An open-source governance and verification layer for AI-assisted software engineering — every AI-generated code change runs in an isolated sandbox and is cryptographically verified before a human decides whether to apply it.
An open-source governance and verification layer for AI-assisted software engineering — every AI-generated code change runs in an isolated sandbox and is cryptographically verified before a human decides whether to apply it.
Project Details
Updated 07/01/26 · Edited by orgThe One Shot is an open-source AI governance framework for safely executing AI-generated software changes.
Current AI coding systems can propose code modifications, but they do not reliably ensure that those changes are correct, reproducible, or safe before execution. This creates risk when AI-generated outputs are applied directly to real systems.
The One Shot addresses this by separating generation from execution. Every AI-generated change is executed inside an isolated sandbox, validated through deterministic execution, and recorded with verifiable execution evidence before any human approval step.
The system replaces trust in model reasoning with trust in reproducible execution outcomes.
The project will deliver a complete open-source reference implementation, including:
- deterministic execution pipeline
- sandboxed execution environment
- verification and trace generation system
- reproducible execution logs
- documentation and local runnable setup
The framework is model-agnostic and designed to integrate with existing AI coding tools without requiring changes to underlying models.
Theory of Impact
Updated 07/19/26 · By grantmaking.aiAs AI systems become more capable of generating software, the primary risk shifts from generation quality to execution trust and validation.
The One Shot reduces deployment risk by ensuring that every AI-generated change is executed in an isolated environment before it can affect real systems. Each execution produces a deterministic, reproducible record that can be independently verified.
This creates an auditable execution trail that replaces reliance on model explanations with objective execution evidence.
The system strengthens AI safety infrastructure by ensuring that improvements in model capability are matched with improvements in execution governance and verification.
People
Updated 07/19/26 · By grantmaking.aiTeam Member
Funding Details
Track Record
Correct Track Record wording
Track Record:
This project builds on prior experience developing automation workflows and structured execution systems in UiPath-based environments. A working hackathon project demonstrating applied automation logic is available in the repository below.
Repository:
https://github.com/itz1508/hackathon-uipath - recent submitted on 06/29/26
Audisor solves the problem of unverified AI-generated configuration repairs reaching production. AI can propose fixes fast, but enterprises need proof that one exact result is safe and explicit human authority before applying it.
https://github.com/itz1508/hackathon-band - submitted 2 weeks ago
Built: Five specialized Band agents collaborate through structured context to resolve software issues — with one controlled retry when requirements are not met.
Private comment. Only shown to approved funders and grant reviewers.