I’m saying if you provide the same prompt to Opus 5.5 across 12 different people, you’re going to get some similarities that immediately arise.

I’ve mentioned this across many HN threads - I call it the LLM corollary to “garbage in, garbage out”: "generic in, generic out". If you prompt with more specificity, you’re obviously going to get different and potentially better results. But at that point, the artist is doing more of the work.

A while back, I tried a similar test asking for LLMs to come up with engaging D&D puzzles that would fit organically into an underground labyrinth environment. This is a pretty general prompt. Let's just say that the results were about as inspired as a tepid bowl of tapioca.

https://www.youtube.com/watch?v=B084LKO6ezI

Now imagine the same thing but each person is using Opus 5.5 - I’d wager there would be a lot of commonality among each of the outputs.

> Leaving there and proceeding for three days toward the east, you reach Diomira, a city with sixty silver domes, bronze statues of all the gods, streets paved with lead, a crystal theatre, a golden cock that crowns each morning on a tower...

Text in the second person is causing my my D&D instincts to kick in.

Even more so if it worked with PDFs.

I already have way too many projects on the backburner right now... wink wink nudge nudge to anyone who wants to take this.

Reverse engineering Redhook's Revenge binary (an old DOS game) before the advent of LLMs cost me way more hours than I'd care to admit back in the day - so I can't wait to put an LLM to work on some more obscure games like Sword Quest.

On a side note I should really give Oregon Trail II a shot. I never got into any of the successors like Yukon Trail, Amazon Trail, etc.

https://news.ycombinator.com/item?id=49676394

They often ship the original assets in a somewhat brazen disregard for basic copyright law even when the games are still for sale on places like GOG though.

This creates Sierra AGI-style adventure game scenes painted live from simple Turtle-esque drawing instructions so you can basically provide it an empty canvas and then position text labels on the canvas where you want certain things (tavern, oak tree, etc) and it will generate a custom script for rendering them in a EGA graphics style.

https://kq-styles.specr.net

The more banal your prompt is, the more banal the output is going to be. People have been testing LLMs with little things like “write a short fantasy story,” for years now and most of the stories are exactly what you’d expect: prosaic drivel.

I call this “generic in, generic out,” an LLM corollary to the classic GIGO (“garbage in, garbage out.”)

QB64 is pretty popular as a sort of QBASIC-dialect with additional features like OpenGL, multimedia support, etc.

FreeBasic is decent as well but its focus is more on being a compiler whereas QB64 aims to be more batteries-included with its builtin IDE.

https://github.com/QB64-Phoenix-Edition/QB64pe

https://github.com/freebasic/fbc

Combat was great fun though because you could "curve" the bullets!

Another thing to remember: the NES sort of cheats in a sense, because the cartridges themselves could include onboard chips like the MMC3/5 [1] that extended the capabilities of the original NES, stretching its lifetime far further.

[1] - https://nicole.express/2022/the-nes-as-an-artifact.html

  TYPE PERSON
    AGE AS INTEGER
    FIRSTNAME AS STRING * 12
  END TYPE

  DIM KURTZ AS PERSON


  PRINT KURTZ._

However you are right that if you hit F1 on KURTZ it would then show User defined PERSON, and then you could hit F1 again and it would show built-in help of user defined TYPE with AGE/FIRSTNAME properties. Sort of like context-sensitive help I guess.

As far as subroutines/functions go, if you tried to pass data into a subroutine where the data couldn’t be coerced, it would just give you a generic type mismatch.

  DIM Y AS SINGLE
  Y = 10.5

  PRINTIT  Y   ' Throws parameter type mismatch
  PRINTIT (Y) ' Succeeds through a self-evaluated expression which is effectively BYVAL because Microsoft takes pain in your suffering.

  SUB PRINTIT (X%)
    PRINT X%
  END SUB

I've spent the last 6 months building a similar 1:1 faithful recreation of QBASIC and QuickBasic 4.5 that runs natively in the browser [1] - so I know way more about this than anyone should by right.

[1] - https://mordenstar.com/share/qbasic-core-demos

https://www.twitch.tv/videos/2890955408

Here’s the most egregious example:

https://imgur.com/a/ju2tg75

I feel like that kind of event-driven programming model was very intuitive, even for younger audiences.

Prompt results are graded based a weighted calculation which includes: adherence to the prompt, image fidelity, and steerability.

My comparison benchmark also tends to favor prompt adherence, which a lot of others don’t. Most of ones that I've seen tend towards rather simplistic prompts (e.g. "neon-lit city facing a robotic uprising, with high-tech battles, in anime style"), whereas the prompts I've created try to test high specificity.

I’ve been running them all the way back to SDXL.

You can compare specific models using the "View All Models" so if you want to see the progression of open-weight models, or model X vs model Y, you can do so.

Just a heads up - I haven't added Flux 3 as I'm waiting until BFL drops the open-weights version.

Generative Comparisons:

https://genai-showdown.specr.net

Editing Comparisons:

https://genai-showdown.specr.net/image-editing