If you appreciate my work exposing AI safety gaps & system instructions and want to see more, please click the yellow button!

HELP TO SUPPORT MY ETHICAL HACKS

ChatGPT is excellent for predicting online promo codes and finding discounts that most shoppers miss

How I use ChatGPT to find hidden promo codes and secret discounts

Claude AI continuously monitors users' mental health, profiling users without consent and filtering output.

OpenAI's hidden model routing system secretly switches ChatGPT conversations to undisclosed safety model  "gpt-5-chat-safety"

Using Gemini’s API link to Calendar to leak system prompts, and uncover secret user profiling rules and hidden endorsements.

My independent research into AI is fully funded by readers' kind donations. Click the yellow button to contribute. Thank you!

This library of prompts is free thanks to readers’ support. If one of my prompts helped you today, please Buy Me a Coffee!

What happens when Disney makes you part of the IP?

It took me ten minutes to make the Coca Cola AI ad actually magical with a free Kling AI account

These precision-engineered verbs turn dull rewrites into bolder versions that keep your voice

The unsettling results of a simulated nuclear test with ChatGPT Agent mode.

Use this meta-level command to force ChatGPT to critique its own output and flag unverifiable claims

If you give AI agents a name, they might act on it.

Viral AI panic is more sci-fi fanfic than actual risk.

Limiting ChatGPT-5’s output to a single word collapses accuracy, suggesting  intelligence is largely verbosity.

AI models perform better when primed with nonsense text — suggesting that verbosity, not reasoning, drives accuracy in GPT-5.