I’ve painted and sculpted my whole life, and I have watched this technology arrive with the same knot in my stomach you probably have. So I’m not going to tell you the concerns are overblown. Some of them are exactly right. But “all of it is poison” and “all of it is fine” are both lazy positions, and the useful answer sits between them.
The line that actually matters
Here is the distinction I keep coming back to, and it has held up better than any other framing I’ve tried:
Does this touch the part where the judgment lives, or the part around it?
Almost all creative work is surrounded by labour that is not the craft. Writing the artist statement. Naming the piece. Drafting the pitch. Chasing the invoice. Making forty variations of a caption. Reading the submission guidelines. Working out what to say to a gallery.
None of that is the work. All of it eats the hours that were supposed to go to the work. That is where this earns its place.
Where it genuinely helps
- The admin moat. Applications, statements, descriptions, correspondence, listings. The paperwork that stands between you and making things.
- Getting unstuck. Thirty titles when you need one. Fifteen ways into a composition problem. It doesn’t give you the answer — it gives you enough material that your own taste can start operating again.
- Talking about the work. Many good artists are genuinely bad at explaining what they do. Having something to argue with while you work out what you actually think is legitimately useful.
- Learning something adjacent. A patient tutor for the framing, the business, the technical thing you keep avoiding.
- Research and reference gathering. Orienting yourself in an unfamiliar subject before you commit weeks to it.
Where it hollows things out
I’m going to be blunt, because I think the soft version does people harm.
If you hand it the judgment, you lose the thing. Not morally — practically. Taste is built by making thousands of small decisions and living with the consequences. Outsource the decisions and the taste stops developing. You will not notice this happening for about a year, and then you will notice it very sharply.
The default aesthetic is an average. It is trained toward the middle of everything, which means it produces competent, pleasant, entirely unsurprising work. If your work has a voice, unedited output will sand it down. This is easy to miss because average output looks fine.
The consent question is real. These systems learned from work whose creators mostly did not agree to it. I don’t think there’s a clean answer to that. I do think pretending it isn’t a question is worse than sitting with the discomfort.
A test I use
Before using it on anything creative, I ask one question: if this comes out well, will I be proud of it, or relieved?
Relief means I’ve handed over something I should have done myself. Pride means it did the labour and I did the work. It’s a crude test and it has been right nearly every time.
What this looks like in practice
For me: it writes nothing that goes out under my name without heavy rewriting. It never makes a compositional decision. It does help me name things, prepare for conversations I’m dreading, get through applications, and understand technical subjects faster than I would alone.
The paintings are still made by hand, badly, slowly, and with a great deal of swearing. That part is not for sale to a machine and never will be.
If you decide against it entirely
That is a completely respectable position and I’m not going to argue you out of it. Plenty of excellent work will be made by people who never touch this, and the people telling you that you’ll be left behind are mostly selling something.
But if you do want a way in that doesn’t cost you what matters — start with the admin. Let it fight the paperwork. Keep the craft.