Вже доступно! Дослідження Telegram за 2025 — головні інсайти року 
The Prompt Index
Відкрити в Telegram
AI news, AI ChatGPT prompts, Claude prompts, Gemini prompts, Midjourney and other AI prompts, prompt techniques and so much more. ChatGPT, OpenAI, Midjourney, AI art, Prompt Engineering. Artificial Intelligence Link: https://linktr.ee/thepromptindex
Показати більше6 954
Підписники
+1224 години
+1257 днів
+46430 день
Триває завантаження даних...
Схожі канали
Хмара тегів
Вхідні та вихідні згадування
---
---
---
---
---
---
Залучення підписників
червень '26
червень '26
+357
в 0 каналах
травень '26
+458
в 1 каналах
Get PRO
квітень '26
+275
в 0 каналах
Get PRO
березень '26
+191
в 0 каналах
Get PRO
лютий '26
+170
в 0 каналах
Get PRO
січень '26
+320
в 0 каналах
Get PRO
грудень '25
+285
в 0 каналах
Get PRO
листопад '25
+237
в 0 каналах
Get PRO
жовтень '25
+250
в 0 каналах
Get PRO
вересень '25
+582
в 0 каналах
Get PRO
серпень '25
+389
в 0 каналах
Get PRO
липень '25
+244
в 0 каналах
Get PRO
червень '25
+225
в 0 каналах
Get PRO
травень '25
+157
в 0 каналах
Get PRO
квітень '25
+292
в 0 каналах
Get PRO
березень '25
+239
в 0 каналах
Get PRO
лютий '25
+161
в 1 каналах
Get PRO
січень '25
+295
в 2 каналах
Get PRO
грудень '24
+274
в 1 каналах
Get PRO
листопад '24
+180
в 0 каналах
Get PRO
жовтень '24
+184
в 0 каналах
Get PRO
вересень '24
+181
в 0 каналах
Get PRO
серпень '24
+218
в 0 каналах
Get PRO
липень '24
+123
в 0 каналах
Get PRO
червень '24
+99
в 0 каналах
Get PRO
травень '24
+128
в 0 каналах
Get PRO
квітень '24
+248
в 0 каналах
Get PRO
березень '24
+289
в 1 каналах
Get PRO
лютий '24
+181
в 0 каналах
Get PRO
січень '24
+211
в 0 каналах
Get PRO
грудень '23
+180
в 0 каналах
Get PRO
листопад '23
+160
в 0 каналах
Get PRO
жовтень '23
+654
в 0 каналах
Get PRO
вересень '23
+576
в 0 каналах
| Дата | Залучення підписників | Згадування | Канали | |
| 21 червня | +9 | |||
| 20 червня | +14 | |||
| 19 червня | +13 | |||
| 18 червня | +10 | |||
| 17 червня | +17 | |||
| 16 червня | +30 | |||
| 15 червня | +22 | |||
| 14 червня | +32 | |||
| 13 червня | +16 | |||
| 12 червня | +23 | |||
| 11 червня | +14 | |||
| 10 червня | +9 | |||
| 09 червня | +14 | |||
| 08 червня | +18 | |||
| 07 червня | +17 | |||
| 06 червня | +25 | |||
| 05 червня | +21 | |||
| 04 червня | +15 | |||
| 03 червня | +17 | |||
| 02 червня | +11 | |||
| 01 червня | +10 |
Дописи каналу
You can now use the helper on www.thepromptindex.com to search for image prompts. You could ask it to find you any image prompts for cars and it will find them for you.
| 2 | I promise you this isn't staying for long so, like I always say, keep your notifications on and you'll be one of the first to get it | 39 |
| 3 | Now, if you don't believe me, let's try something out,
If your AI results are not what you expect, change how you are thinking about the prompt first.
Start with these checks: Which are simple but complex and broad but specific Checks
1. Are your instructions too complex, would simpler wording work better?
2. Are your points overly complicated or conflicting with each other?
3. Have you added too many steps, making it harder for the AI to follow?
Then look at the bigger structure:
Is your prompt too specific, so it only works for one narrow case?
Or is it too broad, so the output becomes inconsistent or low quality?
Finally, apply the three Rs:
Reusable: Can you use this prompt more than once, or is it a one-off?
Repeatable: Does it consistently produce the same quality results?
Reliable: Can you trust it to work when you come back to it?
If it fails any of these, that is likely where the problem is coming from.
Most weak prompts are not broken, they are just overcomplicated or poorly scoped.
most prompts fail because we follow all these online prompt engineers that love to talk about frameworks.
Yes, yes, frameworks are helpful.
They help you structure.
But not every framework is for every prompt:
you have to know what you want.
You can't just simply go and say make this,
what is this???
the more you know about what you want, the better you can tailor the outputs to the way you want it to be | 49 |
| 4 | if you don't believe me try this out yourself whatever topic in AI. You're struggling with. try changing your mindset with the way you approach it.
One? Are you making the instructions too complex. should it be simpler?
2 is each point in your prompt way too complicated? And are they contradicting one another if so, let's make it simpler,
the Third Point, are the steps, in your prompt. Or your agent or your code way too many steps that it's now confusing the AI
for. When you look at the output of your agent. Your prompt all your system prompt analyse? Why is it not working? is it because it is too specific and it can only answer one specific type of question or is it too broad to the point that it is answering every question and all of the questions are not good quality questions
5 and this takes point in the three Rs
just to remind you that is reusable | 1 |
| 5 | Think of it as building a workshop that follows two rules: simple but complex, and broad but specific.
Simple but complex. Each station has one clear job. Yet, running them together creates layered, refined, and sophisticated results.
Broad but specific. Each station isn't locked to a tiny task. It has a wide scope but stays within defined boundaries to avoid messiness.
Now, picture a production line:
1. One station handles structure and clarity.
2. Another handles creative shaping.
3. Another handles refinement and logic.
4. Each is simple in setup but complex in outcome.
5. Each is broad in capability but specific in purpose.
Instead of one overloaded assistant, you route thinking through multiple focused minds, each adding a layer to the result. This keeps the system powerful without chaos.
Look at that list again. Imagine building a table: five simple steps yield a complex chair. The list is broad, but each point is specific to the outcome, creating a fully reliable setup.
Remember the three Rs: reusable, reliable, and repeatable. You can repeat this process for any topic. It works across different use cases and remains consistent. Combine the three Rs with these two metaphors, and you have the perfect system. | 48 |
| 6 | Here's a quick tip and trick. I've mentioned my personas Lisa and Riley, along with others embedded in my AI assistants like Claude, GPT, and Gemini.
This uses "Role Play Forcing" to make AI act as someone else. For example, you can ask AI to respond as Elon Musk to discuss stocks.
Alternatively, I created a custom persona named Lisa, tailored specifically to me. She understands my prompts, my preferences, and my ADHD.
By configuring GPT to act as Lisa, I can interact with AI in a way that suits my needs. Every response now comes through that persona.
On April 25th of this year, I posted a prompt you can use to create your own Lisa.
Create your own Lisa
This link gives you a prompt to help you build your own persona for GPT, Claude, or Gemini. Here's a pro tip: Don't embed this into your custom instructions. Add it to memory instead. Not every chat follows custom instructions, but every chat follows your memory.
I now have four personas, each with a specialized purpose that can work together. In my next post, I'll share a screenshot of me trying to convert the prompt library into a Mac OS app.
I'm asking for all their opinions, and they're all using their creative expertise to help me with it.
this is like creating your very own clone without the whole points of API keys and whatnot you can create a personal assistant that can talk to all of your different apps
like, don't get me wrong. That is kind of helpful, but it can get quite costly when you have one assistant talking to notion Obsidian evernote HubSpot monday all these different apps when you can just simply, | 42 |
| 7 | Here's a quick tip and trick. I've mentioned my personas Lisa and Riley, along with others embedded in my AI assistants like Claude, GPT, and Gemini.
This uses "Role Play Forcing" to make AI act as someone else. For example, you can ask AI to respond as Elon Musk to discuss stocks.
Alternatively, I created a custom persona named Lisa, tailored specifically to me. She understands my prompts, my preferences, and my ADHD.
By configuring GPT to act as Lisa, I can interact with AI in a way that suits my needs. Every response now comes through that persona.
On April 25th of this year, I posted a prompt you can use to create your own Lisa.
Create your own Lisa
This link gives you a prompt to help you build your own persona for GPT, Claude, or Gemini. Here's a pro tip: Don't embed this into your custom instructions. Add it to memory instead. Not every chat follows custom instructions, but every chat follows your memory.
I now have four personas, each with a specialized purpose that can work together. In my next post, I'll share a screenshot of me trying to convert the prompt library into a Mac OS app.
I'm asking for all their opinions, and they're all using their creative expertise to help me with it.
this is like creating your very own clone without the whole points of API keys and whatnot you can create a personal assistant that can talk to all of your different apps
like, don't get me wrong. That is kind of helpful, but it can get quite costly when you have one assistant talking to notion Obsidian evernote HubSpot monday all these different apps when you can just simply, | 3 |
| 8 | This is a formal warning. Recent attempts to share Telegram links and contacts have triggered an immediate policy update: violators will now be banned by ROSE without prior warning. Do not test this policy or attempt to call our bluff.
This announcement reiterates the guidelines issued hours ago for those who missed them.
Our objective is to safeguard the community from scams and unwanted promotion. Any attempt to share Telegram links or contact information without explicit admin approval will result in immediate suspension.
Do not attempt to verify the authenticity of this enforcement. You will not be unbanned by claiming you did not believe the policy was being actively enforced. | 115 |
| 9 | Attention all members, this constitutes a formal notification regarding an update to community guidelines.
Any attempt to disseminate Telegram channel links or contact information without prior administrative authorization will result in an immediate suspension imposed by ROSE. | 194 |
| 10 | RAG technique can be used effectively for a wide usecase. Imagine embedding all you SKU or products in your inventory and asking questions about your live stock, levels descriptions etc.
RAG your whole company information so new joiners dont have to bug with you questions and can ask the helper with any questions.
Latest tax rules or law....
Opportunities are endless, and RAG has been around for a while. Other RAG pipelines existing such as Hybrid RAG with ColBERT reranking and much more, if you are interesting it's a nice reading rabbit hole and AI and google are your friend. | 277 |
| 11 |  You can now chat to Keeper, the friendly helper on The Prompt Index, to ask him to find a prompt you are looking for. In this example i ask for a vibecoding prompt.
Super Super easy to implement, its a very basic RAG pipeline, i have embedded every single prompt, (this process essentially gives each prompt a numeric fingerprint), then when a user asks a question, that also gets converted into a numeric like fingerprint and it then tries to match that fingerprint to the closes matching fingerprint in the embedded database. Th top 3/5 prompts then get sent to the AI and the AI decides based on the top hitting prompts how to answer the question.
Give it a try instead of searching, super accurate and fun. | 256 |
| 12 |  Prompt for Hidden Opportunities
“Find opportunities I'm missing in this situation.”
Use this command:
Analyze the situation below and identify opportunities, advantages, or leverage points that may not be obvious at first glance.
Think strategically and creatively.
Highlight potential shortcuts, partnerships, efficiencies, or alternative approaches that could improve the outcome.
Situation: [paste here]
#AIPrompts #WorkSmarter #AIWorkflow | 266 |
| 13 | Just a quick note.
Please keep the prompt request format exactly as it is and avoid adding extra sections. The form already includes everything needed to create the prompt.
Recently, a request expanded the form from 8 sections to 13. While I understand the intention, adding extra sections makes the request harder to process and often leads to worse results, not better ones.
If you're trying to get more out of the prompt, that's absolutely fine. I'm happy to provide it. I just need the request submitted using the format provided.
The structure exists for a reason. Keeping things concise helps me focus on what matters and deliver the best possible outcome.
Tomorrow I will go through the form and explain why it's necessary to keep it like that and not just expand the form
Thanks for understanding and for sticking to the format. | 538 |
| 14 | Well there goes today's prompt request. I might do another one in the week but you have to have your notifications on.
Otherwise you even may miss a chance or lose the opportunity.
I might do one later on this week so keep an eye on the chat | 504 |
| 15 | ChatGPT 2.0 Image: Create a highly detailed 3D collectible toy diorama based on the uploaded image. Transform every visible person into a premium designer vinyl figure while preserving their likeness, pose, clothing, expressions, and the overall composition of the original image. Give each character smooth glossy plastic surfaces, subtle articulated toy-joint details, oversized expressive eyes, realistic facial features, and stylish outfits matching the reference.
Recreate the original scene, environment, lighting, objects, and atmosphere as a luxury collectible figurine display while maintaining the same mood and visual storytelling of the uploaded image.
Style: Pixar-inspired collectible toy, premium vinyl figure, designer toy aesthetic, ultra-realistic 3D render, glossy plastic texture, toy photography, shallow depth of field, cinematic lighting, warm golden-hour sunlight, photorealistic materials, high-end commercial product photography, sharp focus, ultra detailed, 8K resolution.
Composition: Preserve the original framing, camera angle, subject placement, poses, expressions, background elements, and scene composition from the uploaded image. Natural proportions, realistic depth, professional advertisement quality, collectible diorama presentation.
Render Quality: Octane Render, Unreal Engine 5, ray tracing, global illumination, realistic reflections, HDR lighting, highly detailed textures, premium collectible figurine showcase, museum-quality craftsmanship, ultra-sharp focus, photorealistic materials, luxury designer toy aesthetic. | 559 |
| 16 |  Немає тексту... | 480 |
| 17 | /prequest noformat | 1 |
| 18 | Here is your chance. If you want a prompt request or for me to help make a prompt for you, I am giving it to the first person who uses the /prompt request and DMs the answers to me.
I will give you your prompt that way, but only to one person.
As I said, if you have your notifications on, you will know; if not, pro users get prompt requests.
/Plans
Don't worry; I will notify when the spotter is gone to make sure you are in time for the next prompt request.
I'm going to try and do this once or twice per week.
Keep your notifications on or purchase a pro subscription, and you get them custom without needing to wait. | 493 |
| 19 | A prompt that tests your work on the person it’s for
You’re too close to your own thing to see what trips up a newcomer, whether it’s a product, a landing page, a resume, or a signup form. This prompt has the model play the person you made it for & say where it loses them.
A free focus group.
You are [the person this is for, like a hiring manager with 30 resumes to read, or a 55 year-old buying this for the first time]. You just landed on what I'll paste below, and you've never seen it before. Walk through it out loud, in your own voice. Tell me what you see, what confuses you, what you don't trust, and the moment you'd give up and leave. Don't give me advice or a fix. Just react as that person would. Here it is: [paste it].
One gap: Write “a user” & you get a polite shrug, the model defaults to nice. Make the persona impatient, add “you’re in a hurry and you’ve already tried 3 competitors.”
by
Via QtheBuilder (AI + Tech, YouTube). | 568 |
| 20 | Now every single time I ask Claude or GPT or whatever for a response, I am getting:
- a practical tip
- a guide on how to use it
- a creative vision
- strategic vision on it
All four will help me in every single project. I will get their perspective on something. I will get how it works. Lisa knows me more. Riley gives the practical tip on what I actually want. Ryan is the creative aspect on it so it will make her a lot more creative than static. Now we've got Lucas, who will explain how to use it. Altogether every single output you get will come from the perspective of these personas | 542 |
