Closing the loop is the secret to effective AI coding: AI agents must be able to test and validate their own work through compilation, linting, and running tests. This is why AI excels at coding but struggles with creative writing - code can be validated, prose cannot.

Closing the loop is the secret to effective AI coding: AI agents must be able to test and validate their own work through compilation, linting, and running tests. This is why AI excels at coding but struggles with creative writing - code can be validated, prose cannot.

Code reviews and PRs are becoming obsolete: Peter prefers 'prompt requests' over pull requests and reads prompts more than code. He rewrites most community PRs with his agent rather than reviewing them line-by-line, focusing instead on architecture and product vision.

Code reviews and PRs are becoming obsolete: Peter prefers 'prompt requests' over pull requests and reads prompts more than code. He rewrites most community PRs with his agent rather than reviewing them line-by-line, focusing instead on architecture and product vision.

AI development requires a fundamental shift in thinking: Developers must transition from writing code to being 'builders' or 'architects' who design systems, make taste decisions, and orchestrate multiple AI agents working in parallel. The mundane plumbing is automated away.

AI development requires a fundamental shift in thinking: Developers must transition from writing code to being 'builders' or 'architects' who design systems, make taste decisions, and orchestrate multiple AI agents working in parallel. The mundane plumbing is automated away.

Companies could operate with 30% of current staff using AI: The future requires fewer but more senior engineers with high agency, product vision, and system understanding rather than traditional coding skills. This will require companies to completely restructure how they work.

Companies could operate with 30% of current staff using AI: The future requires fewer but more senior engineers with high agency, product vision, and system understanding rather than traditional coding skills. This will require companies to completely restructure how they work.

Working in the open and building with AI is the new competitive advantage: For developers, especially new grads, the key differentiator is demonstrating ability to build things using AI agents effectively, sharing prompts and processes publicly, and developing deep curiosity about system architecture.

Working in the open and building with AI is the new competitive advantage: For developers, especially new grads, the key differentiator is demonstrating ability to build things using AI agents effectively, sharing prompts and processes publicly, and developing deep curiosity about system architecture.

The creator of Clawd: "I ship code I don't read"

Name: The creator of Clawd: "I ship code I don't read"
Uploaded: 2026-02-04T19:17:05.429443
Duration: 1 h 54 min 5 s
Description: Peter Stainberger, creator of PSPDFkit and Clawbot, discusses his journey from building a PDF framework used on over 1 billion devices to burning out, taking a 3-year break from tech, and returning to build software in a radically different way using AI agents. He explains how he now ships code with

By The Pragmatic Engineer

Watch on YouTube (1:54:05)

Overview

Peter Stainberger, creator of PSPDFkit and Clawbot, discusses his journey from building a PDF framework used on over 1 billion devices to burning out, taking a 3-year break from tech, and returning to build software in a radically different way using AI agents. He explains how he now ships code without reading most of it, merges 600 commits in a day, and why traditional software development practices like code reviews and PRs are becoming obsolete in the AI-assisted development era.

Key Takeaways

Closing the loop is the secret to effective AI coding: AI agents must be able to test and validate their own work through compilation, linting, and running tests. This is why AI excels at coding but struggles with creative writing - code can be validated, prose cannot.
Code reviews and PRs are becoming obsolete: Peter prefers 'prompt requests' over pull requests and reads prompts more than code. He rewrites most community PRs with his agent rather than reviewing them line-by-line, focusing instead on architecture and product vision.
AI development requires a fundamental shift in thinking: Developers must transition from writing code to being 'builders' or 'architects' who design systems, make taste decisions, and orchestrate multiple AI agents working in parallel. The mundane plumbing is automated away.
Companies could operate with 30% of current staff using AI: The future requires fewer but more senior engineers with high agency, product vision, and system understanding rather than traditional coding skills. This will require companies to completely restructure how they work.
Working in the open and building with AI is the new competitive advantage: For developers, especially new grads, the key differentiator is demonstrating ability to build things using AI agents effectively, sharing prompts and processes publicly, and developing deep curiosity about system architecture.