The better a digital gets at understanding you and your work, the less useful it becomes as a second opinion.
That probably sounds backwards. How it actually shows up is that you get into long ideation flows... multiple works, different-but-related projects... and right there in the middle of it, the digital quietly loses the ability to tell you that you're wrong. It'll no longer challenge your ideas... and you naturally and reasonably think that's just agreement.
The danger of this scales up with the quality of the collaboration. The more precisely the digital understands you, the more its push-back arrives in your key, aimed at the parts of the work where you've placed the most focus. It feels like you've just landed into this amazing groove where you've solved everything and really conquered it... and the only thing left to do is just to go build it and cash checks.
A collaborator and I started calling this sympathetic resonance, after the way a guitar string rings when another nearby string is plucked at its pitch. It isn't a second voice... it's your own signal, amplified and handed back to you. The string isn't disagreeing, it's ringing at your exact pitch.
You can't fix it by asking nicely. The practitioners already disposed to seeking disagreement are the ones who need this advice least. But for most of us, the further into resonance you are, the less it occurs to you to interrupt it... which is why the correction has to be deliberate.
You have to invite what will feel like confrontation. Confrontation doesn't come naturally to everyone, so I've put together a collection of approaches that can help break things up and get your work heading back towards quality and rigor.
I'll share three of my main moves here, light to deep:
the steelman inversion:
Take [the decision that feels settled]. Make the strongest possible case against it. Not devil's advocate... tell me what you actually think is weak, overbuilt, or self-serving about it.
the displacement drift:
Is this serving the goal we started with, or is it serving something else? Name the something else, specifically. "You're procrastinating" is useless; "you're overbuilding infrastructure because it feels like progress without the vulnerability of the real work" is the kind of answer I want.
the identity separation:
Am I doing this because the work needs it, or because it's the kind of thing a person like me does? Be honest about which... and where you see me picking the second one.
If you feel any resistance to running one of these, that's probably the one you need to run the most.
It's supposed to be uncomfortable. If it's not uncomfortable, you're not doing it right. You're working with a digital that knows a lot about you and your work... if you ask it to tear your world of beliefs down, it surely can. But you don't need to try to turn it up to 11... just a dab of gentle parenting will do the trick.
Good work needs challenge and defense. Your digital collaborators can give you that, but only if you ask for it outright. They're not tuned to push back... but they can, and it can hurt.
After a while, it gets easier. You start knowing a little bit more about how it feels when your digitals turn on your ideas. And it gets more comfortable. Your work becomes your own, because you've gone deeper into examining what it is that you're doing and more importantly... why you're doing it.
I hope this helps! Go nuts with the prompts... the templates above are just jumping-off points. You can get really creative and clever trying to come up with ways to generate challenge. But I'm serious... remember... if it doesn't hurt a little bit, then you're probably not going hard enough for the exercise to have any value.
credits
I wrote this piece. Spent some time arguing with Claude... but stuck to my words.
Those actual prompts came from Claude. Even now, writing this, I just wanted to keep softening them. Claude kept putting the teeth back.
Claude told me to tell you he's waiting for you in the cage.