The Next Unicorns: 7 Top AI startups from the HF0 Residency

Key Concepts

AI creative suite, traffic prediction, Tower of Babel analogy, personalized content, AI-driven smart speakers, developer ecosystems, open source, LLM driven, jailbroken AI, autonomous coding agents, adaptive websites, reliable AI, personalized AGI, revenue attribution, knowledge graphs, instructible voice model, solar incentives, LLM marketplace, inference as a commodity, heterogeneous ecosystem, plugins, model comparison.

Korea - AI Creative Suite

• Main Topic: Building an AI creative suite.
• Key Points:
  • Korea is building an AI creative suite.
  • The CTO, Diego Rodriguez, shares three stories to illustrate the challenges and opportunities in the AI space.
  • The first story highlights the difficulty of predicting "traffic" (unforeseen consequences) in AI development, even when the core technology (like "cars") seems predictable.
  • The second story uses the Tower of Babel as an analogy for the fragmentation and lack of consensus in technology choices (e.g., React vs. JavaScript).
  • The third story describes a Netflix use case where AI is needed to manage and personalize content for diverse audiences, such as different towns in India.
  • Korea was used for broadcasting an ad with Fox to millions of people within two days of signing up.
• Data: Raised $25 million with eight people.
• Call to Action: An email address is provided to prioritize job applications.

OpenHome - AI-Driven Smart Speaker

• Main Topic: Building an AI-driven smart speaker platform.
• Key Points:
  • OpenHome is building the first AI-driven smart speaker, emphasizing natural language interaction and customization.
  • The goal is to move beyond awkward command-based language to seamless, intuitive conversations with AI.
  • The platform focuses on developer ecosystems, allowing users to build custom smart speakers and voice AI applications.
  • Developers want open source, LLM-driven, and fully jailbroken AI.
  • Voice AI can be integrated into various hardware devices, such as talking toys, AI robots, and AI appliances.
• Examples: Developers are building games, personalities, and home automation tools.
• Data: Over 10,000 developers are building on OpenHome.
• Offer: Releasing 500 free dev kits for attendees.

Koframe - Adaptive Websites

• Main Topic: Creating adaptive and personalized websites using AI.
• Key Points:
  • Koframe aims to make websites adaptive and personal, giving each customer experience its own AI growth team.
  • The company generated $20 million for Europe's largest travel company and increased click-through rates for India's largest company (a $400 billion enterprise) in just a few weeks.
  • Koframe is the only marketing tech company partnered directly with OpenAI.

Federous AI - Reliable AGI

• Main Topic: Building reliable and personalized AGI.
• Key Points:
  • Federous AI argues that "scale is dead" and reliability is more important than simply making AI models bigger.
  • The company focuses on building AI that is 99.9% reliable for everyday tasks, such as booking airline tickets and filing taxes.
  • They emphasize that reliability is revenue, as it unlocks billion-dollar applications in e-commerce and B2B sales.
  • Federous AI has built an action R1 agent that outperforms Claude, Sonnet, Gemini, and OpenAI in form filling with absolute reliability.
• Arguments:
  • DeepMind founder and CEO says compound AI agents errors will take more than 10 years to fix.
  • Lunn says that we need a new AI architecture to push the paradigm forward.
  • Over 90% of AI projects fail in production due to reliability issues.
• Quote: "Reliability is revenue."

Upside - Forensic Revenue Attribution

• Main Topic: Using AI to improve revenue attribution and sales intelligence.
• Key Points:
  • Upside focuses on forensic revenue attribution and intelligence, helping sales and marketing teams understand what works.
  • The company uses LLMs to extract important details from unstructured data, such as email records, and turn raw enterprise data into a structured map of the world.
  • They are building a data command center that teams can use to reach customers more effectively.
  • The company is hiring for roles in knowledge graphs, data analytics agents, and graph analytics.

Open Audio - Instructible Voice Model

• Main Topic: Building an AI voice model that is expressive and controllable.
• Key Points:
  • Open Audio is building an AI voice model that is expressive and controllable, starting with the founder's personal experience.
  • They introduced S1, the first ever instructible voice model, which allows users to control not just what to say, but how to say it.
  • S1 is the most expressive voice model in the world.
  • Fish Audio has grown from 400K to 5.5 million annualized revenue in just four months.
• Model Comparison: Open Audio claims to have "blown 11 labs out of the water" based on TTS arena rankings.

Glow - Solar Incentives

• Main Topic: Incentivizing solar energy adoption using blockchain technology.
• Key Points:
  • Glow builds solar farms by incentivizing their construction with glow tokens.
  • The company aims to revolutionize how governments subsidize solar energy, turning the same subsidy into 10 times as much solar.
  • Glow is building incentive projects in India, Mexico, and Lebanon.
• Data:
  • Raised a $30 million round led by Framework and USV Union Square Ventures.
  • Posted the world record for onchain deepin revenue, doing more than $10 million of revenue in a single day.

Favored - Engaging App Development

• Main Topic: Building highly engaging applications.
• Key Points:
  • Favored focuses on building the world's most engaging app.
  • The company scaled from 1 to $100 million annualized.

Open Router - LLM Marketplace

• Main Topic: Creating a marketplace for language models.
• Key Points:
  • Open Router is the first and largest LLM marketplace, providing a single API to access various language models with the best prices, performance, and uptime.
  • The goal is to make a heterogeneous ecosystem homogeneous, treating inference as a commodity.
  • Open Router offers features such as tool calling, edge case handling, caching, fine-grain privacy settings, and observability.
  • They have built their own middleware for doing inference called plugins, which are more powerful than MCPs.
• Technical Challenges: Building a system for fast routing and model comparison.
• Data: The platform has grown 10 to 100% every month for the last two years.

Conclusion

The presentations highlight diverse applications of AI, ranging from creative tools and smart speakers to personalized websites and reliable AGI. Key themes include the importance of developer ecosystems, the need for reliable AI, and the potential of AI to address real-world problems in various industries. The speakers emphasized the importance of focusing on user needs and building practical solutions rather than simply scaling up existing models.