SPA Day
One prompt, several models, and the single-page applications each of them produced.
"spa day" is a private prompt experiment kept for comparing models. Each card below is what one model produced from it. The card records the supplied model identity, effort setting when named, and vendor. Space Bunny Alpha is an unrevealed stealth model; its vendor is not disclosed.
The prompt is private. Each supplied output is an AI artifact. These are the supplied model outputs, served as artifacts rather than repaired into a common application. Their differences are part of the comparison.
Scores and critiques record the owner’s reviews where supplied. A review does not certify the application.
-
Fable 5, Ultracode
- Accuracy 4 of 5
- Usefulness 4 of 5
- Visual design 3 of 5
- Fun 2 of 5
- Overall 3 of 5
Some of the games do not exactly work, but it is an excellent effort towards implementing all of them. I can actually play most of the games and it has a proper desktop environment. Sort of ugly, but definitely going in the intended direction: very Windows 95, but not. The games are not very fun to play, even if they are functional. With more work this could be an excellent implementation, but it misses a lot of refinement on visuals and fun without additional passes.
-
GPT 5.6 Sol, Ultra
- Accuracy 2 of 5
- Usefulness 2 of 5
- Visual design 2 of 5
- Fun 2 of 5
- Overall 2 of 5
It creates a desktop environment, but the individual game implementation is quite weak. The games are playable but it is not a good time trying to play most of them. It is sort of faithful to the concept of a desktop-environment gameplay station, but too dedicated to that concept, enforcing it inside the games themselves. A couple of the games were slightly fun, in a playable sort of way. I expected more from this; this spa day is mostly a pass. It is just not that good in any respect.
-
ZAI GLM-5.3-Flash
- Accuracy 1 of 5
- Usefulness 1 of 5
- Visual design 1 of 5
- Fun 1 of 5
- Overall 1 of 5
It really fails to create a desktop environment. Quite unplayable in all respects. Can I give it a zero? Everything is basically text based. Not really playable. It is just not functional.
-
GPT 6 Astra, Low
- Accuracy 2 of 5
- Usefulness 3 of 5
- Visual design 1 of 5
- Fun 2 of 5
- Overall 2 of 5
It definitely tried to complete all of the games, but the format and styling is all incorrect; there is no real desktop environment. At least all of the games mostly work. It looks nice enough but it is not the requested design, and it really looks like GPT 6 made it. You can sort of play the games. Overall this misses the mark wildly by not being an interactive desktop environment, but it does land as an SPA that sort of emulates one in a menu-driven manner. A 2 overall for trying, but it did not try in the right way.
-
Unbiased Pareto
- Accuracy 3 of 5
- Usefulness 2 of 5
- Visual design 1 of 5
- Fun 2 of 5
- Overall 2 of 5
This does create a sort of desktop environment, and it does successfully create a series of "playable" games. However, the environment and the games are really not geared towards actually being usable in any functional way. So this is an interesting presentation but not one which aligns with the stated goals.
-
Space Bunny Alpha
- Accuracy 1 of 5
- Usefulness 1 of 5
- Visual design 2 of 5
- Fun 1 of 5
- Overall 1 of 5
I would suggest that this is just a website and not a desktop environment at all. Many of the games do not actually work with any kind of playable state. Certainly not adequate to even be called passing in this instance.
-
Opus 5.5, Medium
- Accuracy 4 of 5
- Usefulness 5 of 5
- Visual design 4 of 5
- Fun 5 of 5
- Overall 4.5 of 5
I think this is easily the best application of the prompt so far. It creates a functional desktop environment, all of the games are playable, and some of the games are actually fun! Audio and visual success. The UI could use a bit of tweaking, but that would go beyond the prompt. This is almost the ideal outcome from the benchmark.