Sourcery | LinkedIn (original) (raw)

About us

Never wait for a code review again Instantly review your code in IDE, CI or GitHub with our AI review tool - https://sourcery.ai/code-review

Industry

Software Development

Company size

2-10 employees

Headquarters

London, England

Type

Privately Held

Founded

2018

Sourcery's code reviews and security analysis are now available for any team building with Vercel as part of their AI Agent marketplace! See how Sourcery can help catch bugs, fix security vulnerabilities, and more for any of your Vercel projects - https://lnkd.in/eQVTJNAh
I wanted to see if bad quality code slows down/disrupts coding agents the same way that it impacts developers. I took a set of tasks and through them at Gemini CLI, Codex, and Claude Code and then made the code quality for each task progressively worse to see how things changed. Here's what I saw: - Traditional "code smells" didn't really bother agents - Moments of very high complexity or very low maintainability index made it way less likely that an agent could complete a task - More complexity = more cost per task But beyond that I saw agents do some wild things: - Deleting their entire test suites to make the tests "pass" - Finishing a task and then getting stuck in a loop of deleting the entire project and rebuilding it from scratch The full blog post with all of the details is in the comments

Last Round

External Crunchbase Link for last round of funding

US$ 1.8M