- Embedding VC Newsletter
- Posts
- Embedding VC Newsletter - Mar 2025
Embedding VC Newsletter - Mar 2025
Hi friend,
We’re at the end of March—bringing longer days, fresh energy, and maybe even a hint of spring 🌱. Whether you're wrapping up Q1 goals or just grateful for the extra sunlight, it's a great time to reset and refocus 🔄.
Welcome to the March edition of the Embedding VC Newsletter, where we provide insights into the latest developments in our portfolio and the AI ecosystem.
👋 Refresh about Embedding VC:
If it is the first time you received our newsletter. As a fresher, we invest $100K to $1M checks into early-stage generative AI startups. If you or your founder friends are building something exciting, feel free to reach out to us at [email protected].
✨ Insight
![]() |
In the rapidly evolving landscape of artificial intelligence, AI agents have emerged as one of the most promising frontiers for innovation and investment. These autonomous systems represent a fundamental shift in how AI interacts with the world, moving from passive response models to active, goal-oriented execution. More about AI agent, Manus, and Deep Research on our Substack.
🗞️ Vectors In The News
Updates from Embedding VC portfolio companies:
AI-native platform that makes practicing trademark law easier
Domain-Specific Model Outperforms: Huski AI’s new 8B model achieved top performance in handling “Likelihood of Confusion” refusals, which is the denial of a trademark application due to its similarity to an existing trademark, potentially causing consumer confusion. It is one of the most complex areas in trademark law and often requires thousands of word long arguments, and now Huski has outperformed leading open and closed-source models in this area.
Upgraded Agent Capability: The Huski AI agent now guides users through the full trademark process: name generation, clearance, office action simulation, and even image mark creation, significantly boosting attorney efficiency and confidence. They are also testing a fully autonomous trademark agent that runs the entire process from a single prompt, with more to be revealed at INTA 2025.
AI coding assistant for professional software engineers and large codebases
Anthropic Case Study: Augment AI launched its developer tool on Google Cloud’s Vertex AI, integrating Claude to help engineers understand and manage massive codebases. The rollout marks a major step toward Augment’s mission of streamlining software development through AI-powered context awareness and real-time support across complex engineering environments.
Outperforming Cursor: In a side-by-side test, Augment Code proved significantly faster, more accurate, and more cost-efficient than Cursor, with its full-repo indexing and concise responses. Unlike Cursor’s limited context window and verbosity, Augment handles large, evolving codebases with ease (another video comparison) and integrates seamlessly with major developer tools.
New Releases: Augment released the #1 open-source agent on the SWE-bench Verified leaderboard, hitting a 65.4% success rate on real-world GitHub issues using Claude Sonnet 3.7 and OpenAI’s O1—no custom models, just smart orchestration. They also released AugmentQA, a real-world code retrieval benchmark built on authentic engineering questions and full repos, revealing how top synthetic leaderboard models fall short in practice.
First AI “software engineering mentor” to support the entire software lifecycle
PR Review Leaderboard Launch: Entelligence AI launched the first-ever Human Leaderboard for Code Review, flipping the script by letting developers test their skills against top models like Sonnet 3.7 and o3-mini. The new PR Review eval platform aims to drive bottoms up growth and awareness.
Manager Insights Portal: They also launched a Manager Insights Portal to surface team performance trends. The portal provides detailed qualitative and quantitative insights over team performance, sprint reviews and individual performance reviews.
💡 New Embedding Vectors
Embedding VC joins Autograph’s Pre-Seed round as the first VC investor
Round details: $2.6M pre-seed raise led by NFX, with Long Journey Ventures and Cibersons participating as follow-on investors
What they do: Autograph AI provides a personal storytelling platform powered by generative AI. Its service centers on an AI agent, Walter, who regularly calls users (or their loved ones) to interview them about their life history and reflections for digital legacy preservation. Autograph's mission is to record and retell the history of humanity in every character's own words through Walter.
Event with Neal Stephenson: Founder Cristian Cibils Bernardes hosted a standout panel at South Park Commons with sci-fi legends Neal Stephenson and Ken Liu, investor Cyan Banister, and cognitive scientist Joscha Bach. Cristian is also a sci-fi writer on the side; the event showcases parallels between writing and building Autograph—both use narrative as a tool to shape how we imagine and create the future through technology and collective curiosity.
Product waitlist: Sign up for the waitlist to get free lifetime access to the platform before April 14!
🎤 Builder Spotlight
Pratik Verma, Co-founder and CEO of Okahu (right)
Prasad Mujumdar, Co-founder and CTO of Okahu (left)
Meet the Founders: Pratik Verma, Co-founder and CEO, is a serial entrepreneur with multiple exits. Before founding Okahu, Pratik was founder of BlueTalon, which raised $26M in total and was acquired by Microsoft in 2019. Pratik joined Microsoft Azure as part of the acquisition, and helped launch the data governance product as Group Product Manager. Pratik received his PhD from Stanford University.
Prasad Mujumdar, Co-founder and CTO, is an expert on data, security and monitoring. He was Principal Product Manager at Microsoft, who also joined Microsoft through the BlueTalon acquisition. Before BlueTalon, Prasad worked at Cloudera and IBM. Prasad has been Apache Open-source Project Management Committees (PMC) in multiple impactful big data projects such as Hive, Sentry and Flume for 10+ years, which is a prestigious role for engineers in the open source ecosystem. The two co-founders have worked together for more than 8 years.
Founder Fireside Chat - Q&A with Pratik and Prasad
Q: What inspired you to start Oakhu? | A: We saw adoption of LLMs based AI apps proliferating but getting stuck in going to prod. We learned that cloud AI developers found it difficult to instrument and analyze telemetry to understand where their apps were breaking because they were using lots of different frameworks, models and tools. This was during our time at Microsoft where we’d spent a few years after acquisition of our previous startup that built data governance at scale for cloud users. We felt like we had a unique approach to simplify observability of LLM based apps given that most of the issues in such apps are really understanding prompts and data content. This is turning out to be even more true in an agentic apps world we’re seeing emerge. |
Q: What are you most excited about for the next three years? | A: Widespread adoption of agentic apps in the enterprise. Most of the apps that we have seen in prod using LLMs in the enterprise are mostly using summarization, automated data processing or iteratively assisting users with finding information. With agentic apps, focus can move to doing the task rather than describing what tasks are to be done. Understanding where LLMs vs agentic code breaks in this situation becomes even more impactful. |
Q. What have you enjoyed about working with Roger and Embedding VC? | A: Roger has been a wonderful thought partner who’s very responsive and supportive. We feel like we can always openly share what’s working well or where we are stumbling, especially when the issue is a thorny technical problem. Roger has been helpful in thinking through such problems with us and pointing us to other resources in the Embedding and wider ELC community. |
👋 Here’s to spring with fresh momentum!
See you soon! We are Embedding VC.
Embedding VC invests in early-stage Generative AI startups.