--- title: "Blog" description: "Perspectives on AI code review, review agents, and shipping quality software, from the team building Baz." canonical: "https://baz.ai/resources/blog" --- # Blog Perspectives on AI code review, review agents, and shipping quality software, from the team building Baz. - [Baz pricing is getting simpler](https://baz.ai/resources/blog/simpler-pricing): Starting November 3, 2026: Pro includes 50 reviews per developer each month, then $1 per review. It applies from your next billing cycle. - [Building an AI code review layer for your software factory: four lessons from a CTO](https://baz.ai/resources/blog/ai-code-review-layer-software-factory): Baz’s CTO shares four lessons for building an AI code review layer with shared context, continuous learning, and plan review before implementation. - [Coding is dead. Long live software engineering](https://baz.ai/resources/blog/coding-is-dead-long-live-software-engineering): Baz CTO Nimrod Kor explains how AI coding agents reshape software engineering, from shared skills and planning to code review and continuous verification. - [It's Not Just the Model, Dummy: How We Rebuilt Our Code Review Harness](https://baz.ai/resources/blog/its-not-just-the-model-how-we-rebuilt-our-code-review-harness): Recall up 31% (from 51% to 67%), 60% faster, half the cost. And no, it wasn't just a model upgrade. - [Agent code is the new attack surface](https://baz.ai/resources/blog/agent-code-is-the-new-attack-surface): Teams are shipping agents through SDKs, plugins, skills, hooks, MCP servers, and CI workflows. Baz now runs a dedicated security reviewer on every Change that touches agent code. - [Skill Maintainer: keeping your instruction files true](https://baz.ai/resources/blog/skills-maintainer-keeping-your-instruction-files-true): Your CLAUDE.md was accurate the day it was written. Skill Maintainer checks each rule against the code and the pull requests that came after, and opens a PR when a rule no longer holds. - [Awesome Reviewers, rebuilt: instructions across AI and infrastructure domains](https://baz.ai/resources/blog/awesome-reviewers-rebuilt-around-ai-and-infrastructure-domains): We rebuilt awesomereviewers.com around the domains people actually build in, gave every instruction a date derived from its source discussions, and made the whole corpus fetchable as raw text. Here is the refresh, and the road that led to it. - [SAST-inside: expanding the hypothesis space of agentic security review](https://baz.ai/resources/blog/sast-inside-expanding-the-hypothesis-space-of-agentic-security-review): SAST-inside adds a second source of vulnerability hypotheses to Baz Advanced Security. Specialized static checks nominate candidate conditions in the changed code, and the agent investigates each one against the repository before it becomes a finding. - [Spec Reviewer now runs your application before reviewing it](https://baz.ai/resources/blog/spec-reviewer-now-runs-your-application-before-reviewing-it): Spec Reviewer can now launch the application from the Change branch inside an isolated sandbox, open it in a real browser, and validate the running implementation against the specification. Private application review, which runs that sandbox inside your own Kubernetes environment, is available as a limited technical preview. - [Bring AI Code Review to Azure DevOps with Baz](https://baz.ai/resources/blog/announcing-azure-devops-support): Azure DevOps AI code review is live. Baz connects directly to your Azure DevOps organization and reviews every pull request raised in Azure Repos. - [How to know when an autonomous merge agent is working](https://baz.ai/resources/blog/how-to-know-when-an-autonomous-merge-agent-is-working): Merger Agent Stats tracks evaluated PRs, ready-to-merge rate, merged-without-human-input rate, outcomes, and throughput, so trust in Merger is backed by numbers. - [Code review needs an inbox](https://baz.ai/resources/blog/code-review-needs-an-inbox): PR Inbox adds a My pull requests tab to Changes, with Waiting for review and My open PRs sections, so developers stop hunting repo lists to find what needs their attention. - [Usage-based AI needs budget controls](https://baz.ai/resources/blog/usage-based-ai-needs-budget-controls): Account settings now show live usage, let admins set spend limits, and break down Fixer costs, so usage-based pricing comes with real guardrails. - [Approve the plan before the code exists](https://baz.ai/resources/blog/approve-the-plan-before-the-code-exists): Plans generated by Planner now go through a dedicated approval workflow, with versions, section comments, and Mermaid diagrams, before implementation starts. - [Agent work should be inspectable](https://baz.ai/resources/blog/agent-work-should-be-inspectable): Baz Sessions now trace Planner, Merger, Spec Reviewer, and full repo security scans, not just PR reviews. - [Security review should not wait for the next pull request](https://baz.ai/resources/blog/security-review-should-not-wait-for-the-next-pull-request): Full codebase security scans find risk already sitting in the repository, tracked as Sessions with severity-tagged findings and fixer PRs. - [Bringing agentic code review to Azure Repos](https://baz.ai/resources/blog/bringing-agentic-code-review-to-azure-repos): Baz now supports admin-scoped project and repository selection for Azure Repos, so Azure DevOps teams get the same agentic review workflows as GitHub. - [Measuring the review loop before and after Baz](https://baz.ai/resources/blog/measuring-the-review-loop-before-and-after-baz): Engineering Impact's pickup time and time-to-merge graphs now support a pre-Baz baseline, so teams can see review velocity change, not just its current state. - [Session Logs: the execution record for Baz agents](https://baz.ai/resources/blog/session-logs-the-execution-record-for-baz-agents): Session Logs give every Baz run a structured execution record. They show how a run was triggered, where it executed, its current state, major completed stages, outcome, and cost. Instead of exposing raw internal logs, Baz presents a timeline of product events so teams can inspect reviewer runs, fixer runs, scheduled scans, benchmark runs, and future Codex or Claude Code workflows from one place. - [Auto Fix and Auto Resolve Cut PR validation time](https://baz.ai/resources/blog/auto-fix-and-auto-resolve-cut-pr-validation-time): Auto Approve issues a formal PR approval once every Baz comment is resolved and the corresponding fixes are verified against linters and CI. Auto Resolve closes review threads only after the system correlates a verified transform to the original comment and confirms required checks are green. Both features are governed by admin-configurable rules, supported-language lists, and confidence thresholds, and they emit immutable metadata, reverts, and telemetry to help teams reduce manual validation and advance toward an autonomous codebase. - [Advanced Security Reviewer: security review for the cyber-capable model era](https://baz.ai/resources/blog/advanced-security-reviewer-security-review-for-the-cyber-capable-model-era): Baz Advanced Security Reviewer uses cyber-capable models, Agent Harness, and Context Broker to find exploitable security issues in code changes with real codebase context. - [cyber.md: AI-native posture that speaks agent](https://baz.ai/resources/blog/cyber-md-ai-native-posture-that-speaks-agent): cyber.md is a proposed Markdown-based security posture file that helps coding agents preserve security intent during normal development workflows. Rather than acting as a vulnerability report or attack playbook, cyber.md captures protected assets, trust boundaries, coding invariants, defensive patterns, and testing expectations in a format agents can consume safely and contextually. - [Getting Ready for Cyber-Capable Models Like Mythos](https://baz.ai/resources/blog/getting-ready-for-cyber-capable-models-like-mythos): Our short take on why cyber-capable models like Mythos will disrupt classic code management. We argue that the real shift is from diff inspection to consequence analysis: models now reason across migrations, shared types, schemas, and interfaces to surface logical bugs, breaking changes, and downstream impact. - [SRE Agent: Baz listens to production and opens the fix PR for you](https://baz.ai/resources/blog/sre-agent-baz-listens-to-production-and-opens-the-fix-pr-for-you): We just shipped a new Baz coding agent, SRE, that listens to traces and logs from APM (Application Performance Monitoring), maps them back to the exact code that caused them, and opens concise fix pull-requests. - [A complete guide to building autonomous repos in GitHub](https://baz.ai/resources/blog/a-complete-guide-to-building-autonomous-repos-in-github): This guide shows how to introduce autonomy into pull request workflows safely - covering configuration, guardrails, and staged authority so teams can automate without creating irreversible risk. - [Baz #1 in Precision in independent Code Review Bench](https://baz.ai/resources/blog/baz-1-in-precision-in-independent-code-review-bench): Baz #1 in Precision in independent Code Review Bench - [Baz CLI: Interactive code reviews](https://baz.ai/resources/blog/baz-checkout-live-local-and-intelligent-code-validation): Today we are introducing the Baz CLI. The Baz CLI introduces an interactive review loop in the terminal, helping developers walk through changes, review them against requirements, ask questions, and resolve issues faster before merge. - [One agent to control all Rule Files](https://baz.ai/resources/blog/one-agent-to-control-all-rule-files): My clear, practical take on the agent rule file problem: why Markdown sprawl and token bloat happen, and how Baz fixes it by turning agent prompts into a single, versioned policy that is translated into tool-specific rule files for consistent, auditable agent behavior. - [Fixer Agent - Stop Wasting Time on Routine Pull Request Fixes](https://baz.ai/resources/blog/fixer-agent-stop-wasting-time-on-routine-pr-fixes): Fixer Agent resolves lint, format and build failures in your pull requests, verifying each fix in an ephemeral sandbox before it commits. - [Introducing Private Mode](https://baz.ai/resources/blog/introducing-private-mode): When Private Mode is enabled, Baz analyzes your changes locally and does not store or retain your source code. This is designed for teams with strict security, compliance, or data residency requirements who still want deep, structured code review. - [Pixel Perfect by Baz - Make designs and implementations match. Every time.](https://baz.ai/resources/blog/pixel-perfect-by-baz): Automated visual review that surfaces missing UI elements and tooltip regressions and maps them to code. - [Introducing Fixers: Environment-compatible auto-fixing agents](https://baz.ai/resources/blog/fixer): Fixer is a sandboxed auto-fix agent that applies and executes PR fixes inside an ephemeral, isolated runtime. It runs format, lint and build checks and only commits changes when validation succeeds, saving reviewers hours reworking incoming feedback. - [Spec Reviewer Design Spotlight](https://baz.ai/resources/blog/spec-reviewer-design-spotlight): Spec Reviewer Design Spotlight turns screenshot evidence into a single, review-ready artifact for each requirement. - [The AI Coding Loop Hits Runtime](https://baz.ai/resources/blog/the-ai-coding-loop-hits-runtime): Developers found flow in the IDE, now agents are chasing truth in execution. - [From Bots to Maintainers: What Bot Comments Teach Us About AI Code Review](https://baz.ai/resources/blog/from-bots-to-maintainers-what-bot-comments-teach-us-about-ai-code-review): We read thousands of PR comments across nearly 100 top open source repos. AI code reviewers develop personalities, and generic ones create friction inside a company. - [Spec Reviewer: Bringing Figma and Jira Context into Every Pull Request](https://baz.ai/resources/blog/spec-reviewer-bringing-figma-and-jira-context-into-every-pull-request): By embedding design and product understanding directly into review, Baz removes an entire layer of back-and-forth between design, engineering, and QA. - [Engineering Intuition at Scale: The Architecture of Agentic Code Review](https://baz.ai/resources/blog/engineering-intuition-at-scale-the-architecture-of-agentic-code-review): We benchmarked AI coding agents on real production bugs and built a multi-agent reviewer that maps context, infers intent, and proves risks like a senior engineer. - [What Open Source Pull Requests Teach Us About AI Code Review](https://baz.ai/resources/blog/what-the-top-open-source-projects-teach-us-about-ai-code-review): Pull request trends from LangChain, Kubeflow, Terraform, PyTorch and Spring, and what they reveal about how AI should review code. - [Code Review Comments to Dev Standards: 4 Internal Code Prompts We Turned Into AI Reviewers](https://baz.ai/resources/blog/from-code-review-comments-to-dev-standards-4-internal-code-prompts-we-turned-into-ai-reviewers): Every engineering team has unspoken rules, the kind of standards you’ve repeated in PRs a hundred times but never bothered to write down. With AI code reviewers, those patterns don’t need to stay scattered. - [5 Ways to Measure the Impact of AI Code Review](https://baz.ai/resources/blog/5-ways-to-measure-the-impact-of-ai-code-review): How to track outcomes, developer sentiment, and real ROI for your AI - [Code reviews are broken: How GitHub’s poor UX Is hurting developers](https://baz.ai/resources/blog/code-reviews-are-broken): Why GitHub's outdated code review experience is frustrating developers and stifling progress. Explore the key flaws in its workflow and what needs to change for modern development. - [From Review Thread to Team Standard: How We Built AwesomeReviewers](https://baz.ai/resources/blog/from-review-thread-to-team-standard-how-we-built-awesomereviewers): AwesomeReviewers demonstrates how code review discussions can turn into guidelines that capture the team’s collective wisdom. In this post, we’ll share how it works under the hood, and what we learned building it. - [Building an AI Code Review Agent: Advanced Diffing, Parsing, and Agentic Workflows](https://baz.ai/resources/blog/building-an-ai-code-review-agent-advanced-diffing-parsing-and-agentic-workflows): Enhancing AI-driven code reviews with Git Diff, Difftastic, and Tree Sitter. Discover how syntax-aware diffing, structured parsing, and historical tracking improve AI’s ability to understand and optimize code changes. - [Baz launches AI code review with $8M in seed funding](https://baz.ai/resources/blog/ai-code-review-baz-is-now-available-backed-by-8million-seed-round): Baz is now available to everyone. An AI code reviewer that reads across repositories to catch what single-file tools miss, backed by an $8M seed round. - [Baz Reviewer: Automated Code Review With Context](https://baz.ai/resources/blog/baz-reviewer-one-more-step-towards-automated-code-review): Baz Reviewer gives automated code review the context it usually lacks: what changed, why it matters, and what to do about it. - [Beyond LGTM: Master Code Reviews with AI](https://baz.ai/resources/blog/beyond-lgtm-master-code-reviews-with-ai): A streamlined approach to code review - focused on clarity, precision, and outcomes - transforms the process into a faster, more effective workflow that drives stable releases. - [Copilot vs. Code Quality: Are Current Metrics Enough to Define AI Success?](https://baz.ai/resources/blog/copilot-vs-code-quality): Copilot vs. Code Quality: Are Current Metrics Enough to Define AI Success? - [How to Achieve Structured Output in Claude 3.7: Three Practical Approaches](https://baz.ai/resources/blog/how-to-achieve-structured-output-in-claude-3-7-three-practical-approaches): Achieving reliable structured output in production can be tricky, especially with extended thinking mode. By using approaches like no thinking mode, careful prompting, or combining reasoning with a separate structuring model, you can effectively work around these limitations and integrate structured outputs into your GenAI applications. - [Bazzy: An AI Code Review Assistant With Context](https://baz.ai/resources/blog/meet-baazzy-the-ai-code-review-assistant): Bazzy is the AI code review assistant giving context-aware suggestions across multi-repo, multi-language codebases, for faster reviews with fewer errors. - [Practical AI: 5 Real-World Code Review Wins](https://baz.ai/resources/blog/practical-ai-5-real-world-code-review-wins): Examining five real-world scenarios that demonstrate how AI already helps developers maintain coding standards, reduce downstream issue, and deploy faster. - [Scaling AI feedback onto Codebase Context: A Primer on RAG Code Reviews](https://baz.ai/resources/blog/scaling-ai-feedback-onto-codebase-context-a-primer-on-rag-code-reviews): Discover how Baz Reviewer leverages Retrieval-Augmented Generation (RAG) to transform code review. Learn how dynamic retrieval, semantic analysis, and AI-powered collaboration help developers detect redundancy, enforce best practices, and streamline reviews. - [The Next Evolution in Code Review: Tracing-Driven Insights](https://baz.ai/resources/blog/the-next-evolution-in-code-review-tracing-driven-insights): Learn how integrating telemetry, OpenTelemetry, and AI-driven insights can enhance code reviews and optimize development workflows. Explore how Baz is leading this evolution in software development practices. - [Too Many (Unsung) Heroes – The AI Echo Chamber](https://baz.ai/resources/blog/too-many-unsung-heroes-the-ai-echo-chamber): AI-generated content is flooding the internet, and models are increasingly being trained on their own outputs. This self-reinforcing cycle is creating an echo chamber - one where originality diminishes, clichés reign, and the quality of AI-generated insights declines over time. - [Can AI Code Review Tools Catch Breaking Changes?](https://baz.ai/resources/blog/where-ai-succeeds-and-fails-in-predicting-breaking-changes-in-code-review): We tested whether AI code review tools can predict breaking changes. Where they succeed, where they fail, and what closes the gap. - [Why AI Code Assistants Fail on Large Codebases](https://baz.ai/resources/blog/why-ai-code-assistants-fall-short): AI code assistants like Copilot promise efficiency but fall short on multi-repo, multi-language projects. How tracing and observability close the gaps. - [Why Your AI Code Gen Doesn’t Understand Diffs](https://baz.ai/resources/blog/why-your-code-gen-ai-doesnt-understand-diffs): Traditional diffs make code reviews noisy and inefficient. AST diffing offers a smarter way to detect meaningful changes, reduce false positives, and improve developer workflows. Learn how Baz is streamlining code review with AST-based analysis.