# How We Solved Agent Building — Andrew Qu, Vercel

## Executive summary

Andrew Qu details Vercel's journey in solving agent building, evolving from initial manual data querying to complex, multi-step agent architectures. The major breakthrough was realizing that the core capability was not a clever toolset, but the simple file system (list, read, bash). This insight led to the development of Eve, a framework designed to bring file system conventions (like Next.js) to agents, enabling developers to build highly customized, production-ready agents with minimal effort.

## Key takeaways

- The File System Unlock: The most significant breakthrough was realizing that powerful agents could be built using a minimal set of tools—specifically, file system operations (list, read, run bash)—which models are already well-trained on. This approach allowed agents to explore and write where needed, leading to a massive increase in performance (the eval score basically doubled).
- Focus on Company-Specific Knowledge: While off-the-shelf agents are useful, the speaker argues that the most effective agents are those stuffed with deep, company-specific knowledge. These custom agents provide the most 'juice' and save significant time across various departments (e.g., legal, finance, data science).
- The Need for a Framework (Eve): To simplify agent development, Vercel built Eve, described as the 'Next.js for agents.' It adopts file system conventions, allowing developers to easily declare skills, tools, and channels, making the process of building and iterating on agents highly manageable.

## Technical details

- Agent Evolution Stages: The agent development progressed through three versions: 1) A 'mega prompt' using a pasted Snowflake schema for SQL generation; 2) A chain of narrowly scoped agents (e.g., query agent -> planning agent -> execution agent); and 3) A single agent managing its own state and memory, allowing it to reflect and correct errors.
- Vercel's Agent Infrastructure: Vercel built the AIDK (AI Developer Kit) to standardize model interfaces, allowing developers to switch out provider-specific code with a single line while maintaining model fallbacks, secure code execution, and durability.
- Agent Context Management: To improve efficiency, Vercel implemented a recurring job that distills common queries into roughly 100 'skills.' This ensures that new agent runs start with accumulated context, rather than starting from scratch.

## Practical implications

- Adopt a framework like Eve to simplify agent development by leveraging file system conventions, rather than building complex, bespoke agent chains.
- Prioritize building custom agents that incorporate deep, proprietary company knowledge over relying solely on general, off-the-shelf agent solutions.
- Utilize the concept of 'skills' to pre-process and inject common context into agent runs, improving efficiency and reducing the need for the agent to start from zero context.

## Topics

AI Agents, LLMs, Agentic Applications, File System Abstraction, Build Tools, Data Science Automation, Eve, AIDK, Claude Code / Opus 4.5

Source: https://www.youtube.com/watch?v=9dYcwOkpCE8