AI Code Review: Baz is now available, backed by $8million Seed Round

Source control with code comprehension, starting with Baz Reviewer

A few years ago, a team I had managed found itself in an intense high-stakes integration project. I committed to what I thought was a plausible deadline but quickly realized that merging codebases and unifying product roadmaps is a marathon, not a sprint. Despite having the best development talent, we encountered friction on the most basic programming tasks. Our single biggest pain? We got stuck in endless pull request queues and struggled to keep track of how each change impacted the next team over. Communication overhead skyrocketed - Email, Slack, Zoom but mostly discussions on GitHub. Even with modern codebases, common integrations and build frameworks, and a robust CI/CD pipeline, the missing linchpin was getting developers off of each other’s toes. 

Fast forward to today’s AI-enabled coding experience: Teams still need to agree on what to build, how to build it and track how changes impact existing flows. The old approach—get it to work locally and pass a bunch of diffs in hope they magically converge—just doesn’t cut it when a quarter half the lines are AI-suggested. I’ve experienced this pain first hand over and over again, and with the latest improvements in model comprehension of complex coding patterns, I believe we are on the brink of some substantial breakthroughs that will impact every developer.

Today we’re excited to share that our first core product, Baz Reviewer, is generally available. We’ve spent these past few months working with design partners that were sceptical about the prospect of making code review more effective. Together we worked to articulate, refine and perfect a system to use models to perform the best, most comprehensive code review imaginable every single time.

‍

This first step not only addresses the immediate inefficiencies in reviewing and merging code but also paves the way for a transformative vision: a coding system that truly understands code. This vision is what led to our $8million investment co-led by Battery Ventures and Boldstart Ventures, with participation from Vermillion Ventures, Secret Chord Ventures and Fusion VC. 

The fastest and most effective way to code review

Baz Reviewer tackles the limitations of legacy review platforms and Git-based diffing with unique capabilities that reduce downstream incidents and improve code quality by keeping developers focused on critical issues. By combining code diffs and application telemetry, Baz evaluates how code changes impact running services, endpoints and APIs.

Integrated into the pull request workflow, Baz categorizes code changes, identifies which code elements were modified and identifies potential breaking changes both within GitHub or optionally in a curated developer experience on the Baz platform. It accomplishes this through:

• Programmatic Software Verification Baz replaces manual workflows with an opinionated, automated review flow, analyzing usage patterns and data flows to detect changes in endpoints, HTTP methods, parameters, responses, enums, validation rules, authorization requirements, and behavior. 
• Harnessing Application Behavior Data A key differentiator of Baz is its ability to read and interpret code changes as if they are already running, giving developers a unique downstream outlook into future application states. This form of code analysis helps teams pinpoint potential breaks before code is merged into production.
• Elevating beyond Git-Centric Reviews Baz extends reviews well beyond diffing commits to introduce actionable, inline annotations that highlight precisely where problems may arise. It parses actual code grammar instead of the legacy text diff to provide high-fidelity context with accurate search, completions, and suggestions.

‍

Read through the docs here and development teams can try Baz out today, fully featured for an initial free trial.

For developers, AI that gets code

I’ve spent too many late nights playing a game of GitHub whack-a-mole. The AI shift makes it painfully obvious that we need a better system that honors dev collaboration. One that can handle the context of a codebase as well as good old-fashioned human quirks. If we can crack that, we don’t just make coding a bit faster or slightly less painful, we open the door for a whole new generation of builders, people who maybe never thought they could ship serious production code.

We’re here to solve for the trust gap that forms when AI code slips in a bug nobody saw; the endless confusion over who “really” owns a commit; the onboarding nightmare where new hires spend days retracing half-explained AI refactors; the style-guide turf wars that erupt when the AI rewrites code in ways half the team hates; the back-and-forth scramble when an AI suggestion arrives after the PR was “final”; the communication overload that grinds distributed teams down; and, of course, the massive AI-driven refactors that bury us in a thousand-line monster PR. The issues that a copilot or GitHub won’t fix.

 I am lucky to be partnering with Nimrod Kor, my long-time friend, and a true 1000x engineer. Nimrod has been building developer tools since forever with distinguished contributions in promoting open source Cloud & AI engineering. Nimrod has assembled a team of absolute rockstars that have instantly bonded and committed to solving this problem. If this blog made you believe, we are hiring aggressively.

Check out our Product Hunt profile here