Confessions of an Unrepentant Slop Snob

A framework for thinking about when AI involvement is additive or a violation

This is a peek into the research behind our “AI Norms and Values” docs, which we developed over the summer and recently shared on our site.

Earlier this year, I was spending a lot of time stewing over why everyone around me seemed so frazzled and on edge.

Half the company was spitting mad about all the slop they were getting. Instead of receiving five crisp bullet points, they were getting twenty-page docs full of padding and slop. They would ask a colleague a question, and the reply would begin with “Claude says…” These folks felt like their time and attention were being not just taken for granted, but actively abused.

The other half of the company felt equally injured. They were working faster and delivering better outcomes than ever, and wasn’t this exactly what we had asked of them? They were more upset about the fact that some people hadn’t updated their workflows in years. Of course you can’t keep up if you aren’t willing to adapt, they protested.

Everyone was mad. One side wanted to place limits on AI (“I am so sick of reviewing docs the sender didn’t even read!”). The other side wanted us to force people to use AI (“I am so sick of reviewing docs riddled with basic errors that a single pass would have caught!")

I was, at different times, in both camps. But the thing that bothered me most was how often I kept hearing the word “dehumanizing”. People said they no longer felt like they had a human connection with their coworkers anymore. This was new.

On the one hand, AI is not special. On the other hand…is it?

My first instinct was to say that tools are tools. The quality of the work is all that matters — outcomes are all that matter — not how the work was made. Do our customers care if we use AI or not? Probably not. They care a lot about the quality of the product and whether it meets their needs, not so much about how we made it.

So maybe we all just need to be ruthlessly outcome-oriented. Build the best thing we can, as fast as we can. AI is a powerful and versatile tool, so we should use it wherever we can to make our work better and do it faster.

Is it really that simple? I had nearly convinced myself that it was, when I noticed how contradictory my own behavior had become.

Meanwhile, behind the scenes, I begin harshly judging everyone who sends me AI slop

At the same time I was repeating “it doesn’t matter how it was made, it matters how good it is” every day, I was developing a violent disgust reflex for AI-generated text on the side.

I’m not sure exactly when it happened. By spring, I was snapping at the tendons of any poor soul who showed up in my inbox with an “I’d value your take on this” or “the call most leaders still won’t make”.

It started with DMs and emails, but it didn’t stop there. By summer, my reactive rage-response to AI-generated text had spread to include most forms of writing. If I’m reading a newsletter and I start to sense AI-isms, I delete and unsubscribe. If I’m reading a blog post, I close the tab; if I’m on social media, I unfollow or unfriend.

More importantly, I judge them. Yes, I look down on them. If they don’t care enough about their own point of view to do the work and refine it themselves, if they don’t care enough about me to write me a note, then why the fuck should I give them a single morsel of my precious attention?

Language does many different jobs for us

Writing is thinking on paper, as William Zinsser once said. Writing is language, encoded for posterity. But language, and writing, do many different jobs for us.

Language evolved as a way to connect — person to person, mind to mind, one mind to many. In software, for example, we convert natural language into bits and bytes that computers can use to do math on. Lawyers convert language into legal text and taxonomies. The technical and professional worlds are awash in dialects where language has been abstracted from its roots as an emotional and relational tool and given functional, depersonalized meanings.

In many of these contexts, substituting AI-generated language can be wholly acceptable. No one blinks an eye if you use structured data generated by AI, or a formal proof generated by AI (as long as it’s accurate). The situations where the use of AI tends to land jarringly, causing frustration, rage, even a sense of betrayal, are the ones where the value of the communication is less abstract, and more personal or relational.

Every job language does for us is either functional or relational, or some combination of the two, and knowing which one you’re in the middle of tells you a lot about whether AI belongs there, and how it’s likely to be received.

Does the value lie in the idea itself, or the fact that a particular person said or thought something?

It’s less about how much AI is being used, and more about in which contexts people are using AI, and secondarily whether or not they disclosed and I consented to it being used there.

Sometimes, yes, the quality of the work is all that matters. If you and I are collaborating on a document or a diff, all our collective comments and edits are in shared service of making the ideas better. It isn’t about whose idea it was or which tools we used, we are just iterating and improving until it’s as good as we can make it. The use of AI here is just another tool, one of many.

Other times, the value of a piece of writing derives from the fact that a particular person said it, thought it or felt it, or its value is grounded in your relationship. Why do you care more about what your skip level says about your performance than you would care about reading the same advice in a book? Likely because this is someone you know and respect, someone with influence over your career prospects, someone who knows what you’re capable of and has a vested interest in helping you succeed. In these circumstances, people expect to hear your voice; if they don’t, they may invent all kinds of terrifying reasons why.

The more personal the interaction, the more AI-generated text can cause a loss of trust

When I started talking to my coworkers, trying to figure out why they were so angry and frustrated all the time, one thing I heard over and over was, “I asked for my colleague’s opinion, and they sent me back a Claude snippet. I wanted to know what THEY THOUGHT.” When someone wants your opinion, AI generated text registers as a violation.

Another common source of friction was performance reviews. “My performance review was obviously written by ChatGPT. Did my manager even read it, or just push a button and spit it out?” We encourage our managers to use AI to develop systems that help them become better managers, but this is a clear risk of using AI to aid in the writing or editing of reviews: if your voice is lost in the process, it may destroy trust between you.

The more any interaction is personal or relational, the more the use of AI in that context tends to cheapen it and degrade trust, unless AI has been specifically invited into that relational context.

The slick, uncanny valleyness of AI communication

AI is just a tool, and we should use it anywhere and everywhere we can to do better work, faster, and deliver better outcomes for our users. But relationships matter. A respectful environment matters too. And whereas lines of code are indifferent to their origin, and can be validated by harnesses and tests, people are intensely attuned to the language people use with them. Relationship maintenance cannot be automated.

Yes, AI is just software. But it is special in one way: the slick, sycophantic, uncanny valleyness of the way it communicates. Human-like, but not human, which is somehow vastly more alienating than messages that are plainly automated. This is why communication that sounds like AI is so degrading to trust.

In the absence of universal norms, conventions will do

A lot of things about communication that used to be clear, no longer are. And a lot of hurt feelings, anger, and frustration are roiling about in the breach. Old norms no longer apply, and new norms are not yet broadly developed or agreed upon.

At times like these, having an agreed-upon convention or standard, any convention or standard, can really help.