---
title: "Research"
description: "Applied research on code review accuracy, diffing, cross-repo context, and the agent architecture behind Baz."
canonical: "https://baz.ai/resources/research"
---

# Research

Applied research on code review accuracy, diffing, cross-repo context, and the agent architecture behind Baz.

- [Ship: Inference Routing for Agentic Code Review](https://baz.ai/resources/research/ship-inference-routing-for-agentic-code-review): We evaluated Martian's Ship as an execution layer beneath Baz's code review pipeline. The question was narrow: can we change how model calls are executed without changing the review behavior the agent architecture produces?
- [The vulnerability classes that don't come back](https://baz.ai/resources/research/vulnerability-classes-that-dont-come-back): Six classes of vulnerability slipped past traditional tooling for years. A study of 28 high-confidence findings from advanced security review shows why cyber-capable models now catch them at review time - before they ship.
- [Model vs. SAST](https://baz.ai/resources/research/model-vs-sast): Internal Baz research on how newer cyber-capable models are changing AppSec detection for pull-request review.
- [Can Cyber-Capable Models Fix AppSec?](https://baz.ai/resources/research/can-cyber-capable-models-fix-appsec): A technical analysis of why modern AppSec continues to miss critical vulnerabilities despite an abundance of tools and where upcoming cyber-capable models can help.
- [Improving cross-repo context for Baz code-reviewers](https://baz.ai/resources/research/improving-cross-repo-context-for-baz-code-reviewers): Why diff-local review breaks down at scale, and how cross-file, semantic, and cross-repo chaining improve review quality for structural bugs, org-wide security policies, and multi-service engineering teams.
- [The anatomy of code review accuracy](https://baz.ai/resources/research/the-anatomy-of-code-review-accuracy): How Baz rewired its AI code reviewer from a demo into a production system by treating retrieval, decomposition, extraction, and evaluation as first-class engineering problems.
- [Spec Agents as Design Reviewers](https://baz.ai/resources/research/spec-agents-as-design-reviewers): Spec Agents are our attempt to move that work into the agent layer. Assuming you built the required testing infrastructure, the goal is to have an agent that reads the spec, walks the product, and returns a clear yes or no with evidence.We’ve been running this model across a cohort of early users and watched how they behaved end to end.
- [Extending OpenTelemetry to Pinpoint Code Elements: Our Journey to Close the Gap](https://baz.ai/resources/research/extending-opentelemetry-to-pinpoint-code-elements-our-journey-to-close-the-gap): Enhancing OpenTelemetry's capabilities to pinpoint specific code elements responsible for system behaviors.
- [Ordering Files for AI-Assisted Development](https://baz.ai/resources/research/ordering-files-for-ai-assisted-development): Explore how ordering files and understanding code grammar can enhance AI-assisted development.