World's Fair 2025
The emerging skillset of wielding coding agents — Beyang Liu, Sourcegraph / Amp
Overview
This talk explores the evolving skillset required to effectively utilize coding agents, moving beyond early chatbot paradigms. It argues that the current era demands a new approach to agent interaction, focusing on empowering agents to perform tasks autonomously rather than micromanaging them. The core thesis is that mastering coding agents is a high-ceiling skill that will significantly enhance developer productivity, akin to learning a new programming language or editor.
Who should watch
- AI engineers and developers looking to leverage coding agents for complex tasks.
- Product Managers and builders interested in the future of AI-assisted software development.
- Individuals struggling with current AI coding tools and seeking more effective interaction patterns.
- Those curious about the shift from chatbot-style AI interactions to more agentic workflows.
Key takeaways
- The effectiveness of coding agents has rapidly advanced, necessitating new best practices that differ from earlier chatbot interactions.
- Effective agent use involves empowering them to make edits directly, rather than seeking approval for every step, to maximize developer efficiency.
- The trend is moving towards simpler, command-driven interfaces for agents, reducing the need for complex, feature-rich clients.
- Agents are becoming deeply coupled with specific LLMs, making model swapping more challenging than in the chatbot era.
- Pricing models for agents are shifting from fixed subscriptions to usage-based, reflecting their token consumption and the value of human time saved.
- The Unix philosophy of composable tools is becoming increasingly relevant for agent development, favoring command-line interfaces and modularity.
- Power users are writing longer, more detailed prompts, recognizing that LLMs are programmable with sufficient context, leading to better results.
- Agents can be used to accelerate onboarding, improve code reviews, and help developers understand complex codebases more deeply.
Notable quotes
*The number one mistake that people are using with coding agents right now is they're trying to use coding agents the same way they're using AI coding tools six months ago.*
*The human is still you're ultimately responsible for the code that you ship and you shouldn't view this as a human replacement. It's really a tool that you can wield to make yourself 10 100x more effective.*
Unofficial community note. Prefer the recording for nuance.