NavieAI breaks the glass ceiling on AI-powered software development by topping the AI software engineer leaderboard, solving issues in record time at the lowest recorded cost
BOSTON, MA, June 26, 2024 / AppMap, the leading AI-powered developer-centric observability platform, today announced the launch and general availability of Navie, a first-of-its-kind runtime aware AI-powered coder that is now one of the leaders on the AI software engineering leaderboard at a fraction of the cost. SWE Bench is as much about efficiency and stability, as it is about raw power or ability to spend millions of tokens for issue solutions. AppMap Navie leverages deep runtime data generated inside the code editor – such as traces, API calls, sequence diagrams, and performance metrics – to provide a new AI-powered coding solution at less than $1 per issue, 20x less than other AI software engineer agents. AppMap Navie is available today in the VSCode and JetBrains marketplaces.
“We’re forging a movement in the application of AI for the most challenging software and real-world engineering tasks, the issues that consume 40% developer time. But to make it practical for use today, AI coding agents need to be cost effective,” says Elizabeth Lawler, CEO of AppMap. “There are AI coding approaches using huge token limits and lots of iterations, working harder but not smarter and running an open meter on the AI bill. Navie’s deep runtime data context means AI can now excel at complex issues with better context. Developers get accurate code suggestions that will work at runtime, which means less rework. Budget is top of mind for everyone and Navie is a cost effective AI approach anyone can adopt and use in enterprise environments.”
Boosting developer productivity when working on the most challenging development tasks has been elusive to even the most widely adopted AI coding assistants. According to research from Microsoft and others, AI coding assistants provide a widely reported 10% productivity gain. The challenging tasks where Navie excels include application modernization and refactoring tasks, deep troubleshooting, and improving application performance and runtime security which account for up to 40% of developer effort. These are some of the most costly engineering activities where AIs with the least amount of context struggle. With Navie, developers now have an AI partner in the analysis, code design and implementation of AI code suggestions for these complex issues at the most cost-effective solution of any AI agent in the market today. Incorporating the application’s runtime behavior as context with code assistance saves developers time, effort, and real money.
AppMap is known for its open-source and enterprise friendly architecture, and remains consistent to that vision with Navie. AppMap uses a first-of-its-kind open context approach which allows developers to see and inspect and edit the context used in the prompt to improve accuracy. Navie leverages both agentic and human-in-the-loop methods to meet developers where they are in their AI code assistance journey.
AppMap data remains local to your environment and AppMap Navie also runs locally or in CI with a bring your own AI backend model approach. Navie works with a variety of LLMs including OpenAI, Mixtral, and more. Navie is available via the AppMap extension for JetBrains and VSCode. Learn more at https://appmap.io/product/appmap-navie
