World's Fair 2025
Claude Code & the evolution of agentic coding — Boris Cherny, Anthropic
Overview
The talk by Boris Cherny from Anthropic discusses the rapid evolution of AI models in coding and the challenges in developing user experiences (UX) that keep pace. It highlights how programming itself has evolved through layers of abstraction and interface changes, from punch cards to modern IDEs. The core thesis is that while models are advancing exponentially, the product development for these AI coding tools is still in its early stages, with a focus on building unopinionated, minimal-viable products to learn and adapt to the unknown optimal UX.
Who should watch
- AI Engineers
- Product Managers
- Developers building with or evaluating AI coding assistants
- Those interested in the evolution of programming interfaces
- Anyone exploring how AI can integrate into software development workflows
Key takeaways
- AI models are improving at coding at an exponential rate, outpacing current product development for AI coding tools.
- The UX of programming has historically evolved through increasing levels of abstraction and new interfaces, from physical punch cards to graphical IDEs.
- Claude Code aims to provide a low-level, unopinionated interface to AI models, allowing users to integrate them into existing workflows without prescriptive UI.
- The tool supports various integrations, including terminal-based use, IDE plugins, GitHub app integration, and a programmatic SDK for custom solutions.
- Teaching AI models to use existing tools (like bash or MCP) is a key capability, requiring less complex integration than traditional IDE plugin development.
- Effective workflows with AI coding agents include having them explore and plan before coding, utilizing Test-Driven Development (TDD), and providing ample context for iteration.
- New features like Plan Mode in Claude Code allow users to approve AI-generated plans before execution, enhancing control.
- Providing more context to AI models, through mechanisms like a `quadm` file or slash commands, significantly improves their performance and utility.
Notable quotes
*The model is moving really fast. It's on exponential. It's getting better at coding very, very quickly.*
*We feel like there's so many more products that could be built for models that are this good at coding. And we're kind of building the bare minimum.*
*We want the model to be useful for you. We also want to get we want to be unopinionated and we want to get out of the way.*
Unofficial community note. Prefer the recording for nuance.