AI Briefing
KO

Devin Launches 'Code Scans' Based on Agentic MapReduce to Automate Code Quality and Performance Improvements

·2026.09.16 09:00

Key point

Devin has released the 'Code Scans' feature, which utilizes an Agentic MapReduce architecture to investigate entire codebases and generate pull requests.

Details

Devin has released the Code Scans feature based on the Agentic MapReduce architecture, automating various engineering goals beyond security scans, including performance, databases, test coverage, and code quality.

Architecture and How It Works

The architecture, built for Devin Security Swarm, has been expanded to parallelize large-scale codebase investigations.

  • Plan: Analyzes the repository to define code identification rules relevant to the goal.
  • Shard: Splits the code into focused batches according to the rules.
  • Map: Parallel Devin agents investigate each batch and report findings.
  • Reduce: A final agent consolidates findings, removes duplicates, and prioritizes them to generate a report and PR.

Performance and Demonstrated Results

In pre-tests, it recorded a 96% PR merge rate across multiple repositories, with an estimated savings of over 700 hours of engineering time during a short test period.

  • Rust Compilation Optimization (Dioxus): Reduced clean debug build time by 64%, from 58.6 seconds to 21.0 seconds, through dependency and build configuration optimization.
  • SEO Improvement: devin.ai's Ahrefs health score rose from 87 to 92, and slow pages decreased by 73%. On cognition.com, all missing image alt text was fixed.

Supported Scan Types

Users only need to define the goal using the /scan command. Supported types are as follows:

  • Performance: Detects slow paths, redundant calculations, and missed caching opportunities
  • Database: Inefficient data access patterns such as N+1 queries and unnecessary round trips
  • Test Coverage: Identifies critical flows and edge cases not covered by tests
  • Dead Code: Unused functions/modules, deprecated feature flags, and provides justification for removal
  • SEO Optimization: Missing metadata, incorrect canonical URLs, structured data gaps, etc.
  • Custom: Define your own patterns such as team coding standards, accessibility checklists, and migration requirements

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.