Software engineering
Explain an unfamiliar codebase, propose unit tests and draft documentation inside an authorized development workflow.
Software, technology & cybersecurity
Build, inspect and improve the tools.
American Super Intelligence will give software teams and independent developers models to build upon and technical education to use them well. Coding capabilities will support prototypes, testing, debugging and application modernization.
A proposed American Super Intelligence use case.
Explore model development and release
Practical uses
AI products
Evaluate an open model for a product, design task-specific evaluations and test adaptation or retrieval with permitted data.
Defensive security
Organize authorized security findings, explain secure coding patterns and practice incident analysis in controlled exercises.
AN EXAMPLE IN PRACTICE
A software team explores a coding assistant for a documented internal library.
- Engineers create representative tasks, approved examples and tests before comparing models.
- They inspect generated code for correctness, license implications and security, then measure review and rework.
- A controlled rollout keeps code review, access restrictions, monitoring and rollback under the team’s authority.
THE HUMAN CAPABILITY
What people will learn
- Write evaluations that can fail a plausible but wrong answer.
- Protect source code, credentials and customer data.
- Understand model updates, dependency risks and recovery.
THE DELIVERY PATH
How access will work
Teams will run and adapt Liberty OSS under its release license. Liberty OSS Code will support local development, while Liberty Frontier Code will be the hosted specialist track through approved service access; evaluated multimodal variants will connect visual references to code. Production infrastructure and access controls will remain the team’s responsibility.
Understand the model familiesHow success will be measured
Track tested correctness, defects, review effort and maintainability alongside development time and resource use.
Requirements for this use
AI-generated code is not automatically correct, secure or cleared for reuse. Cybersecurity work requires authorization; the public service is not an unrestricted tool for attacking systems. Require review and controlled, explicitly authorized execution of generated code; do not grant repository or production access merely because the model can propose a change.
This scenario describes the proposed experience. Product evaluation and the responsible organization’s approval will precede deployment.