Devin, world's first AI software engineer, gets a major upgrade
What's the story
California-based Cognitive Labs's Devin AI has undergone substantial improvements, enhancing its capabilities across various dimensions.
The AI model is now faster, more accurate in code editing, more reliable in following instructions, and demonstrates improved independent decision-making.
The tool was launched in March this year and is touted as the "world's first fully autonomous AI software engineer."
Performance and reliability
Devin AI can now execute tasks faster and more accurately
One of the most notable advancements is the significant speedup in Devin's task execution.
According to internal evaluations, Devin can now complete common tasks, ranging from full-stack development to code refactoring and cleanup, up to 80% faster than before.
This improvement is attributed not only to faster individual actions but also to Devin's enhanced decision-making abilities, resulting in reduced looping and quicker problem-solving.
Improved collaboration
AI model now automatically responds to comments on pull requests
Devin's integration with GitHub and GitHub Mobile has been further refined. The AI model now automatically responds to comments on its pull requests (PRs), streamlining the collaboration process.
Additionally, you can provide feedback directly through chat, and Devin will automatically suggest new additions to its Knowledge base, a repository of tips, documentation, and facts that it can draw upon during its tasks.
Enterprise features
Custom Devins available for enterprise use
For organizations with stringent security and compliance requirements, Cognitive Labs offers Custom Devins and Devin in the Virtual Private Cloud (VPC).
Custom Devins are tailored for specific use cases and proprietary datasets, providing faster and more reliable performance within a narrower scope.
Devin in the VPC ensures that your data remains within the enterprise's controlled environment, addressing security concerns.
Team of AIs
A team of Devins can now work in parallel
The introduction of MultiDevin allows managers to create and manage a team of Devins working in parallel.
This feature is particularly beneficial for large backlogs and repetitive, isolated tasks such as migrations and refactors. By distributing tasks across multiple Devins, productivity can be significantly increased.
The new features further solidify Devin's position as a valuable AI tool for code generation and automation.