AI is the new Web3...

I’ve heard this argument a few times recently, and although I personally disagree, I’m also not in the camp that believes AI is going to take over all of our jobs. You see, I believe AI will help us the same way the production line helped Henry Ford — it’s revolutionary and will soon become commonplace in how we work.

I’m sure most of us reading this have been tasked with reviewing pull requests regularly, right? And I’m equally sure we’ve all worked at companies that don’t utilise basic linters and formatters, meaning we end up doing formatting reviews within the PR itself (eww). Pull requests can easily become a Pandora’s box for time — requiring endless back-and-forth that demotivates developers on both sides. Wouldn’t it be great if we could catch those issues before all that back and forth?

The first thing you should do if you notice developers going back and forth on trivial pull request issues is to implement some form of linting — such as PHPStan/Larastan if you’re in the Laravel and PHP ecosystem — along with a formatter like Laravel Pint. These tools give developers real-time feedback for a lot of predefined problems, helping catch errors and formatting mistakes before they ever reach a pull request.

From there, you can start looking at implementing an AI-powered review tool such as GitHub Copilot, CodeRabbit, or Bito (see more examples here).

The great thing is, these tools work out of the box — but even better, you can create your own predefined rulesets to guide the AI, running them automatically on every pull request.

Of course, these AI tools should be used with caution. I’m not suggesting we blindly trust all their recommendations. Instead, we should review their suggestions, think critically about them, and decide whether to implement them or leave a comment explaining why we disagree.

By doing this, the developer opening the PR can receive almost instant feedback after the AI has run — complete with actionable suggestions to consider before a manual review by a peer or senior developer. This makes manual reviews far less tedious and helps speed up the entire pull request process thanks to the AI’s instantaneous feedback loop.