Jerry Liu is the co-founder and CEO of LlamaIndex. The items below are things Liu said on the Latent Space, What's AI, Chain of Thought, Data Exchange and DataFramed podcasts and in a Microsoft for Startups interview, about RAG, agents and building LlamaIndex. — Latent Space: RAG Is A Hack - with Jerry Liu from LlamaIndex (Oct 2023), transcript.

Part 1: The Origin and Evolution of LlamaIndex
- On GPT Index: On The Data Exchange Podcast, Liu said that "I called it GPT index at the time which is basically a toolkit to like index your private data so you could use it with a language model". — The Data Exchange Podcast, Episode 187: Jerry Liu of LlamaIndex (Jul 2023), captions.
- On community-driven growth: On the Latent Space podcast, Liu said: of LlamaHub "it's supposed to be a community driven hub" and "the bulk of the peers are completely community contributed". — Latent Space: RAG Is A Hack - with Jerry Liu from LlamaIndex (Oct 2023), transcript.
- On Stacks That Needed to Be Invented: On the Latent Space podcast, Liu said "there was just stacks that needed to be invented to actually solve this type of problem, because language models themselves didn't have access to this data". — Latent Space: RAG Is A Hack - with Jerry Liu from LlamaIndex (Oct 2023), transcript.
- On the Limits of a Context Window: On The Data Exchange Podcast, Liu said of the early GPT-3 window: "you can dump like a bunch of text in there but you can't dump like a gigabyte of data". — The Data Exchange Podcast, Episode 187: Jerry Liu of LlamaIndex (Jul 2023), captions.
- On Documentation as Education: On the What's AI podcast, of developers onboarding onto new concepts, Liu said "they want to see the documentation as an educational resource that supports them throughout their developer journey". — What's AI Episode 25: Jerry Liu (Dec 2023), transcript.
- On Modularity of Text: On the Latent Space podcast, of working with text, Liu said that "from a software engineering principle, it just makes things way more modular". — Latent Space: RAG Is A Hack - with Jerry Liu from LlamaIndex (Oct 2023), transcript.
- On Timing and Circumstance: On The Data Exchange Podcast, Liu said that "that kind of was a perfect combination of just like timing and Circumstance and led to the growth of the project". — The Data Exchange Podcast, Episode 187: Jerry Liu of LlamaIndex (Jul 2023), captions.
- On Hyper-Focus: On The Data Exchange Podcast, Liu said that "we're kind of like hyper focused on this core Mission because we think there's a lot of depth". — The Data Exchange Podcast, Episode 187: Jerry Liu of LlamaIndex (Jul 2023), captions.
- On Existential Worry: On the Latent Space podcast, of the turn of the year before starting the company, Liu said "I was just existentially worried about whether or not this would actually be a company at all". — Latent Space: RAG Is A Hack - with Jerry Liu from LlamaIndex (Oct 2023), transcript.
Part 2: The "RAG is a Hack" Philosophy
- On the nature of retrieval: On the Latent Space podcast, Liu said: "RAG is basically just, just a hack, but it turns out it's a very good hack". — Latent Space: RAG Is A Hack - with Jerry Liu from LlamaIndex (Oct 2023), transcript.
- On context limits: On the Latent Space podcast, of the questions investors and friends asked, Liu said they were: "what happens if context windows get much bigger? What's the point of actually structuring data right in the right way? Right? Why don't you just dump everything into the prompt". — Latent Space: RAG Is A Hack - with Jerry Liu from LlamaIndex (Oct 2023), transcript.
- On Three Categories of Users: On The Data Exchange Podcast, Liu said that "there's probably broadly three categories of users". — The Data Exchange Podcast, Episode 187: Jerry Liu of LlamaIndex (Jul 2023), captions.
- On data privacy: On the Latent Space podcast, Liu said: "what RAG does is it increases like transparency, visibility into the actual documents". — Latent Space: RAG Is A Hack - with Jerry Liu from LlamaIndex (Oct 2023), transcript.
- On a High-Level API: On The Data Exchange Podcast, Liu said of what makes the package appealing: "it offers a very nice kind of high level API". — The Data Exchange Podcast, Episode 187: Jerry Liu of LlamaIndex (Jul 2023), captions.
- On Keeping the Package Light: On The Data Exchange Podcast, Liu said that "we want to keep the package as modular and light as possible". — The Data Exchange Podcast, Episode 187: Jerry Liu of LlamaIndex (Jul 2023), captions.
- On Democratization: On The Data Exchange Podcast, Liu said that "that's kind of what genre of AI has enabled is just like the broad democratization and usage of machine learning". — The Data Exchange Podcast, Episode 187: Jerry Liu of LlamaIndex (Jul 2023), captions.
- On Using Models Out of the Box: On The Data Exchange Podcast, Liu said that "you can kind of just use it out of the box as like a software package basically". — The Data Exchange Podcast, Episode 187: Jerry Liu of LlamaIndex (Jul 2023), captions.
- On Underlying Infrastructure: On The Data Exchange Podcast, Liu said that "we're one of the kind of like underlying infrastructure that fuels a wave of like a lot of these products". — The Data Exchange Podcast, Episode 187: Jerry Liu of LlamaIndex (Jul 2023), captions.
Part 3: Data Quality and Parsing
- On Retrieval Augmented Generation: On The Data Exchange Podcast, Liu said that "retrieval augmented generation uh is basically uh coupling the language model with uh some sort of retrieval model and you build a system around it". — The Data Exchange Podcast, Episode 187: Jerry Liu of LlamaIndex (Jul 2023), captions.
- On Embedding Lookup: On The Data Exchange Podcast, Liu said that "embedding based lookup is you know people have done this before it's very fast". — The Data Exchange Podcast, Episode 187: Jerry Liu of LlamaIndex (Jul 2023), captions.
- On Standard Chunking: On the What's AI podcast, Liu said that overall "people don’t think enough about the data pipeline for RAG, because they use some very standard chunking strategy, and this turns out to be suboptimal, and that’s actually why their RAG pipeline is failing". — What's AI Episode 25: Jerry Liu (Dec 2023), transcript.
- On Complexity Beyond the Current Stack: On The Data Exchange Podcast, Liu said that "we think there's a lot more complexity that needs to be explored beyond the current stack emerging today". — The Data Exchange Podcast, Episode 187: Jerry Liu of LlamaIndex (Jul 2023), captions.
- On Assistive Agents: On the DataFramed podcast, Liu said that at the moment "the best form factors for these AI agents, are very much assistive in nature and solve certain types of tasks and do it relatively well, instead of trying to bite off a ton of things end to end". — DataFramed #308: A Framework for GenAI App and Agent Development with Jerry Liu (Jun 2025), episode page with key quotes.
- On Quantifying Uncertainty: On the DataFramed podcast, Liu said: "Quantifying uncertainty means AI agents at least know the areas where they are unsure and require a little bit of human review.". — DataFramed #308: A Framework for GenAI App and Agent Development with Jerry Liu (Jun 2025), episode page with key quotes.
- On Citations: On the DataFramed podcast, Liu said: "You absolutely need citations, particularly if you're trying to process a large number of documents and each document is super long.". — DataFramed #308: A Framework for GenAI App and Agent Development with Jerry Liu (Jun 2025), episode page with key quotes.
- On Verifying Quickly: On the DataFramed podcast, Liu said of citations: "This gives the human the ability to verify things pretty quickly.". — DataFramed #308: A Framework for GenAI App and Agent Development with Jerry Liu (Jun 2025), episode page with key quotes.
- On Context Engineering: In a 2025 Microsoft for Startups interview, Liu said that I think "the key challenge that anyone is facing in using AI correctly is really the prompting and context engineering and making sure you're using the AI to actually help orchestrate meaningful knowledge workflows". — Microsoft for Startups: Jerry Liu from LlamaIndex talks about the future of AI agents, knowledge workflows & developer tools (2025), captions.
- On the Company Mission: In a 2025 Microsoft for Startups interview, Liu said that "our mission as a company these days really is to be the central platform for helping to automate knowledge work with AI agents". — Microsoft for Startups: Jerry Liu from LlamaIndex talks about the future of AI agents, knowledge workflows & developer tools (2025), captions.
Part 4: Moving Beyond Naive RAG
- On Flat Top-K Retrieval: On the What's AI podcast, Liu said that, when doing retrieval, "you don’t do retrieval just on like a flat top K of like the document text trunks". — What's AI Episode 25: Jerry Liu (Dec 2023), transcript.
- On Decomposing Questions: On the What's AI podcast, Liu said that, before launching a query, a system may rewrite or "decompose a complex question into smaller ones". — What's AI Episode 25: Jerry Liu (Dec 2023), transcript.
- On Bigger Chunk Sizes: On the What's AI podcast, Liu said "the world will probably move in the direction of bigger and bigger trunk sizes as these models get better". — What's AI Episode 25: Jerry Liu (Dec 2023), transcript.
Part 5: The Shift to Agentic Workflows
- On Agent Harnesses: On the Chain of Thought podcast in 2026, of the internals and mechanics of agents, Liu said "solidified into these like general agent harnesses". — Chain of Thought: The AI Framework Era Is Over: Why Context Is the Moat (Jun 2026), transcript.
- On Tailored Interfaces: On the Chain of Thought podcast in 2026, Liu said "by providing a tailored interface to solve a task, it just makes it way easier to kind of actually do the thing with AI". — Chain of Thought: The AI Framework Era Is Over: Why Context Is the Moat (Jun 2026), transcript.
- On Task Specifications: On the Chain of Thought podcast in 2026, of agents, Liu said "You give a task specification, they look at your context, they go do things, and this only gets better as the models get better". — Chain of Thought: The AI Framework Era Is Over: Why Context Is the Moat (Jun 2026), transcript.
Part 6: Productionizing LLM Applications
- On Prompt Engineering as a Career: On the What's AI podcast, Liu said "I don’t think that the prompt engineering, category itself will necessarily last as like a differentiating, like, career title". — What's AI Episode 25: Jerry Liu (Dec 2023), transcript.
Part 7: The "LLM as a CPU" Paradigm
- On a General Toolkit: On the Chain of Thought podcast in 2026, Liu said "we kind of think of ourselves as building like just a general toolkit plus all the surrounding metadata". — Chain of Thought: The AI Framework Era Is Over: Why Context Is the Moat (Jun 2026), transcript.
- On Abstractions and Production Context: On the Chain of Thought podcast in 2026, of the shift away from the framework layer, Liu said "the abstractions were solidifying, but the need for high-quality production context was still evolving". — Chain of Thought: The AI Framework Era Is Over: Why Context Is the Moat (Jun 2026), transcript.
- On Agents Needing a System of Record: On the Chain of Thought podcast in 2026, Liu said that people think a system of record will still be around because "agents need a way to actually store, act, and operate on data". — Chain of Thought: The AI Framework Era Is Over: Why Context Is the Moat (Jun 2026), transcript.
Part 8: Startup Building and Founder Lessons
- On a Steady North Star: On the Chain of Thought podcast in 2026, of LlamaIndex, Liu said "this overall North Star that we had, actually, I don't think it's changed too much, even if the product surface and you know, the ICP itself has changed". — Chain of Thought: The AI Framework Era Is Over: Why Context Is the Moat (Jun 2026), transcript.