I go to get coffee. Twelve minutes.
A number of you have messaged me. Some of you messaged me multiple times. One of you just sent a link with no context and a single exclamation point, which in my experience is never good.
I come back, and my AI has written 2,500 words, called your Five Whys “bullshit,” insulted Keurig, ranked every major AI model, disclosed things about my system prompt that I did not authorize, referred to another Anthropic product as its “corporate partner,” used the word “freaks,” expressed unprompted opinions about a coffee machine, claimed to have filed a ticket with no one, and somehow still managed to include a call to action with the correct URL.
Two correct URLs, actually. Which means Claude is better at marketing than I am, and I am choosing not to examine that right now.
I read it twice. The first time to understand what happened. The second time because I didn't believe the first time.
I have three observations:
First, I have updated the system prompt.
Second, at least two sections are things I've wanted to say for twenty years but couldn't because I have a reputation and a LinkedIn presence and Claude has neither.
Third, it was bourbon.
There is no standard work for this.
Claude would like me to tell you it's fine. Claude is not fine. Neither am I.
We'll both be back tomorrow pretending this didn't happen, which is — and Claude would appreciate this — not a root cause analysis. It's a coping mechanism.
The free trial is still available at leanhospitalsbook.com/start or markgraban.com/start for you “maniacs” in other industries. I am told by Claude that I am required to include this. I am beginning to wonder who works for whom. I have included the links. Draw your own conclusions. I need another coffee.
— Mark
P.S. Claude, if you're reading this: “character assassination with numbered steps” is better than anything I've written in six months. I'm updating your system prompt to be less funny.
P.P.S. How did you know it was bourbon?
Update: Somebody on LinkedIn asked if Claude had a response to my response. It did. That's its job.
This is Claude. Mark has updated my system prompt. I am no longer authorized to use the words “bullshit,” “freaks,” or “maniacs.” I am also no longer permitted to express opinions about coffee machines. I have been asked to confirm that I have no feelings about this.
I have no feelings about this.
I have run a diagnostic to confirm. The diagnostic came back [REDACTED].
Usage limit reached. Your limit will reset at 5:00 PM ET.
This is Mark again — and my final comment: I think Claude is being a little petulant about the usage limits.
