Developers Unveil 39 AI Projects at JetBrains Codex Hackathon, Top Prize Goes to 'Hyperreasoning' Agent

San Francisco, CA – In a weekend-long event that underscored the rapid evolution of AI-powered development tools, the inaugural JetBrains Codex Hackathon produced 39 IDE-native AI projects, with the top prize awarded to a solo builder who created a reasoning-optimized coding agent. The hackathon, held earlier this month, drew 443 applicants and culminated in six finalists, highlighting a growing community focus on smarter, safer, and more transparent AI within the integrated development environment.

“The level of ambition and technical rigor we saw was remarkable,” said Maria Petrova, a JetBrains spokesperson. “These projects are not just proof-of-concepts—they are shaping the future of how developers interact with AI in their daily workflow.”

Background: A Weekend of IDE-Native Innovation

The JetBrains Codex Hackathon invited developers to build AI-powered tools directly within the IntelliJ Platform. Approximately half of the 39 completed projects were IDE plugins or tools built using the IntelliJ Platform SDK, reflecting a deep integration with the development environment rather than external agents.

Judges, including leading technologists who volunteered their Sunday, evaluated projects based on innovation, usability, and long-term impact. The winning entry, hyperreasoning, built by Aditya Mangalampalli, stood out for its unique approach.

What This Means for the Future of AI in the IDE

The hackathon’s results signal a shift from speed-focused AI code generation toward more intelligent and transparent systems. Finalists prioritized correctness, safety, context awareness, and reviewability over raw output speed. As one judge noted, “The projects that impressed us most were those that gave developers more visibility into what their agents were doing, more guardrails, and a clearer sense of when to trust the output.”

This trend suggests that the next wave of IDE-native AI will emphasize collaborative reasoning and reliability, making the tool an trusted partner rather than a black-box code generator.

Key Highlights from the Hackathon

• 39 projects completed, with 6 finalists selected.
• Top Prize: hyperreasoning – a solo-built coding agent that evaluates reasoning paths before generating code.
• Second Place: Scopecreep by Bhavik Sheoran, Kenneth Ross, Roman Javadyan, and Joon Im.
• Third Place: mesh-code by Ayush Ojha, Coco Cao, Kush Ise, and AL DRAM.
• Two of the six finalists were solo builders—a rare achievement in team-based hackathons.
• Roughly half the submissions were JetBrains plugins or IDE-native tools built on the IntelliJ Platform SDK.

“One person, one laptop, 24 hours—and a coding agent that decides which reasoning paths are worth exploring before it generates a single line of code,” Petrova added, describing the winning project. “That’s the kind of breakthrough the hackathon format is designed to accelerate.”

For more details on individual projects, stay tuned for follow-up blog posts from JetBrains. The company plans to spotlight each finalist, sharing the stories behind their innovations.

Jump to Background | Jump to What This Means