry, once a hotbed for marketplace startups that struggled with effective candidate evaluation, is now ripe for innovation. AI codegen and LLMs allow for immediate and nuanced assessment of technical skills, enabling platforms to launch with robust evaluation capabilities from the outset. This shift also expands the market, allowing for more sophisticated screening of senior candidates, not just entry-level or international applicants, as demonstrated by companies like Apriora.

PERSONALIZED LEARNING AND EDUCATION REVOLUTIONIZED BY AI

AI is poised to deliver on the long-held dream of truly personalized learning. Platforms can now offer tailored educational experiences that adapt to individual student needs, a significant leap beyond traditional e-learning. Examples like Revision Dojo for exam prep and Adexia for grading assignments showcase AI's ability to assist both students and educators, addressing pain points like tedious grading that contribute to teacher churn and improving learning outcomes.

THE EVOLVING ECONOMICS OF AI AND CONSUMER APPLICATIONS

While raw intelligence remains costly, its price is rapidly decreasing due to model distillation and advancements in model training. This trend suggests that consumer AI applications, once cost-prohibitive, may soon become accessible, potentially for pennies per use. This economic shift could revive the subscription models common in Web 2.0, where a vast user base receives services for free, with a smaller segment paying for premium features, driving significant business models.

THE RETURN OF FULL-STACK STARTUPS ENABLED BY AI AGENTS

The 'tech-enabled services' wave of the 2010s, where companies tried to integrate software with extensive human operations, often failed due to poor gross margins. However, AI agents are now making full-stack models viable again. These agents can automate many of the operational tasks that previously required large human teams, allowing these 'full-stack' companies to function more like lean software businesses with better margins and scalability.

THE CRITICAL NEED FOR PLATFORM NEUTRALITY IN AI

The current AI landscape, much like the early internet, faces challenges regarding platform neutrality. The slow development of voice assistants like Siri and the fragmented integration of AI within large tech companies suggest a need for open competition. Historical examples like net neutrality and browser choice in Windows highlight how government intervention can foster free markets, leading to innovation and abundance, a principle that could be applied to AI assistants and platforms.

NAVIGATING THE AI LANDSCAPE: INNOVATION VS. INCUMBENCY

While major players like Google possess advanced AI capabilities (e.g., TPUs for efficient large context windows), they struggle to integrate these effectively into consumer products, often due to internal org structures and a focus on existing revenue streams. This presents an opportunity for startups that can move faster, focus on specific use cases, and deliver superior user experiences, even if they appear smaller relative to tech giants. Companies that stick with emerging tech often find their moment.

THE STRATEGY SHIFT: FROM CUSTOMER DISCOVERY TO TECHNOLOGICAL CURIOSITY

The traditional startup advice of 'sell before you build' and intensive customer discovery, while still valuable, is becoming less critical in the AI era. The rapid pace of technological advancement means that following one's curiosity and exploring new AI capabilities can lead directly to groundbreaking ideas. By working at the 'edge of the future' and experimenting with prompts, data, and prompts, founders can identify and develop 'magical' outputs that create new markets.

BUILDING DURABLE BUSINESSES: MODES AND GROSS MARGINS

For AI startups to achieve long-term success, they must focus on building durable businesses with strong modes, such as brand loyalty and switching costs. While AI lowers the cost of intelligence, founders still need to consider gross margins. High gross margins often indicate simpler, more scalable products, allowing companies to focus on product improvement and user acquisition, essential for sustained exponential growth and avoiding the pitfalls of complex, operation-heavy businesses.

THE FUTURE OF AI INFRASTRUCTURE AND TOOLING

There remains a significant opportunity in building the foundational infrastructure and tooling necessary for deploying and managing AI. This includes areas like prompt management, model evaluation, and agent orchestration. The rapid evolution of AI capabilities means that the tools to build, deploy, and effectively utilize these systems are still largely in their nascent stages, representing a fertile ground for new startups.

Mentioned in This Episode

●Products

●Software & Apps

●Companies

●Organizations

●Concepts

●People Referenced

Common Questions

AI has unlocked new possibilities for startups, particularly in areas like recruiting marketplaces where evaluation was previously a bottleneck. Areas like personalized education and complex service marketplaces can now be built more effectively.

Topics

Recruiting Technology Marketplace Models Full-stack Startups Platform Neutrality

Mentioned in this video

Companies

Revision Dojo

A company that helps students with exam preparation, offering a more engaging and tailored version of flashcards, with significant daily active users.

Apriora

A YC-funded company that builds AI agents to handle screening for technical interviews, expanding the capabilities of pre-screening tools with LLMs.

Web Van

A cautionary tale of a failed startup from the past, used as a comparison to highlight how new technologies can enable similar models to succeed later.

Adexia

A company providing tools for teachers to grade assignments, addressing a common pain point that contributes to teacher churn.

Meror

A company identified as a hot AI startup that operates a marketplace for hiring software engineers, leveraging LLMs for the evaluation piece from day one.

Atrium

A startup that offered tech-enabled services for law firms, founded by Justin Khan, which ultimately did not succeed due to issues common to full-stack startups of that era.

Deepgram

Concepts

MLOps

Machine Learning Operations, a field that was previously overlooked by YC but has become crucial with the advancement of AI, highlighting the importance of timing and perseverance.

codegen

Code generation, a field that emerged as a new focus for companies, including Windsurf, after pivoting from MLOps.

Web 2.0

A past era of the internet characterized by the premium subscription model, which may see a return with the decreasing cost of AI intelligence.

Software & Apps

Gemini Pro

A Google AI model whose consumer usage is significantly lower than ChatGPT, despite its comparable or superior performance in some tasks.

Clevver

A platform mentioned in the context of school integrations, where authentication and login would be important for AI companies like Speak.

BERT models

A type of deep learning model previously hosted on Hugging Face, which were functional but not widely adopted due to limitations compared to more recent AI advancements.

A former PM at Google and partner at YC, who wrote an essay detailing why Google's Gemini integration with Gmail was built incorrectly.