Play game
dreamseam's itch.io pageResults
| Criteria | Rank | Score* | Raw Score |
| Overall Enjoyment | #1 | 3.900 | 3.900 |
| AI Direction & Adaptability | #3 | 3.900 | 3.900 |
| Theme Fidelity & Creativity | #5 | 3.900 | 3.900 |
| Production Velocity | #8 | 3.400 | 3.400 |
Ranked from 10 ratings. Score is adjusted from raw score by the median number of ratings per game in the jam.
The Biggest Challenge
Finding the core loop, and then calibrating it. The AI could produce plausible systems far faster than we could tell which one was the game, so the loop was found by cutting rather than by adding, and every number in it was set by a playtest rather than by a guess. The loop that survived: flying close to the shapes pays, and flying wing to wing with Ito charges a bond that fuses the two of you into one being — the only thing that fits through the Climb. The decision that made it a game was coupling those two: flying together in safe open sky charges nothing at all. Before that, the Buddy was a shelter you could hide next to; after it, togetherness is something you can only earn where it is dangerous. That single rule is now an assertion in the test suite, because it is the one thing the whole design rests on. Then the calibration, and almost all of it came from watching people play. A first-time player could not get past layer 1 at all until the director sat beside her and explained each mechanic — that verdict produced the guided first dream and a one-per-session mercy on the opening layer. A flown probe showed that nobody, human or synthetic, was passing the Climb: auto-UNISON was spending the merge 300 m before the one wall it was the key to, so the bond is now held and spent at 150 m. The scoring pocket went to 9 m with hysteresis at 13 and one number per pass, because a stream of small numbers reads as noise and one number reads as a reward. Six layers were measured across two seeds and two pilot skill levels to check the curve actually rises. The hardest part was not building any of it — it was accepting how much had to be thrown away between the version that sounded good and the version that played.
AI Tools Used
Claude (Anthropic), used through Claude Code, as the pair programmer for all of the code. Image generation for the crystalline logo and the cover art. The six per-layer music tracks were AI-generated (album "Dusk Drift", Treblo) and then processed by hand — low-passed at 3 kHz, normalised to -20 LUFS, faded — to the director's brief of "no highs, always fades, never loud". Every sound effect is synthesised in code; there are no audio asset files.
Areas Where AI Was Used
Code: all of it — flight, procedural world generation from a seed, the sky, the HUD, the 3D bedroom that opens the game, and the automated test suite. Music: generated, then processed by hand. Sound effects: synthesised in code. Images: the logo and the cover art. Not AI: every design decision, playtest verdict, difficulty call and art-direction call was the human director's. Several AI proposals were rejected outright — a throw/launch mechanic for the Buddy (twice), an angel-winged version of the fused form, and, in this last session, making the recording pilot invulnerable so the trailer clips would stop ending in crashes. That last one was cut with "it's no use making it survive the crash, it has to be a well-flown scene."
Leave a comment
Log in with itch.io to leave a comment.






Comments
No matter how many times I try, I always get stuck just past the entrance to the building and end up crashing. I still don't understand what exactly I did wrong.
The 3D graphics are stunning, and the sensation of floating is delightful. I really wanted to see a little more of the game, but I gave up.
Still, I think this game has genuine appeal. 🙂
Yes, I think you're absolutely right about the location. Thank you. That made sense, so I tried it again. However, even though the gauge was at about 80% on a few of my attempts, pressing the F key didn't seem to have any effect, and in the end, the result didn't change. It seems that even understanding the theory can't make up for my lack of skill...
After playing about 10 more times, I finally made it to the next, darker stage🌟
I now understand what “UNISON” means, and I feel like I’ve gotten a sense of that pleasant feeling it brings.
Thank you so much for letting me have such a wonderful experience! 🙂
Beautiful entry. The engineless glider — no throttle, no brake, just pitch and momentum — made "DIVE" a mechanic instead of a label, and following Ito's red thread with the SYNC/UNISON meter is one of the most genuinely load-bearing takes on BUDDY I've played; your buddy isn't a passive companion, it's the line you read to fly well. The dream-world art and the hand-processed, always-fading music sell the mood completely. Loved reading that you rejected the throw/launch-the-buddy idea twice — the restraint shows. Congrats on a standout; wishing you well in the ranking.
I really enjoyed this game although I didn't like how it stopped in the tutorial. Instead, I wish it let me learn from my mistakes
Of course. I'm still new to game dev, so I'm glad I actually had some good feedback despite my lack of experience. I really did enjoy your interpretation of "dive" its vary fun, and I can't wait to see what you do with it post-jam.
It’s fun—but pretty challenging!
The experience of flying alongside a partner is really entertaining.
I kept getting a "Game Over" because I wasn't skilled enough,
but it’s the kind of game that makes you want to keep trying so you can see what lies ahead.
It’s amazing to create a game of this quality in just three days!
It feels absolutely amazing to fly around this beautiful world. ✨