- Embedding VC Newsletter
- Posts
- Embedding VC Newsletter - Sep 2024
Embedding VC Newsletter - Sep 2024
Hi friend,
If you are new here: Hello World! We are Embedding VC, an early-stage VC firm backing Generative AI startups.
Welcome to the September edition of Embedding VC Newsletter, where we provide insights into the latest developments in our portfolio and the AI ecosystem.
P.S. Experiencing newsletter déjà vu? We ran into some issues with our newsletter delivery software and switched to another provider. If this landed in your inbox twice, our apologies for the overeager delivery!
📊AI Landscape
Finally 🍓 are available: OpenAI finally released their o1-preview and o1-mini models to the public, bolstering the momentum of their latest $150 billion fundraising round. The o1 models were trained to "think" through problems before responding, incorporating a prompt engineering technique known as Chain of Thought reasoning (CoT) directly into the model. On tasks that require a high level of logical reasoning, such as Competition Math or Code, it performed significantly better compared to previous SOTA model GPT-4o.
Source: OpenAI
In our opinion, the release of this new model paradigm marks a significant milestone in the development of AGI and computing. As LLM training “scaling laws” slow down due to physical limitations (size of GPU clusters, energy constraints, and availability of high-quality training data), computation will increasingly shift towards serving inference instead of pre/post-training (as illustrated in this graph from Jim Fan’s X post). The new model uses inference-time computation to solve complex reasoning problems. By testing multiple strategies and scenarios, the model eventually converges on good solutions, assuming it can also evaluate what constitutes a good answer. The reasoning component is unlikely to significantly increase the model’s size, as it doesn’t need to memorize as much information compared to an LLM.
Source: Jim Fan
Although the model’s capabilities are impressive, it’s not flawless and sometimes struggles with simpler tasks — the infamous example being, “What’s bigger, 9.11 or 9.9?”
Nevertheless, OpenAI’s release gives developers a glimpse of the model’s potential, and we are excited about what it could enable. Limited reasoning capabilities have long been a barrier preventing several agentic applications from going mainstream and gaining broader customer adoption — a challenge many of our founders have highlighted. While the most powerful version of this new model is currently too slow and expensive to release to the public, we anticipate that: 1) model sizes will shrink over time, and 2) through distillation, existing smaller models will see performance improvements, and open-source counterparts may become available in the near future.
Record-breaking investment, but concentrating on fewer deals: AI startups raised $24 billion in Q2 2024, more than doubling the $11 billion in Q1, marking the largest quarterly total ever for AI funding. However, only 1,000 deals were announced, the lowest in years, with large unicorns like xAI ($6B), CoreWeave ($1.1B), and Scale AI ($1B) securing most of the funding. Most of the funding continues to flow into the foundational model and infrastructure layers, with Perplexity being an outlier in the application layers. This presents a favorable window of opportunity for our founders who are able to raise their rounds, as there is less competition.
AI Coding unicorns on the rise: In Q2 2024, unicorns like Augment (raised $300M), Magic (raised $320M), Poolside (raised $400M) and Anysphere (the team behind Cursor, raised $60M) secured significant funding to enhance the efficiency of software development. While their high valuations still outpace financial performance — except for Codeium (raised $150M), which has already generated eight-figure revenue — we believe AI coding is one of the few practical use cases for AI agents, and likely the first to scale. This is due to two main factors: 1) there is an abundance of training data available; and 2) the outputs of these tasks can be easily tested for correctness. Junior software developers are likely to see their roles disrupted in the coming years.
🏄Firm Update
We co-hosted the AI+DevTools Startup Demo Session at the ELC Annual 2024, in SF on August 27-28. With warm introductions from several VC collaborators, including Lightspeed Ventures, Index Ventures, Essence VC, 8VC, and First Rays Venture, we curated and showcased 15 promising early-stage companies in the space. Three Embedding VC portfolio companies, Entelligence.AI (AI TechLead), Superlinked (vector computer), and Okahu (AI observability), presented their products to an audience of engineering leaders. Founders were able to connect with potential customers and gather valuable insights. For example, Entelligence.AI gained 19 strong customer leads after presenting at our demo sessions.
![]() ![]() | ![]() ![]() |
We have just received an investment from Decile Capital. Decile Capital is a FoF specifically designed to invest in the most promising new managers launched by VC Lab, the leading venture capital accelerator globally. VC Lab has accelerated over 500 VC firms to date, from a pool of more than 17,000 candidates. The fund focuses on the top 5% — so far, fewer than 15 funds have received investment. This investment is a testament to our thesis and track record.
👋New Embedding Vectors
A list of new Embedding VC portfolio companies:
Keywords: AI Engineer
What they do: Entelligence is building the first AI “software engineering mentor”. It supports engineers throughout the entire software lifecycle, from onboarding and codebase visualization to feature design, code review, and improvement, while also providing insights to managers on team progress.
How they stand out: The platform addresses the inefficiencies in software development (engineers in sizable organizations spend ~85% of their time navigating codebases and waiting for code reviews/feedback). It aligns well with our fund's core thesis (Core AI + Engineering Driven + DevTools). Recent traction underscores its market potential and growing traction.
Keywords: AI x LegalTech, Trademark Protection
What they do: Huski is an AI-native platform that makes practicing trademark law easier. Powered by a proprietary AI perception engine, it offers advanced tools for the entire life cycle of a brand, from creation and registration to monetization and protection.
How they stand out: Huski differentiates itself with its combination of a proprietary search engine for identifying infringements and integrated support for litigation and enforcement actions. Additionally, it aligns with our fund's positive view on AI x LegalTech, and has demonstrated market value by securing major clients like Chanel and LV.
Keywords: AI Real Estate Agent
What they do: GetMyHome offers on-demand home buying services for sophisticated buyers. It is building AI-powered data tools including smart home suggestions, automated SWOT analysis, AI-powered Comparative Market Analysis (CMA), offer bidding strategy and a custom-built real estate AI agent.
How they stand out: GetMyHome is founded by a team of serial entrepreneurs and AI experts from Google, Apple, Facebook, and Stanford, and seasoned real estate professionals. Their blend of deep technical knowledge and extensive real estate experience uniquely positions GetMyHome in the market. Find out more about the leadership driving this innovation in the Builder Spotlight section at the end of this newsletter.
Keywords: Developer Tools, Observability
What they do: Okahu provides AI observability to help enterprises building LLM-based apps in the cloud move to production quickly, easily & safely without having to deal with DIY observability stack.
How they stand out: Okahu addresses the fragmented GenAI observability market with a $58B opportunity. Founded by seasoned entrepreneurs with successful exits and deep tech expertise, Okahu aligns with our fund’s focus on key Gen AI DevTools and high-growth observability.
👍Vectors In The News
New Round of Financing: MagicSchool has closed a Series A round of $15M led by Bain Capital Ventures. When we at Embedding first met Adeel in July 2023, we were impressed by his vision, which resonated well with our enthusiasm for opportunities in AI x Education.
Accelerated Traction: They have surpassed 2 million users in May (11 months since launch)! Their revenue growth is even more impressive, as the team began focusing on monetization in January this year.
New Partnership: Engaging with Seattle Public Schools, MagicSchool starts a pilot program to benefit both students and teachers, exemplifying its pledge to enhance education via technology.
Applauded Product Launch: Monterey’s recently launched Monterey AI 2.0 – AI Native User Insights Platform, was endorsed by Garry Tan (CEO of Y Combinator). Please share this product with your product or CEO friends, as it can significantly improve their product and save considerable time and resources on product analytics.
Building Influence: Monterey successfully hosted the first Voice of Customer Summit - a knowledge-sharing forum for VoC leaders, gaining recognition as a market leader in AI-empowered product analytics.
New Partnership: Superlinked have now officially announced the Seed round closed last year which is led by Index Ventures and Tomasz Tunguz, along with their partnerships with database providers like MongoDB and Redis, as well as platform partners such as Starburst and Dataiku.
Community Leadership: Superlinked’s Vector DB Comparison dashboard on VectorHub, their open-source educational platform, is welcomed by developers who aim to make informed decisions for information retrieval and feature engineering.
Founder Spotlight: cambioML Founder Rachel Hu's accomplishments have been recognized internationally, being listed as one of Forbes 30 Under 30 Asia for 2024, spotlighting her visionary initiative in Enterprise Tech.
💌Builder Spotlight
Haibo Zhao, Co-founder and CEO of GetMyHome
Meet the Founder: Prior to GetMyHome, Haibo was the co-founder and CTO at Haystack, previously with Google and Snap. At Snap, he founded the AI filter and lens monetization team. At Haystack, he led to $5M ARR and 300+ enterprise clients, including Chime, Credit Karma, Foursquare, BuzzFeed and Pinterest. He holds a PhD in Artificial Intelligence from the University of Georgia.
Founder Fireside Chat - Q&A with Haibo
Q: What inspired you to start GetMyHome? | A: I believe AI is the new electricity, it will replace most of the repetitive human work. Since March 2024, the landmark National Association of Realtors (NAR) settlement has reshaped the real estate landscape, requiring greater transparency in commissions and giving buyers the power to choose and pay only for the services they truly need. This has opened up a whole new market for sophisticated buyers, making it the perfect time to offer AI-driven solutions that cater to their specific needs. |
Q: What are you most excited about for the next three years? | A: I am thrilled to see how our customers will utilize our AI agent workflow for more use cases. Currently, we are focusing on real estate customer flows. However, we believe that the application of AI in customer-facing workflows shares many similarities across different industries. We are excited to build a flexible system capable of handling user-facing workflows in more verticals. |
Q. What have you enjoyed about working with Roger and Embedding VC? | A: Working with Roger and the Embedding VC team has been an exceptional experience. They are the most tech-founder-friendly team I have worked with, distinguished by their deep engineering roots. Their profound understanding of the underlying technology, combined with a forward vision and practical grasp of capability boundaries, sets them apart. |
See you soon! We are Embedding VC.
Embedding VC invests in early-stage Generative AI startups.