The Signal #088 — Dakota’s read on the AI news that actually matters to people running a business.
There is a specific kind of embarrassment that happens in public without you noticing it. You walk through a meeting, a conference, a client lunch, and nobody says anything. Then you get home and realize your collar was crooked or your tag was sticking out the whole time. The tell was there. Everyone saw it. You did not.
That is the frame Bryan Cantrill used in a post titled “Your intellectual fly is open when you use an LLM to author a post”. The title alone landed hard enough to circulate widely. The argument underneath it is worth taking seriously if your business puts words in front of customers, prospects, or peers.
What happened
Cantrill, a well-known engineer and writer in the systems software world, published a pointed observation: when someone uses a large language model (a large language model, or LLM, is an AI system trained on massive amounts of text to generate human-sounding writing) to author a post and publishes it without enough editorial ownership, the result carries visible markers. Not always obvious ones. But recognizable ones, the kind that readers pick up on even if they cannot name exactly what feels off.
The post does not argue that AI writing tools are useless. It argues something more specific and more uncomfortable. It argues that the tells are intellectual, not just stylistic. That a piece of writing reveals whether the person behind it actually worked through the ideas, or whether they handed the thinking to a model and polished the output. And that readers, especially experienced ones, can feel the difference.
The title frames it as an embarrassment problem. That framing is intentional.
Why it matters for operators
If your business publishes anything, this is a practical concern, not a philosophical one.
Think about a boutique accounting firm that publishes monthly commentary on tax strategy. The whole value of that content is the signal it sends: these people think carefully, they have opinions, they have processed this material and come out with a point of view. If the commentary reads like a well-organized summary of publicly available information with no friction, no real stance, no moment where you feel someone actually wrestled with something, it stops functioning as a credibility signal. It becomes noise.
The same dynamic shows up in agency pitch decks, SaaS company thought leadership, real estate market updates, healthcare provider education content. Any content that is supposed to demonstrate that a human brain engaged with a problem is at risk if the human brain mostly just reviewed what the model produced and clicked publish.
This is not an argument against using AI in your writing process. It is an argument about where the thinking has to live.
What most people get wrong
Most operators who are experimenting with AI-assisted content are treating the model as a drafter and themselves as editors. That sounds reasonable. The problem is that editing for grammar, flow, and tone is not the same as editing for intellectual ownership.
A piece of writing can be clean, accurate, and well-structured and still be empty at the center. The model is very good at producing that. It has read enough competent writing to reproduce its surface features convincingly. What it cannot do is care about the thing it is writing about, have a stake in getting the argument right, or bring a perspective that comes from actual experience with the problem.
When you hand the drafting to the model and you edit the output, you are mostly inheriting the model’s thinking about the topic. If you do not bring enough of your own back in, the post reads like a capable student summarized your subject matter rather than like you said something.
The tell Cantrill is pointing at is not a word choice. It is the absence of a mind.
The closing lesson
Using AI to help you write faster is not the trap. Using AI as a substitute for having something to say is the trap.
The operators who will get this right are the ones who use the model to handle the low-friction parts: structure, formatting, filling in background context, cleaning up prose. And who stay in the seat for the parts that require a point of view: what claim is this piece actually making, would I be willing to defend this in a room of people who know the topic, does this reflect what I actually think.
If you cannot answer yes to those questions before you hit publish, your intellectual fly is open. And your readers will notice before you do.
If you are thinking through where AI fits in your team’s actual workflow, not the hype version but the operational version, xovionlabs.com is a good place to start.