Confessions of a recovering keyboard monkey | Why I review more than I write 

Written by our fantastic Full Stack Developer, Amy Miller.

Welcome to my morning routine. 

It’s 0850, I’m at my desk, a delightful flat white in hand (courtesy of my husband’s obsession with coffee or the fantastic Zudu coffee machine), fully charged laptop, maybe an emergency breakfast after dealing with a hectic school run. I turn on my laptop ready to check Jira, and I spot something different. 

2 tickets in progress 

7 tickets for review, with my face on them. 

I could’ve sworn I cleared that yesterday. 

That’s the moment I realise, I review more than I write these days. 

It’s slowly crept up on me, and it’s completely changed my way of working, but if you know me, you know I love a challenge. 

When I started as a dev a mere 6 years ago, I’d left a cosy career in marketing, one I loved, to get out of my comfort zone and try something I was truly passionate about.  

I entered the world of development. I put the hours in as a new parent and learned the difference between those array methods. Even the nippy ones. I tried to create my own SVGs from scratch, cause why not? I might need to do that one day. It might be my calling. 

Then I was introduced to AI and my world changed. 

Suddenly I didn’t need to know the difference between map and reduce. I didn’t need to type out lines of code and worry about a missing semicolon. I didn’t even need to create that SVG. AI took care of all of that for me, and more. 

I found myself relying less on writing components from scratch. If I needed something relatively simple, I’d prompt my friendly LLM, and it would take care of it for me. 

That’s the fast-paced world of development. I’m no longer the keyboard monkey I once was; bashing out code, waiting for hours for someone to review it and crossing my fingers that nothing would break post-merge. Back then I barely had time to think. If there was a bug and others were busy, I’d spend hours pulling apart the code, writing tests and scouring StackOverflow. Now, I can just paste that troublesome component into AI, and it’ll instantly point out that I’ve forgotten to do something extremely basic, and, let’s face it, save myself from the embarrassment in front of my colleagues. 

So does this make my career easy? Maybe even obsolete? My sweet summer child, of course not. It’s moved up a notch, we’re now in fifth gear. My little AI pal is only ever as good as the prompt I write, and that is the new normal. The setup of the repo. The careful planning and markdown files we craft. The work that goes in to the skills we create. It takes a lot of effort behind the scenes to have the confidence to be able to sit down, connect AI to figma and trust it to make the designs come to life, rather than fixate for hours over whether you should add 3px to padding, or was it margin? And hold on, shouldn’t I be writing that in grid? 

Along with that, you can do so much more. Previously, you would work on one task; now you can multitask. Why do one ticket when you could have three in the pipeline? When I was previously mulling over grid, I’m now prompting three sessions at once and reviewing the output. Did I mention my new skill set includes context-switching, juggling, and drinking a flat white all at once? Claude: remind me to add that to my LinkedIn. 

But here is my confession: I’ve never found reviewing easy. When you write code from scratch, it feels like a journey. You have a clear start, middle and end. You know the questions you’ve had to ask yourself and the conclusions you’ve made. When a review pops up, you’re entering that landscape cold. 

So while I stare at those 7 polished PRs, summaries all neatly laid out by AI, not a typo in sight, I must remind myself that whilst polished, it lacks the journey. It’s there to make me feel confident, but as a good developer, it’s my job to question it. Yes, the CI passes, but will I be cursing the code in three months’ time? Can I really trust the output from the LLM we’ve trained? 

The answer is a murky yes – and that’s only because I don’t trust it blindly. 

I’ve learned to review differently. I can’t guess what AI was thinking when it put together those lines of code, nor can I retrace a journey that I wasn’t on. Instead of asking myself, ‘what was it thinking?’ I need to pick it apart. How would I have written it from scratch? Why did it take this approach and not another? Has it tied itself in knots? What edge cases has it not tested for? 

It’s slower going, and it’s a completely different mindset than what I was used to in my monkey-days. Less typing, more judgement, this is the new normal. 

Reviewing is writing. You write the prompt, you write the skill, you read the output, you argue with the agent, you own the code. 

So yes, I review more than I write these days. It’s different, but as I said, software development is constantly evolving, and I wouldn’t have it any other way. Now I’d best get to those 7 PRs that need reviews.