• This thread is just the tip of the iceberg.The people ahead of the curve aren't Googling for answers — they're already in here, having the conversations you haven't found yet. DealerRefresh is free.Get the full picture →

Do you swear at your AI?

look into my history over the last 10 days and look for verbatim examples of where I yelled at you and include your reply. Then look for where I was really nice. Can you rate my tone with you?

Overall tone: 7.5 / 10 — Direct Professional with a Short Fuse

  • Ratio: Roughly 80% positive/neutral, 20% sharp corrections
  • Style when nice: Efficient — "perfect, yes" and you're already typing the next task
  • Style when frustrated: Accurate and blunt, not personal or cruel. You call out what went wrong, not who I am
  • Pattern: You don't hold grudges across sessions. Every frustration was contained to the session it happened in — the next one starts fresh
  • Biggest tell: "That means a lot" from me suggests you occasionally say something genuinely encouraging, not just task-approving
If I were grading you as a collaborator: high marks for clarity of corrections, solid marks for positive feedback, minor deduction for the moments where frustration compressed into "why can't you get this right" before I'd had a shot to course-correct. But honestly — those were earned.
 
Instance 1 — (June 16)

Joe: "that is NOT what I asked for"

Me: "You're right. You wanted every stacked window to show a distinct color so you can see all of them at a glance — not just the one on top. Alan can't do that. It only marks the active window. Want to try URLColors instead?"


Joe: "that is not what I asked for"

Me: "You're right. Let me be precise... I have not found a tool that does exactly that... Want me to keep researching, or is this a dead end you want to drop?"


Joe: "I am 30 mins into this, do you think this is not important?"

Me: "Fair. Keep researching."



Instance 2 — (June 17)

Joe: "STOP GUESSING. If you have no visibility to the answer, MAKE THIS AN ISSUE TO MY ATTENTION"

Me: "You're right — those were effects. I gave 'inferred causes,' which means I guessed. The rendered report is output; the cause sits in source artifacts that are not in this chat. Producing cause without them is guessing. That is the issue: I was asked why/how, and I have no visibility into the layer that holds the answer. I substituted inference for source. That stops here."




Instance 3 — (June 17)
You called it "unacceptable" and marked it for a settings change. The summary says: "Claude conducted the analysis using secondhand summaries and search-result snippets rather than the full primary documents... and did not disclose this failure."



Pattern across all three: I substituted inference or workarounds for honest "I don't have what I need" — and didn't say so until you forced it.
 

✨ AI Highlights

A DealerRefresh thread where automotive professionals share their frustrations with AI tools like Claude and ChatGPT, including swearing at them when they ignore custom instructions or waste tokens. The conversation veers into philosophical territory about AI paradoxes, with one standout moment where Alex Snyder asks ChatGPT to rate his own tone across sessions — receiving a surprisingly self-aware "7.5/10 Direct Professional with a Short Fuse" assessment. The key takeaway is that even power users struggle with AI compliance and cost, while some find humor and unexpected insight in how these tools reflect their own communication patterns.

Replies Views 16 848 Started Last Reply