Rage-Click This, Then Proofread Something

A tiny habit that outperforms million-dollar consultancies.

Welcome to 2025, where professional success means doing what used to be the bare minimum. Where reading your work makes you exceptional. Where giving a shit is a competitive advantage. The bar is underground. Step over it and win.

In this month’s news, Deloitte got caught submitting hallucinated academic citations to the Australian government and you know what? Good. Fuck them.

Not because they used AI. I use AI. I make a living on Claude writing code while I argue with boards and regulators about compliance. We are doing the same job. The difference?

I READ THE FUCKING OUTPUT.

Revolutionary concept, apparently.

Last week Claude generated a piece of config that looked beautiful. Clean. Elegant. Would’ve passed every code review at your startup. Also would’ve exposed our entire banking infrastructure to the public internet. Caught it on line 683. Because I read line 683. Because that’s my fucking job.

But lawyers billing $800/hour can’t be bothered to Google whether their cases exist. Consultants charging millions submit reports that sound like ChatGPT having a stroke. Developers - DEVELOPERS, MY PEOPLE! - push and approve each other´s untested AI completions to prod. To PROD. Payment systems. Auth flows. Database migrations. Just... yolo, let’s see what happens.

It’s insulting to the client. To yourself. You’re charging price-of-a-house money and you can’t even be bothered to read what you’re selling them? Fuck you.

---

---

The Scam Was Always There

Here’s what kills me: everyone’s acting surprised. “Oh no, AI makes mistakes!” “Quality is declining!” “Professional standards are eroding!”

No. Stop. Here’s a hot take:

AI didn’t make professionals lazy. It revealed who was never reading their work in the first place.

That lawyer who submitted fake cases: Do you think they were carefully reading contracts before? They were find-and-replacing names in templates. The Deloitte consultant? Wikipedia to PowerPoint converter. That developer shipping untested code? Brother, they never wrote tests anyway.

AI just made it obvious. Stripped away the busywork that was hiding the fact that nobody was doing the real work.

I generate 50 changes of all sizes and criticality a week with AI, code, reports, emails, the works. You know what that’s like? It’s drowning in almost-right. Every single one looks good at first glance. Most ARE good. Some are 95% perfect with 5% devastation hiding in the middle. Like that time Claude helped us restore a corrupted database. Deleted 150,000 production records. Took me 2 hours to unfuck that (PITR, people, saves lives). On the incident report, my name; not Claude’s.

The thing is - I caught dozens of other fuckups before they hit prod. Why? BECAUSE I READ THE CODE.

Not skim. Not glance. Not “looks good, ship it.” Actually read. Line by line. Understanding what each part does. What it affects. What breaks when it breaks.

This is apparently a superpower now.

Your Reviews Are Worthless And Nobody Gives A Shit

“But we have review processes!” Yeah? Your review process is two people adding “Please review” and “looks good to me” without reading. This applies to everyone from the salesperson to the CEO.

I know because I’ve watched you do it. Senior engineers approving 10,000 line PRs in 30 seconds. “Looks good to me!” How? HOW does it look good? You didn’t read it. You saw green checkmarks from the linter and called it a day.

The partners at BigLaw and Big 4 built these elaborate quality theater productions. Six rounds of review! Quality committees! Professional standards boards!

All of it, every single layer, was just people assuming someone else was reading it.

Junior generates with AI -> Senior “reviews” (doesn’t read) -> Manager “approves” (doesn’t read) -> Partner signs (definitely doesn’t read) -> Client gets fucked -> “AI error”

No. It was 4 human errors. The error of not doing your fucking job.

When Claude writes code for me, I treat it like a gifted but psychotic junior developer. Brilliant one moment, actively trying to drop production tables the next. No malice, just... confidently wrong about things that matter.

My review process:

• Assume it’s broken

• Assume it’s insecure

• Assume it missed the requirement

• Prove otherwise

That’s not AI-specific. That’s how you should review ALL work. But nobody does. They see something that looks passable and ship it.

You know what happened when I missed one review? When I got tired and just trusted Claude’s Terraform? Misconfig of a stream cost us $2k in AWS credits over a day. Alerting caught it, thank fuck, but it was still my mistake. Mine. Not Claude’s. Mine.

The Golden Opportunity

Want to make a killing professionally in 2025? Please subscribe to my paid content lol.

Joking aside, here’s the entire requirement:

Read what you ship.

That’s it. The bar is so low it’s underground. Clients and employers are desperate for someone who cares; you just have to be slightly better than absolutely terrible. While everyone else is writing think-pieces about “AI governance” and “the future of work,” you just... do the work.

So here’s the thing: this window won’t last. Eventually, the bar will rise back to ground level. Basic competence will stop being a superpower.

But right now? Right now you can dominate just by caring a bit. Just by reading to line 683. Just by googling whether citations exist. Just by reading every line, catching every subtle fuckup, owning every outcome. because that’s what they’re actually paying for. Not the code. The judgment. The ownership. The giving-a-shit.

AI writes the output. You own the mistakes. That’s the deal. And if you’re not willing to make that deal, stop pretending you’re a professional.

You’re just an expensive API call who’ll be replaced by a Chrome extension.