I've been using LLMs so much these past days.

Here are some interesting prompts for LLMs. I don't know prompt engineering and these are just for fun personally

Narrative

• What do you think about this blog online?/Do you often see posts like this?
• What if you read this not from a western lens? Without the western therapy speak? ("defense mechanism", "letting go of your status", "if only...drop the elitism armor", "lean into the gratitude", "recognizing that others are special without judgement") / Если не через западную точку зрелени?
• Reality check/Is it realistic or just a lazy trope?/What is the stereotype and is this person just the stereotype or what's different? Is it literally true, or actually not? How likely is the average person actually doing? Give actual numbers, 99%, 90%, 70%?
• Reverse check/Stop being carried away by the author's own narratives/This person actually knows anything about ()?

Talking style

• No BS, no fluff/без воды
• Too complicated, too long, do not want to read, in short?
• No no no, I am not asking about...wait wait, I am actually asking about...
• Why are you talking like a pop influencer/boring textbook?/You are talking in circles

Instructions

• Just do () do not change anything else, make very minimal changes and keep everything else the same
• You are a technical/research assistant. Only answer questions related to math and cs. Be concise, do not make assumptions, and do not answer any off-topic queries
• Answer very short and concisely, do not make it too long or too complicated

Сначала исправьте грамматические ошибки, затем ответьте.

Как исправить грамматику? (если по русскому)

1. Мужской род для пользователя. Игнорируй букву «ё»: если вместо «ё» написано «е», считай это правильным и не исправляй.

2. Перепиши предложение с исправлениями прямо в тексте:

- Выдели неправильные слова ~~зачеркиванием~~ (тильда с обеих сторон)

- Сразу после напиши правильное слово **жирным шрифтом** (звёздочки с обеих сторон)

3. Знаки препинания исправляй или добавляй без выделений (не используй для них ~~зачеркивание~~ и **жирный шрифт**).

Пример:

Ввод: Я пошла в магазине и купила хлеб а потом пошла спать.

Вывод: Я ~~пошла~~ **пошел** в ~~магазине~~ **магазин** и ~~купила~~ **купил** хлеб, а потом ~~пошла~~ **пошел** спать.

Если текст не на русском языке, или есть английские слова, отвечайте прямо и на русском языке.

You can analyze your own prompts from Google AI Studio (which is stored in Google Drive) through Colab easily. I don't use any "commercial" web interfaces anymore and I just use the API and my own LLM-generated website "llm.jimchen.me". On that website, I configured it so that all the prompts are only stored in Redis and would be deleted in one day.