Bluebird Count

Bluebird Count: Artist's Statement

Connor Velte · Northeastern University · Class of 2027 · Finance and AI

Bluebird Count is a game about looking closely. I built it around five habitats, from a spring meadow to a snowy field at dusk, each with a short timer and a set number of Eastern Bluebirds hidden in it. Every scene is full of blue things that are almost the bird: cornflowers, blueberries, irises, lake water, winter shadows, and two real look-alike species. The game lives in the gap between seeing something and noticing it. You only find the bird by slowing down and checking for the one detail that gives it away, while the clock keeps running.

The first decision I would defend is that the decoys get harder in a set order. The forest adds Blue Jays, the lake adds Indigo Buntings, and the autumn ridge fills the scene with orange leaves and rose hips that match the bluebird's rusty breast. Each round takes away a shortcut the player learned in the last one, and a card before each habitat names it, so the difficulty reads as a lesson. The second is the penalty. Clicking empty ground costs two seconds, and clicking the wrong bird costs four. Without a cost, the best strategy is to click everything blue. Two seconds is small enough to risk a guess and large enough to make you look twice. The third is the ending. When time runs out, the game circles every bluebird you missed, because "it was right there" is the feeling the whole game is built on.

Claude Code produced things I did not ask for. Its best addition was bird songs panned toward the bird's side of the screen, which rewards listening as well as looking and fits the theme. Where I overruled it: the first version was too easy. I cleared every round with time to spare, which killed the tension the game depends on. After ten rounds of playtesting, I had it cut the timers from 45 to 65 seconds down to 30 to 40, shrink the birds, add more decoys, and hide more bluebirds behind grass, reeds, leaves, and snow.

I consider this game mine. The concept, the core rule, the escalating decoys, and the choice to make carelessness cost you were my decisions. Claude wrote the code and filled in details like the specific habitats and species. I decided what the game was for, judged whether it worked, and sent it back when it did not. The assignment calls fun a human judgment, and every one of those judgments in this game was mine.

Prompts that guided development

"make a game where you have to find and click a number of hidden blue birds throughout different wilderness environments, it should have other similarly colored elements to make finding the birds increasingly difficult, make rounds and a timer for each round"

"the game was too easy so i told you to make it harder and the rounds faster"

AI disclosure

Tools: Claude Code (Claude Opus 5.5), one session. I designed the game and directed every revision; Claude Code wrote all of the code and drafted this statement, which I revised. Iteration: one build, ten rounds of playtesting, and one correction pass, which made these changes: Round timers cut from 45 to 65 seconds down to 30 to 40 seconds.

Bluebirds drawn about 15 percent smaller.

About 40 percent more blue decoys in each habitat.

Look-alike birds increased from 18 to 25 across the five rounds.

More bluebirds partly hidden behind grass, reeds, leaves, or snow.

Bird songs and wing flicks made less frequent.