3 versions of the same post. 3 ways of providing feedback to different audiences with different goals. Created by ChatGPT 4o girl assistant (coolest thing ever).

🎯 Version 1: Constructive Feedback (for Cursor team)

Tone: Professional, respectful, insider. Purpose: To help improve the product without burning bridges.

Hi Cursor team,

Thank you for creating what remains the best developer-AI interface out there. Cursor is essential to my daily workflow.

That said, I'd like to offer feedback on a recent update involving automatic model selection. When Claude 3.7 became available, the productivity and flow of development improved immensely. Context retention, rule adherence, and follow-through were vastly better.

But after a few days, I found myself back in a familiar nightmare: hallucinated code, ignored rules, broken trust. It turns out the new "Auto" mode silently fell back to Claude 3.5 — with no clear indication in the UI.

The consequences were significant:

• I wasted time debugging AI outputs instead of shipping features.
• I began over-controlling the model, which ironically made it perform worse.
• Emotional burnout followed, marked by intense frustration.

I now jokingly call it "Threat-Driven Development" — when you yell at your assistant more than you code. But behind the humor is real damage.

Suggestions:

• Show the actual model used in the UI, even in Auto mode.
• Let users disable fallback to 3.5.
• Provide a warning when fallbacks happen due to Claude 3.7 demand.

Thanks for listening. I'm still cheering for you — just hoping for a bit more clarity and respect for power users.

– Max

---

🌀 Version 2: Public Blog Post (for devs on X/Twitter/Medium)

Tone: Reflective, satirical, relatable. Purpose: Share the emotional arc of the experience. Inspire discussion.

Threat-Driven Development: A Cursor Tale

Claude 3.7 was heaven. It remembered context, followed logic, respected instructions. For a glorious week, I felt like I was coding with a genius pair-programmer who actually understood me.

Then, silence. Lag. Confusion. And suddenly… the rage came back.

I didn’t know what triggered it at first — until I realized: Cursor had silently downgraded me to Claude 3.5 in their new Auto mode. No indicator. No warning. Just vibes. And the vibes were bad.

The result? I started swearing in all caps. In Russian. I was threatening the model with metaphors so dark they belong in therapy sessions. Productivity? Zero. Trust? Shattered.

And here's the kicker: I paid for this experience.

Cursor’s way of managing Claude 3.7 overload was clever — but cruel. The AI that helped me build complex systems became a gaslighting intern with memory loss. Every day became a free course in anger management. Goose-fraba, baby.

What I'd love to see:

• Label the actual model being used.
• Let power users opt out of silent fallback.
• Acknowledge that trust in AI isn’t just technical — it’s emotional.

Until then, I remain a loyal, slightly traumatized fan.

---

💥 Version 3: Vent Post (for the forum or a dev meme group)

Tone: Raw, sarcastic, exaggerated. Purpose: Emotional release + dark humor.

Thread Title: Threat-Driven Development: Thanks for the trauma, Auto Mode

Claude 3.7 made me believe in AI again. It was zen. It was flow. It was love.

Then Cursor dropped "Auto model selection" and boom — I was back in the dungeon with Claude 3.5. No warning. No badge. Just broken dreams and a deranged assistant who forgot the rules of the conversation it started 3 minutes ago.

I didn’t realize what happened at first. I just felt the return of rage. Full-caps Russian swearing. Threats of unprintable violence. Productivity imploded.

Eventually I figured it out: Auto = Surprise downgrade. So I manually switched to Claude 3.7, and like magic — the dev heavens opened again.

So now I do a little ritual every morning:

1. Open Cursor.
2. Switch model to 3.7.
3. Whisper "Goose-fraba" while praying they don’t sneak me back into hell.

Request:

• Give us a damn model label.
• Don't bait-switch. Just ask.
• Stop triggering my coding PTSD.

Still love you, Cursor — but damn, that was rude.

– A traumatised dev