Anthropic’s Code with Claude showed off coding’s future—whether you like it or not

2026-05-21 · Source: MIT Technology Review · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning · Depth: Intermediate, medium

Summary

Anthropic's "Code with Claude" event in London on May 19 showcased a significant shift in software development, with nearly half of attending developers having shipped pull requests entirely written by Claude, often without human review. This reflects a new paradigm where LLM-powered tools like Claude Code are increasingly generating production code, with Anthropic claiming "Most software at Anthropic is now written by Claude." Recent updates, Claude 4.6 and 4.7 (February and April), have enhanced its capabilities. Anthropic's goal is full automation, where Claude self-corrects errors and "prompts itself," exemplified by the new "dreaming" feature that allows agents to learn from past tasks. However, external concerns include the burden of reviewing AI-generated code, potential skill degradation, and increased vulnerability to attacks. Anthropic maintains that traditional best practices still apply, noting Claude is "as good as a midlevel engineer" but aims to eventually build itself.

Key takeaway

For software engineers integrating AI coding tools, recognize that while Claude Code can generate significant code, shipping unreviewed AI-produced pull requests introduces substantial risk. You must prioritize rigorous code review and adapt your role to focus on system design, troubleshooting, and validating AI outputs. This shift demands maintaining traditional best practices to prevent security vulnerabilities and skill degradation.

Key insights

LLMs are rapidly automating software development, shifting human roles from writing to oversight, despite emerging concerns.

Principles

• AI should self-correct and prompt itself.
• Prioritize "getting out of Claude's way."
• Traditional dev best practices remain crucial.

Method

Claude Code's "dreaming" feature enables agents to write and consolidate notes, learning from past tasks and errors to continuously improve code generation for a specific codebase.

In practice

• Delegate pull request generation to Claude.
• Utilize "dreaming" for codebase-specific learning.
• Reinforce existing code review standards.

Topics

• Anthropic
• Claude Code
• LLM Code Generation
• Software Development Automation
• AI Code Review
• AI Safety

Best for: CTO, VP of Engineering/Data, AI Architect, Software Engineer, AI Engineer, Director of AI/ML

Related on AIssential

• Anthropic launches AI-powered Code Review for Claude Code — AI-powered code review can significantly accelerate development cycles by focusing on logical errors in AI-generated code.
• The End of “Software Engineer”: Why Anthropic’s Claude Code Creator Says the Title Will Disappear… — AI coding agents like Claude Code will fundamentally alter the software engineering profession, potentially eliminating the traditional title.
• Claude Code, A Technical Guide for the Business Executive — Claude Code is transitioning from a developer tool to a secure, full-stack agentic OS, leveraging whole-project context.
• Claude Code's product lead talks usage limits, transparency, and the "lean harness" — Anthropic prioritizes agile, user-driven development for Claude Code, betting on rapid model advancements over fixed roadmaps.
• Anthropic Introduces Routines for Claude Code Automation — Anthropic's Routines automate coding workflows via scheduled, API, or event-driven triggers on cloud infrastructure.
• Live blog: Code w/ Claude 2026 — Anthropic is enhancing developer productivity and model autonomy through advanced agentic capabilities and platform integrations.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by MIT Technology Review.