What we learned building a games site with AI

Before AiGILE, our team spent more than twenty years as 3RD SENSE, designing games, gamification and apps for businesses: more than 300 game and gamification projects and over 60 published apps. So when we wanted to put AI-assisted software development to work on something real, something we would have to run and answer for ourselves, a games site was the obvious choice. This month we relaunched Playaholics, and AI did a large share of the building.

What we built

Playaholics went live on 2 October 2026. It has four daily games so far (Pot or Pop, Kwizi, Closed Circlet and Psychomaths), the same puzzle for everyone each day, and leaderboards for today, this week and all time. Players can play straight away as guests, or create an account to keep their scores and streaks. Since launch we've added a virtual currency called P-Nuts with a shop to spend it in, daily reminders, and the option to save the site to a phone's home screen like an app.

Underneath, it's more than a set of web pages. Every score is worked out on our server from the moves the player made, so nobody can simply send in a high score. Each game is its own package, built separately and added to the site without changing the site's code, so a finished game can be added in minutes. The site runs on AWS in Sydney.

What AI did well

The routine code. Most software is plumbing: forms, accounts, database tables, the checks that data is what it should be. AI writes this quickly and consistently, and that is where most of the time went in the projects we used to run.

Following a plan. We wrote down what we wanted before building: the screens, the rules for scores and leaderboards, and the order to build things in. Given a clear plan, AI worked through it in steps, and each step could be checked before the next one began.

Testing as it went. Alongside the code, AI wrote automated tests, and it ran them after every change. When a change broke something, it usually knew before we did.

Content at scale. Kwizi, our quiz, needs a steady supply of new questions. Each week, one AI model writes questions and a second checks them for accuracy and fairness, changing or dropping the ones that don't pass. More than 1,200 questions are live so far.

Where people made the difference

Deciding what fun is. AI can build a game to a description. It can't tell you whether the game is any good to play. That took people playing it, over and over, and changing the design: showing your score on each level as you go, adding a penalty for careless shots, making the daily game's button impossible to miss.

Judgement calls. Small decisions shape a product, and they need someone who knows the players. When the daily reminders were first set for 8am, the answer came straight back: who plays games at 8am? They now go out at 10:30.

Finding the bugs that matter. Some problems only show up on a real phone, in a real hand. Playing Pot or Pop on an Android phone turned up a freeze at the end of a game that no automated test had caught. Once we could describe it, AI tracked down the cause and fixed it the same day, but a person had to find it first.

Keeping it fair and safe. What data we collect, how we treat players under 18, when a score counts and when it doesn't: these are decisions about people, so people made them, and AI built to them.

What it means for your business

The lesson for any business adopting AI is the one we see in our consulting work. AI didn't replace the team. It changed what the team spends its time on: much less on the routine, and more on the decisions, the design and the checking that only people can do. The work moved in small steps, each one tested before the next, which is the agile loop we use for adoption applied to software.

It also changes what's worth building. When we develop software today, we're seeing development time cut by up to 80%. Every project is different, so take that as what's possible rather than a promise, but a small team can now take on work that would once have needed a much bigger one. Software that didn't seem worth building a few years ago, an internal tool, a customer app, a game to teach your people something, may well be worth another look.

If you'd like something built

We still design and build software, games and gamification included, and today we use AI through much of the build. See our software development page, or tell us what you have in mind.

And if you found us through an old 3RD SENSE game, welcome back. Come and play at Playaholics.