Inspiration

The world is built on a fragile foundation. Trillions of dollars in global infrastructure run on decades-old, legacy code. We built Gemini Rosetta to usher in the "Action Era" of AI—where agents autonomously resurrect software archaeology rather than just chatting about it.

What it does

Gemini Rosetta is an autonomous modernization platform that successfully refactored 90s-era legacy code—including the 1993 DOOM Engine and COBOL Banking Systems—into enterprise-grade modern Python in a specialized autonomous "Marathon Mode" sprint.

How we built it

  • Technical Foundation: Built and validated within the Antigravity environment, utilizing Gemini 3's native orchestration capabilities.
  • 1M Context Window: Simultaneously held entire repository states and dependency graphs in active memory.
  • Verification: Integrated Vibe Engineering™ with neon dashboards and neural audio feedback via gTTS.

Challenges we ran into

The biggest hurdle was maintaining Logic Parity while refactoring "GOTO hell" procedural patterns and ancient bit-level arithmetic hacks into type-safe Python logic.

Accomplishments we're proud of

  • Financial Integrity: Successfully migrated bank.cob with verified ISO-8601 compliance.
  • Physics Resurrection: Transformed 1993 procedural movement into frame-rate independent Newtonian physics.

What's next

Scaling the Antigravity Engine for OS-level migrations and autonomous self-healing codebases.

Built With

  • antigravity
  • gemini-3-api
  • gtts
  • python
  • streamlit
Share this project:

Updates