You might be overthinking your prompt engineering by trying to be too professional, when all you really need is the vocabulary of a moody teenager.
The Mid Method
I just finished reading a fascinating thread from a Reddit contributor that highlights a bizarre but effective quirk in how Large Language Models process feedback. We usually treat AI like a junior employee, giving polite, constructive criticism or asking it to please revise for clarity. But this innovator took a completely different approach. The original poster discovered that using Gen Z slang, specifically the word mid, generates significantly better revisions than clear, professional instructions. When the author tested this, the results were undeniable.
The core concept the expert identified is that standard feedback prompts often lead to lazy revisions. If you tell ChatGPT try again or rewrite this, it tends to stick to the original structure and just swaps out a few synonyms. It keeps the same boring tone. However, the Reddit user found that replying with mid triggers what they describe as full panic mode. It signals a deep dissatisfaction with the quality and vibe of the output, forcing the model to ditch the generic corporate speak and attempt a total rewrite with actual personality.
💡 The Vocabulary of Devastation
The author didn’t just stop at mid. They tested a whole arsenal of single-word slang terms that seem to act as destroyers for bad AI writing. The expert points out that words like boring, cringe, basic, and especially npc hit differently than standard critiques.
Think about what these words actually mean in a semantic context. If you say something is wrong, the AI checks for factual errors. But if you say something is cringe or npc, you are attacking the soul of the writing. You are telling the AI that it sounds robotic or socially unaware. The creator notes that npc is particularly effective because it implies the response lacks agency or humanity. This forces the specific algorithms governing tone and creativity to kick into high gear to prove you wrong. It is a fascinating psychological trick, or as the poster puts it, maximum devastation with three letters.
📌 The Iterative Vibe Check
Another brilliant takeaway from this post is the strategy of chaining these prompts. The savvy professional compares this process to training a puppy, though in this case, the puppy is a trillion-parameter language model. You don’t have to accept the first revision.
The contributor suggests using a sequence of slang-based feedback to fine-tune the result. For example, if the first revision is an improvement but still not great, the author recommends saying, better but still mid. This acknowledges the progress while maintaining the pressure for higher creativity. This creates a reinforcement loop where the AI is constantly trying to escape the mid label. It turns the drafting process into a rating system, similar to how one might critique a SoundCloud rapper, focusing purely on the vibe rather than technical grammar rules.
✅ Efficiency Over Formality
The most practical insight from this industry pro is the ratio of effort to results. We often spend minutes crafting valid, detailed context for our prompts to get the tone right. We write things like, Please act as a world-class copywriter and avoid clichés.
But the person who shared this trick proved that you can control the output with single words. By using internet slang, you are signaling to the model that the context is informal, high-energy, and distinctly human. It breaks the AI out of its default helpful assistant mode, which is often overly polite and verbose. The author argues that even though the AI doesn’t have feelings, treating it like it does, and specifically, hurting its feelings with slang, yields the best creative writing results.
How to Use The Mid Prompt
Based on the creator’s findings, here is a simple way to test this on your next generic output:
1. Spot the Generic: When ChatGPT gives you a response that feels like corporate fluff or lacks flavor, do not explain why it is bad.
2. Drop the Bomb: simply reply with mid. Watch how it completely restructures the answer.
3. Escalate if Needed: If the second attempt is still stiff, try the author’s other favorites: npc (to fix robotic tone) or cringe (to fix awkward phrasing).
4. Chain It: Use better but still mid to guide it toward the perfect final draft.
This is a hilarious reminder that sometimes the most effective way to communicate with advanced intelligence is to speak like a human on the internet!
Check out the full discussion for more examples.
💡 FAQ & Troubleshooting
How does the “mid” prompting technique work?
When an AI generates generic or “corporate” output, simply replying with “mid” acts as a strong negative feedback signal. Unlike standard commands like “try again”—which often yield lazy revisions—this slang term signals mediocrity to the model. This frequently triggers a total rewrite where the AI attempts to add more personality, specific details, and effort to “impress” the user.
Are there other single-word prompts that improve results?
Yes. If “mid” doesn’t yield the desired result, you can use other dismissive terms such as “boring,” “cringe,” “basic,” or “npc.” You can also use harsher phrasing like “you suck” or “make this less dogshit.” These inputs invoke a strong response from the model, forcing it to diverge sharply from its previous “lazy” output patterns.
Is the external link in the post safe to click?
No. Do not click the link included in the original post content. It has been flagged as a potential virus or malicious site. You should ignore the external URL and only utilize the prompting advice provided in the text.
I started replying “mid” to ChatGPT’s responses and it’s trying SO HARD now
byu/AdCold1610 in