Oh my, I bet Duffy was feeling pretty smug about this. I won’t comment on this guy’s tone, but:
He is right about one thing: witch-hunting creators who use AI to bridge a skill gap (like an artist using an LLM to write boilerplate code) is unproductive. If a tool helps someone ship a genuinely good game, pure Luddism adds nothing.
Everything else in the rant rests on broken analogies, false assumptions and (willingly so) misrepresentations and strawman arguments. Not that we haven’t discussed these here before, but since you brought it up…
-
The Paintbrush and Drill Fallacy (basically, a Category Error): Paintbrushes, drills, and Photoshop are deterministic instruments of direct human agency andc reativity: Every stroke or cut or pixel is executed by the user. Generative AI is stochastic delegation. Comparing prompting an LLM to holding a paintbrush confuses performing a task with commissioning one.
-
The Compiler False Equivalence: Moving from Assembly to C to Python is not the same as moving to AI. A C compiler is deterministic and predictable - you write the exact logic, and it translates it 1:1 into machine code. An LLM is probabilistic. You describe an outcome and hope the weights guess right.
-
Utility vs. Craft (Who cares how Facebook is coded?): People play new retro games not because it offers peak gaming convenience, they play them to see human ingenuity operate inside 198x hardware limits. Comparing commercial web infrastructure to a retro hobby misses the point of the scene by a mile. If someone finishes a marathon on an e-bike, nobody says, “Why do you care? Just be happy for them.” They point is, it is no longer a marathon.
-
The 90s Shovelware Excuse (Tu Quoque): Pointing out that 199x had plenty of bad games without AI does not justify flooding a niche community with automated slop today. High release volume is useless if gamers (and everyone else) have to wade through dozens of broken, uncurated AI experiments to find one playable title. And more than that, there’s infinite amount of original retro slop that we look back at with fondness right now, whereas AI slop will be forgotten within a month (if that). (also, I used the term ‘shovelware’ because, who knew! we already had a term fo badly designed, mass-produced titles, and nobody ever thought it was good, so I don’t know what the argument is here
)
People ask “Is this AI?” for practical reasons:
-
Signal-to-noise filtering: Users want to know if a release was engineered with care or spat out in an afternoon before investing time in it.
-
Fair categorization: Retro competitions, fanzines, and archives separate human-coded assembly from assisted or generated work, just as photography contests separate raw shots from heavy composites.
And that’s it.
PS I also moved your post to the correct thread.