If AI writes 90% of the code, what is a senior dev actually paid for?

It’s 2030. Which developer will be worth more? A) The one who writes the best code B) The one who manages AI agents best C) The one who understands business best

Which part of a new project do you set up first: CI, tests or deployment?

You’re starting an indie startup tomorrow. You can only have ONE advantage: A) 100K followers B) $100K in funding C) Incredible coding skills

Idempotency is the most underrated word in backend work. If a request can be retried safely, half your failure modes disappear.

Someone builds an app using AI, gets 10,000 paying users, but can’t write code. Are they a developer? A) Absolutely B) Not even close C) Does it matter?

I'm building a football manager game from scratch - CAMPO ⚽️ And this is my first attempt at a 3D match engine. Current stack: • Three.js • Blender • Mixamo • Lots of AI-assisted coding The animations are still rough. Player movement needs work. The crowd isn't quite there yet. But seeing my own football simulation come to life in 3D for the first time feels incredible. Still a LONG way to go. Any Three.js or game dev experts here? What would you improve first?

Scope creep isn't a client problem. It's a missing sentence in the contract.

You can only use ONE AI coding tool for the next 12 months. A) Grok B) DeepSeek C) Gemini

I’m building a Brazilian football manager game called CAMPO. And I think I’ve massively underestimated how difficult it is to simulate 90 minutes of football. The game is still in development. One club per Brazilian municipality. Competitions following the official CBF calendar. But my biggest obsession isn’t the leagues, transfers, or graphics. It’s the match engine. I want players to behave like actual footballers. Not 22 mathematical formulas chasing a ball. So I started experimenting with Grok Bot and Claude Code together. And something genuinely interesting came up. A midfielder with almost no stamina left was still pressing, sprinting, and making perfect decisions in the 89th minute. He had played an exhausting match. Yet he behaved exactly like he did in the first 10 minutes. That bothered me more than it probably should. Because it exposed something fundamental. I wasn’t really simulating football. I was simulating football statistics. I started using Grok to question the match logic and Claude Code to investigate the underlying systems. Fatigue shouldn’t just lower a player’s speed. It should affect decision-making. Positioning. Reaction time. Passing accuracy under pressure. Even whether a player decides to make another run. And fixing one variable creates problems somewhere else. Make fatigue too aggressive, and every match falls apart after 70 minutes. Make it too forgiving, and players become superhuman. I’m still working on it. But this is what makes building CAMPO so addictive. Every time I think I’ve solved something, I discover another layer of football I haven’t properly understood. AI can help me write code. But teaching a computer to understand why a tired midfielder makes a bad decision in the 89th minute? That’s a completely different challenge. I’m curious. If you were building the most realistic football manager possible, what detail would you absolutely refuse to compromise on?

If every developer can use the same AI, what makes one team ship better than another?

Android and Google TV: github.com/Gathin23/cleanyou…

Writing code got faster. Understanding it, reviewing it and operating it didn't. The bottleneck moved. It didn't disappear.

Which do you trust more when debugging in production: logs, metrics or traces?

I think Grok Bot just exposed the biggest mistake I’ve made building my startup. I’ve been building an ERP for barbershops. Appointments, payments, inventory, customer management. Pretty much everything a barbershop needs to run its business. Last night, I connected Grok Bot to my system. I was expecting it to find bugs or suggest a few improvements. Instead, it noticed something I had completely overlooked. One barbershop had a nearly full appointment calendar. Revenue looked decent. Customer numbers were growing. But when Grok analyzed the actual appointments, it found a pattern. Customers were booking haircuts and simply not showing up. The barbers were losing hours every week. And my beautiful dashboard was telling them everything was fine. That honestly bothered me. I’d spent so much time building features, optimizing the database, and making the UI look good. But I’d completely missed something that was costing real people real money. So I started working on a way to predict no-shows and help barbers fill those empty slots. The crazy part? Grok didn’t write any code. It just looked at my business from an angle I hadn’t considered. Sometimes I wonder how many other obvious problems I’m missing because I’m too busy building. Any other founders letting AI analyze their actual business instead of just writing code?

The hard part of a codebase was never writing it. It was knowing why it looks the way it does.

What's the one feature that makes a management game impossible to stop playing?

You delete a photo. Then empty the trash. Your phone says it’s gone forever. But deleted data can still be recovered. So… does anything ever really get deleted?

Rate limiting isn't only protection against attackers. It also protects your database from your own bugs.

An app's onboarding has one job: get the user to the first moment of value before they decide to leave. Everything else can wait.