Live from GTC: AI That Thinks Like A Human with SurrealDB's Tobie Morgan Hitchcock

Summary

---

Podcast Summary: Live from GTC – AI That Thinks Like A Human with SurrealDB's Tobie Morgan Hitchcock

Podcast: Reshaping Workflows with Dell Pro Precision and NVIDIA RTX PRO GPUs
Host: Logan Lawler
Guest: Tobie Morgan Hitchcock, SurrealDB
Date: March 19, 2026
Location: NVIDIA GTC, Inception Area

---

Episode Overview

In this bonus episode recorded live at NVIDIA’s GTC conference, host Logan Lawler sits down with Tobie Morgan Hitchcock from SurrealDB, a database startup making waves in the AI space. The conversation focuses on how SurrealDB’s multi-modal approach to handling data is "revolutionizing the context layer" for AI agents and LLM-powered workflows—a critical step beyond today’s mainstream vector databases. They discuss the future of AI workflows, the unique abilities of SurrealDB, and real-world use cases that highlight the need for richer, more human-like reasoning in AI.

---

Key Discussion Points & Insights

1. What Is SurrealDB and Its Core Innovation?

• SurrealDB Defined:
  • Described as a "multimodal database" focused on revolutionizing the context layer—the data delivered to AI agents or large language models (LLMs) for inference.
  • Goes beyond conventional vector search to incorporate full-text search, graph, and document data.
  • Quote, Tobie:

    "By combining all these different data modalities together and then using that data before you push it into an agent or an LLM, you can get better accuracy and a better response from the LLM." (00:46)

2. Why Go Beyond Vector Databases?

• Typical AI workflows use vector stores for semantic similarity, but human reasoning relies on relationships and meaning that transcend word similarity.
• SurrealDB enables richer connections within data—such as mapping relationships between entities, events, and entire organizations.
  • Quote, Tobie:

    "Humans don't just think around semantic matching... We think about things in terms of relationships and understanding and meaning. And with SurrealDB, that's what we do." (01:46)

• Real-world example: Combining multiple data types allows organizations to unlock insights that would be impossible with just vector matching.

3. Key Use Cases for SurrealDB

• AI Agents & LLM Inference:
  • SurrealDB isn’t focused on LLM training—but rather on powering AI agents and inference with better, context-rich data.
• Large Scale Knowledge Graphs:
  • Enables petabyte-scale mapping of organizational data and relationships.
  • Crucial for digital twins, organizational mapping, and other non-agent-based AI applications.
  • Quote, Tobie:

    "One of our biggest kind of use cases... is massive petabyte scale knowledge graphs." (03:18)

• General Value Add:
  • The ability to contextually map and query disparate data (events, people, organizations) that spans across multiple data modalities.

4. Openness and Accessibility

• Open Source Commitment:
  • SurrealDB is open-source, supporting easy downloads and community-driven development.
  • Quote, Tobie:

    "SurrealDB is open source, first and foremost, and you can try us out and download us and get developing at surrealdb.com." (03:58)

---

Memorable Quotes & Moments

• On What Sets SurrealDB Apart:

  "We go from vector to full-text search, to graph, to document data. And by combining all these different data modalities together... you can get better accuracy and a better response from the LLM."
  (Tobie, 00:46)

• Human-Like Reasoning in AI:

  "Humans don't just think around semantic matching... We think about things in terms of relationships."
  (Tobie, 01:46)

• Clarifying Focus:

  "We're definitely not focused on the training side. We're definitely more focused on the inference, the agent side."
  (Tobie, 03:18)

• Call to Action—Try SurrealDB:

  "You can try us out and download us and get developing at surrealdb.com."
  (Tobie, 03:58)

---

Important Timestamps

• 00:19 – Introduction to the episode, GTC setting, and SurrealDB
• 00:46 – Tobie explains SurrealDB’s multimodal database and context layer focus
• 01:18 – 01:46 – Why vector-only approaches aren't enough and SurrealDB’s advantage
• 03:18 – Main use cases: Agents, knowledge graphs, digital twins
• 03:58 – How to access SurrealDB (open source) and final remarks

---

Tone & Style

Throughout the episode, Logan maintains an energetic, inquisitive tone, pushing Tobie to clarify technical distinctions in a way that’s accessible to listeners. Tobie’s responses are clear, practical, and focused on real business problems, emphasizing the gap between how data is currently used and how SurrealDB aims to close this gap.

---

Conclusion

This episode is an insightful, technical, yet accessible exploration of the future of AI infrastructure. SurrealDB’s approach—combining multiple data types and emphasizing relationships for context-rich AI—stands out as a potential game-changer for inference workflows and enterprise applications. Open-source accessibility ensures that developers and organizations alike can experiment and build on these capabilities today.

To learn more or try SurrealDB:
https://surrealdb.com

Transcript

[Logan]

Foreign.

[Host]

Welcome to Reshaping Workflows with Dell Pro Precision and Nvidia, where innovation meets real world impact in high performance computing.

[Logan]

Welcome back to another episode, well, bonus episode of Reshaping Workflows with Delpro Precision and Nvidia RTX GPUs. I'm Logan, your host. You already know me. This is day three of gtc. We're in the Nvidia Inception area, probably one of my favorite areas, because you get to hear about the newest, latest, greatest kind of startups that have partnered with Nvidia. So I'm with Surreal DB with Toby. So Toby, real quick, give kind of an overview of your position at Surreal DB and then we'll get right into it.

[Toby]

Awesome. Okay, so Surreal DB is a multimodal database that is kind of revolutionizing the context layer. And what that means is the data that gets pushed into an agent or into an LLM. Right now a lot of companies are working with just one type of data that's typically vector. But we go beyond that. We go from vector to full text search, to graph, to document data. And by combining all these different data modalities together and then using that data before you push it into an agent or an LLM, you can get better accuracy and a better response from the LLM.

[Logan]

Okay, so first kind of question, you're right. Data really powers AI, whether it's locally in the data center, et cetera. We're about vector DVs all the time. You know, they're now GPU accelerated. But when we talk graph, other things, what's the real advantage of, you know, Surreal DB that you kind of bring to the table? Like you talk about agents, right? Give me like a, you know, I'm sure customers or anything like that, but like a use case, right, where you, where your platform kind of solves a problem that's, you know, maybe was previously unsolvable.

[Toby]

Yeah, it's a good question. I think if you, if you think about the data that goes into the AI agents right now and then you compare that to how humans think, it's very different. Right? Humans don't just think around semantic matching or similarity of the meaning in a word. We think about things in terms of relationships and understanding and meaning. And with Surreal to be, that's what we do. We bring the similarity search in, we bring that full text search in. But actually we go beyond that and we have. You can build this kind of relationships between your data sets, between the meaning of words, the meaning of entities inside your data. That becomes very relevant when you're not just doing, let's say, a coding task, but when you're trying to map organizational wide data, maybe it's data that looks at organizations and emails and events and people, and that can't just be matched based on the similarity of a word. So it really know whether you're dealing with accuracy of agents at a small scale or whether you're dealing with large data sets which cross an entire organization. That's where SERUTBI benefits.

[Logan]

Okay, so you mentioned agents a couple times. Is that kind of where you fit your kind of focus on agents or. I mean, I can also see, and this is the first time we ever talked, but I could see. I mean, I'll use personal experience from Dell. We've got tons of data, but it's getting access to that data and being able to put it in the right context to, to fine tune models, et cetera. Would you say serialdb is more focused on agents or is it kind of the fine tuning like large language model?

[Toby]

So we're definitely not focused on the training side. We're definitely more focused on the inference, the agent side. But I'd say it's not just about agents. It's about, you know, one of our biggest kind of use cases on Soaribi is massive petabyte scale knowledge graphs. That's the ability, you know, to map entire data set and organization. For digital twins as well, for AI applications that aren't necessarily autonomous agents as well, that's definitely a big use case. And then agents as well. But yeah, we're not really focused on the training side. We're more focused on the infant side.

[Logan]

All right, well, you heard it here first. So Toby, before we go, tell everyone where they can find you, learn more about Cyril db Where can they find everything they want to know if they're interested in learning more?

[Toby]

Awesome. So Cyril DB is open source first and foremost and you can try us out and download us and get [e-mail]

[Logan]

all right, you heard it here. We love open source. So with that Logan GTC 2026. We'll see you on the next one. This podcast was Prod in partnership with Amaze Media Labs.