In my last post [Building Love into My ‘Birdie’ Game for Valentine’s Day.], I walked through the planning phase of adding Valentine’s Day features to Birdie, my educational bird identification game. We nailed the vision, mapped the specifications, and made sure the idea was solid before touching production code.

This time I want to talk about the part that actually matters: shipping the update.

I used Claude Code to add new features to an existing app without breaking what already worked. That sounds simple, but anyone who has shipped changes to a live project knows the real challenge is not writing new code. The real challenge is changing the right things, testing the right places, and not accidentally breaking something that was fine yesterday.

Birdie is an educational bird identification game that turns learning avian species into a small, gamified challenge. Players identify birds by name from a set of options, earn points for correct answers, and unlock achievement badges after perfect streaks.

This Valentine’s update expanded Birdie from 28 birds to 48 species. Each bird has an illustration, scientific name, and educational facts woven into the gameplay.

The centerpiece is the Love Birds Challenge, a special 14-question game mode themed around February 14. It focuses on monogamous bird species, meaning birds that mate for life or form long term partnerships.

Players go through 14 rounds of identification. Each round has one correct monogamous bird and two wrong answers chosen from non-monogamous species. That creates a different rhythm from the main game, which usually rotates through a smaller pool of birds.

The education angle is what makes this feature worth building. Players are not just guessing names. They are learning mating behaviors and long term partnerships in birds, which are surprisingly common compared to mammals.

Normally Birdie shows your score as something like 7 out of 9. During the Heart Multiplier Event, the game uses hearts instead, so it becomes 7 hearts collected.

It is a small change, but it changes how the result feels. A fraction pushes you to think about what you missed. Hearts push you to think about what you gained. That matters when you are trying to create a seasonal moment that feels fun instead of stressful.

The third layer is short bird facts tied to the theme. After a game, players see educational callouts about courtship and long term bonds in specific species.

It is a simple mechanic, but it reinforces the mission of Birdie. The reward is not just points. The reward is learning something that sticks.

This is the part I wish more AI build content talked about. It is easy to generate code. It is harder to update an existing product without rewriting half the project or introducing subtle breakage.

First, as always, i brainstormed in Claude about the changes i want made. I have a project dedicated to the Birdie came so it has quite a lot of context from inception to tech stacks to architecture. It does sometimes get some things wrong so reading and correcting its assumptions was very critical. After the brainstorming, i had it write a prompt for me to feed to Claude Code which is integrated to VS Code (I’ll provide steps on how to set this up in another post so stay locked in). The prompt contained a clear description of intent and specifications. With that, the vibes started flowing. After each change, I tested against the spec. If something felt off, I corrected it immediately. One prompt doesnt do it. Vibe coding is really a communication skill. You communicate and iterate till you get what you envisioned.

The Love Birds Challenge needed enough non-monogamous birds to serve as believable wrong answers. The original roster did not have enough, so I expanded the bird library by 20 species.

Those new birds are not throwaway seasonal content. They become permanent assets that strengthen the main game after Valentine’s Day. I initially used Ideogram for generating the first batch of images, but for this next batch i used Whisk from Google

Planning upfront actually saves time. The execution moved smoothly because the hard decisions were already made.

Reusable assets compound value. If you are generating assets, design them so they can live beyond a single event.

AI works best when you know what you want. The model does not replace clear thinking. It amplifies it.

The Love Birds Challenge, Heart Multiplier Event, and Valentine’s Bird Facts are live. Check it out and let me know what you think.

If you are building with AI and trying to ship real updates, I hope this gives you a framework you can reuse.

Join the Wheels of Automation.

Subscribe