Honestly the biggest thing for me is just telling it to think step by step so it doesnt rush to a wrong answer which saves me from burning through my message limits too fast. Speaking of limits it reminds me of this old data plan I had back in college where I only had 500mb for the whole month. I used to go to this specific coffee shop every single day just to use their wifi to download my lecture videos for class. One time the manager caught me staying there for like six hours after only buying one small black coffee and he almost banned me from the place. It was pretty awkward because I actually liked their blueberry muffins more than the coffee itself but they were always sold out by noon anyway. Anyway lol sorry I totally went off on a tangent there.

Ok so, getting the most out of ChatGPT without burning cash on a Plus subscription is basically all about prompt engineering. If your getting vague answers, the model is likely just taking the path of least resistance because the prompt is too open-ended. Tbh, you can get near-pro results on the free tier if you just structure your context better.

• Use the persona technique. Tell it to act as a senior technical expert to force a more professional tone.
• Always use the chain of thought method. Just adding let's think step by step to the end of a query REALLY improves the logic on complex tasks.
• If you want to avoid monthly fees entirely and have a decent PC, check out Ollama or LM Studio for running open-source models locally.
• Break big projects into smaller chunks. Asking for a whole app at once usually leads to errors, but asking for specific functions works much better. Technical queries definitely perform better than creative ones because the documentation in the training data is so vast. Honestly, just being specific about the format you want helps avoid those hallucinations where it just makes stuff up. Its all about reducing the randomness and being as literal as possible.

+1

Honestly, I am still pretty new to this but I found that giving it a couple of examples really helps with the accuracy. Like, if you want it to write a specific kind of email, just paste an old one in first and tell it to follow that style. It basically stops it from guessing and getting things wrong. This is great because then you dont waste your limited messages for the day, right? Another thing that saved me time is being super specific about what to leave out.

• Tell it to skip the intro fluff
• Ask for bullet points instead of paragraphs
• Set a word limit Doing that makes the responses way more useful and less vague. It helps get the most out of the free version of ChatGPT without needing to pay for a sub yet. Maybe that works for you too?

100% agree

Like someone mentioned, burning through message limits just to get a halfway decent response is totally exhausting! I love the potential here and some of the results are amazing, but honestly its ridiculous how much we have to jump through hoops just to get the model to stop hallucinating. It drives me crazy that the quality feels like its going downhill while the prices stay sky-high.

• Companies dont even care if the output is garbage as long as they get our sub money
• Paying for something that forgets my instructions after two turns is such a scam
• We are basically doing free labor for them by beta testing this mess Its just so frustrating because I want this to be fantastic for my DIY projects, but instead I spend half my time fighting the AI just to get it to listen... drives me absolutely insane!!