04 Feb 2026

1h 56m

Infinite Code Context: AI Coding at Enterprise Scale w/ Blitzy CEO Brian Elliott & CTO Sid Pardeshi

"The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis

Blitzy's CEO Brian Elliott and CTO Sid Pardeshi discuss autonomous software engineering, focusing on how their platform achieves AGI-type effects using non-AGI LLMs. They emphasize the importance of orchestrating LLMs within long-running complex systems, highlighting the limitations of standalone LLMs, such as context window constraints and tool selection challenges. Blitzy employs dynamic agent generation, prompts written by other agents, and iterative planning to overcome these limitations. They use a hybrid approach combining relational and semantic understanding, building a programming language-agnostic knowledge graph to deeply understand code relationships. The discussion also covers model evaluation strategies, the role of taste in assessing model performance, and the labor market implications of AI in software engineering, predicting a shift towards junior engineers skilled in AI.

Outlines

Part 1: Introduction, Blitzy’s Vision

Part 2: LLM Limitations, Context Engineering

Part 3: Technical Implementation, Parallel Environments

Part 4: Dynamic Architecture, Evaluation

Part 5: System Steering, Knowledge Management

Part 6: Human-AI Collaboration, Workflow

Part 7: Model Selection, Memory Strategies

Part 8: Development Fundamentals, Business Value

Part 9: Onboarding, Testing, Future Outlook

Continue

Preview

How to Get Rich: Every EpisodeNaval

Infinite Code Context: AI Coding at Enterprise Scale w/ Blitzy CEO Brian Elliott & CTO Sid Pardeshi

"The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis

Part 1: Introduction, Blitzy’s Vision

00:00Blitzy's AI-Driven Approach to Enterprise Software Modernization

Blitzy's AI-Driven Approach to Enterprise Software Modernization

00:59Autonomous Software Engineering: Blitzy's Architecture and Capabilities

Autonomous Software Engineering: Blitzy's Architecture and Capabilities

02:37Blitzy's AGI Perspective: Orchestration Over Standalone LLM Capabilities

Blitzy's AGI Perspective: Orchestration Over Standalone LLM Capabilities

Part 2: LLM Limitations, Context Engineering

04:35LLM Limitations: Context Window Depreciation and Tool Management

LLM Limitations: Context Window Depreciation and Tool Management

06:04Designing Systems for AGI-Type Effects: Harnessing LLM Limitations

Designing Systems for AGI-Type Effects: Harnessing LLM Limitations

07:30Schematizing Code: Building Relational Understanding for Autonomous Development

Schematizing Code: Building Relational Understanding for Autonomous Development

09:21Knowledge Graphs: A Deep Relational Understanding of Code

Knowledge Graphs: A Deep Relational Understanding of Code

Part 3: Technical Implementation, Parallel Environments

11:32Ingesting Enterprise Code: Building a Programming Language Agnostic Understanding

Ingesting Enterprise Code: Building a Programming Language Agnostic Understanding

12:35Building and Running Applications: Understanding Relationships in Production

Building and Running Applications: Understanding Relationships in Production

14:00Parallel Application Instances: QA and Recursive Correction Loops

Parallel Application Instances: QA and Recursive Correction Loops

15:08Implementation Challenges: Building Instructions and Modernizing Technology Stacks

Implementation Challenges: Building Instructions and Modernizing Technology Stacks

16:18Build Sophistication: Automating Windows Application Builds

Build Sophistication: Automating Windows Application Builds

Part 4: Dynamic Architecture, Evaluation

21:33Dynamic Design: Generating Agents Just-In-Time for Ever-Improving LLMs

Dynamic Design: Generating Agents Just-In-Time for Ever-Improving LLMs

23:34Dynamic Systems: Referencing the Latest Prompting Instructions

Dynamic Systems: Referencing the Latest Prompting Instructions

24:26Evals: Mapping Onto the Real World with Taste

Evals: Mapping Onto the Real World with Taste

25:28Large-Scale Evaluation: Taste and Technical Design

Large-Scale Evaluation: Taste and Technical Design

26:08Blitzy's Technical Taste: Instantiating Technical Taste into the Outcome

Blitzy's Technical Taste: Instantiating Technical Taste into the Outcome

27:25Evaluating New Models: Tracing Agentic Interactions

Evaluating New Models: Tracing Agentic Interactions

Part 5: System Steering, Knowledge Management

28:07Steering the System: Dynamically Addressing Instances

Steering the System: Dynamically Addressing Instances

29:44Algorithmic Steering: Tweaking Algorithms for the Right Outcome

Algorithmic Steering: Tweaking Algorithms for the Right Outcome

31:57Knowledge Graphs and RAG: Hybrid Sources of Truth

Knowledge Graphs and RAG: Hybrid Sources of Truth

32:53Semantic Matching: An Efficiency Search Mechanism

Semantic Matching: An Efficiency Search Mechanism

33:53Guiding Models: Making Judgement Calls on When to Stop Searching

Guiding Models: Making Judgement Calls on When to Stop Searching

Part 6: Human-AI Collaboration, Workflow

35:05Structuring Requests: Creating the Right Interface Experience

Structuring Requests: Creating the Right Interface Experience

36:42Human Approval: Clarity on What Needs to Be Read

Human Approval: Clarity on What Needs to Be Read

37:20System Capabilities: Generating In-Depth Implementation Plans

System Capabilities: Generating In-Depth Implementation Plans

38:31Recognizing Failure: Building Independent Evaluation Systems

Recognizing Failure: Building Independent Evaluation Systems

40:03Project Guides: Identifying Functions That Need Human Help

Project Guides: Identifying Functions That Need Human Help

Part 7: Model Selection, Memory Strategies

41:20Model Scouting Report: Leveraging Multiple Model Families

Model Scouting Report: Leveraging Multiple Model Families

42:06Model Selection: Dynamic Algorithms and Validation

Model Selection: Dynamic Algorithms and Validation

43:39Fine-Tuning vs. Memory: A Bet on Improved Models

Fine-Tuning vs. Memory: A Bet on Improved Models

44:21Enterprise Memory: Sustaining Actions of the Best People

Enterprise Memory: Sustaining Actions of the Best People

45:15Long-Term Memory: A Problem to Be Solved at the System Layer

Long-Term Memory: A Problem to Be Solved at the System Layer

46:32Context Management: Storing Memory and Preferences Based on Actions

Context Management: Storing Memory and Preferences Based on Actions

47:35Local Specificity: Memory on Enterprise Code Bases

Local Specificity: Memory on Enterprise Code Bases

48:46Bifurcating Memory: Global Truths vs. Local Contextual Decisions