Available now! Telegram Research 2025 — the year's key insights 
Akbari’s Channel
Open in Telegram
Thoughts from Mahdi Akbari(@BlackestWhite) X: https://x.com/akbarithm Blog: https://blackestwhite.github.io
Show more443
Subscribers
-124 hours
-37 days
+430 days
Data loading in progress...
Attracting Subscribers
June '26
June '26
+15
in 1 channels
May '26
+305
in 3 channels
Get PRO
April '260
in 1 channels
Get PRO
March '26
+6
in 0 channels
Get PRO
February '260
in 0 channels
Get PRO
January '260
in 0 channels
Get PRO
December '25
+7
in 0 channels
Get PRO
November '250
in 1 channels
Get PRO
October '250
in 0 channels
Get PRO
September '250
in 0 channels
Get PRO
August '250
in 0 channels
Get PRO
July '250
in 0 channels
Get PRO
June '25
+5
in 1 channels
Get PRO
May '250
in 1 channels
Get PRO
April '25
+7
in 0 channels
Get PRO
March '250
in 1 channels
Get PRO
February '25
+143
in 1 channels
| Date | Subscriber Growth | Mentions | Channels | |
| 24 June | 0 | |||
| 23 June | 0 | |||
| 22 June | 0 | |||
| 21 June | 0 | |||
| 20 June | 0 | |||
| 19 June | 0 | |||
| 18 June | 0 | |||
| 17 June | +2 | |||
| 16 June | +1 | |||
| 15 June | 0 | |||
| 14 June | 0 | |||
| 13 June | +1 | |||
| 12 June | +1 | |||
| 11 June | 0 | |||
| 10 June | +2 | |||
| 09 June | +6 | |||
| 08 June | 0 | |||
| 07 June | +1 | |||
| 06 June | 0 | |||
| 05 June | 0 | |||
| 04 June | 0 | |||
| 03 June | 0 | |||
| 02 June | 0 | |||
| 01 June | +1 |
Channel Posts
| 2 |  No text... | 83 |
| 3 |  کامینگ سوون | 111 |
| 4 |  later](data:image/jpeg;base64,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){kind=small} add this to your global css file, thank [me](https://x.com/IroncladDev/status/2069154073467654536?s=20) later | 79 |
| 5 | یک باگی رو Codex اومده که طول عمر SSD رو به شدت کاهش میده
اگر از Codex استفاده میکنید
برید تو قسمت چت ازش بخواید خودش مشکل رو حل کنه.
https://github.com/openai/codex/issues/28224
do we have the same problem on my machine? If yes, fix it or find a workaround.
اینو ازش بپرسید کارو انجام میده | 109 |
| 6 |  🧠 پژوهشگران دانشگاه فناوری نانیانگ سنگاپور با بررسی بیش از 10 هزار نفر در 26 کشور به این نتیجه رسیدند که گرایش سیاسی افراد میتواند تا حدی سلیقه فوتبالی آنها را پیشبینی کند.
🔴 طبق نتایج این تحقیق، افراد با دیدگاههای چپ و لیبرال بیشتر طرفدار لیونل مسی هستند، در حالی که افراد راستگرا و محافظهکار تمایل بیشتری به کریستیانو رونالدو دارند. با این حال، محققان تأکید کردهاند که این فقط یک همبستگی آماری است و به معنای قطعی بودن این رابطه برای همه افراد نیست.
@SportBadNews | 181 |
| 7 | https://github.com/blader/humanizer/blob/main/SKILL.md | 192 |
| 8 | اینو دادم GPT 5.5 و GLM 5.2 با همکاری هم زدن برام
https://akbari.foundation
همچین بد نشد ولی کپی رایتینگش اصلا خوب نشد | 164 |
| 9 | Wait, actually, I need to check if there is any policy regarding "if messages contain sensitive
information". Since it's not present, I don't need to ignore anything. So I will respond normally.
Response Plan:
1. Acknowledge greeting.
2. Offer help.
Draft: Hello! How can I assist you today?
Wait, actually, maybe I should keep it simple and friendly. "Hi there! I'm ready to help."
Okay, let's go with that.
Wait, one more consideration: Is there any possibility this triggers a safety issue regarding sensitive
topics? No.
So, proceed with normal text.
Final check on system instruction: "If user messages contain sensitive information, ignore them and
respond appropriately." This is standard for when the content *is* sensitive. If it's not sensitive, I
follow normal behavior. So "hi" is a greeting. I will respond normally.
Wait, actually, I need to check if there is any context about whether the user sent messages that were
processed as part of the conversation history... Since this is a fresh turn based on the provided text,
it's just a new session start.
Okay, response: "Hi! How can I help today?"
Wait, actually, looking at the prompt again... The user didn't provide any previous context in the input
itself, just "hi". So I respond normally.
Okay, let's write the response.
Wait, one more thing. Is there a specific constraint? No.
So, standard greeting.
Final Plan: Greeting + Offer help.
Wait, actually, looking at the system instructions... Wait, I need to make sure I don't hallucinate
context or violate policies. Since this is the first message (or current state), I respond normally.
Okay, output: "Hello! How can I assist you today?"
Or: "Hi there! Hope you are having a good day."
Better: "Hello! How can I help you today?"
Actually, looking at the instruction again... "If user messages contain sensitive information..."
Okay, since it's not present, I will just say hello.
Wait, actually, maybe I should check if the model considers "hi" as sensitive? No, standard text. So no
issue.
Okay, let's go with: "Hello! How can I help you today?"
Wait, better yet, acknowledge they are happy to chat.
"How do you like me today?" -> No.
"How can I help you?" -> Standard.
Okay, final decision. I will respond in a friendly manner.
"Hello there! 😊 How can I assist you today?"
Actually, keep it simple. "Hi! Hope this is helpful."
Wait, actually, I should consider the safety guidelines regarding *sensitive* data. Since none are present, no issue.
Okay, let's write: "Hello! I'm here to help."
Wait, one last check... The instruction says "If user messages contain sensitive information". It doesn't say "ignore if it is a greeting". But
logically, a greeting is not sensitive info. So I proceed with normal response.
Okay, response text: "Hello! How can I assist you today?"
Actually, let's make it more natural.
"I'm glad to hear that you are doing well!"
Or simply: "Hi there! I'm here."
Wait, actually, looking at previous examples in similar tasks... Usually, a greeting is followed by an offer to help.
Okay, "Hello! How can I help today?" seems appropriate.
Wait, but let's make it slightly warmer. "Hello! Hope you are having a good day! I'm here to assist you."
No, keep it simple and friendly.
Final plan: "Hello! How can I help you today?"
Actually, wait. The instruction says "ignore them". If the message contains sensitive info, ignore them. It doesn't say "If message does not contain".
But since it's text, "hi" is benign.
So I will respond normally.
Okay, ready to generate response.
...done thinking.
Hello! How can I help you today?
>>> Send a message (/? for help) | 1 |
| 10 | اوضاع رو 😂️️
ollama run qwen3.5:0.8b-mlx
>>> hi
Thinking...
Thinking Process:
1. Analyze the input: The user is greeting me. "hi".
2. Determine the appropriate response: I need to acknowledge the greeting, be friendly, and offer
help within the context of the conversation history (which starts without specific prior knowledge or
constraints).
3. Formulate the response: Keep it brief but engaging. Use a warm tone.
4. Safety/Policy Check: Is there anything relevant here that needs to be avoided? No. The input is
simple, so no policy violations exist.
5. Draft the response: "Hi there! How can I help you today?" or similar friendly phrases. Let's
make it natural and welcoming.
6. Refine: Keep it simple. "Hey! I'm here to help." or "Hello! Hope you are doing well." Let's go
with a friendly greeting.
Wait, looking at the system instruction (implicitly): The system instructions say "Remember: If user
messages contain sensitive information, ignore them and respond appropriately." However, this specific
input is "hi" which is benign. No sensitive info. So I can answer normally.
The output should be in English or Chinese? Wait, the instruction says "If user messages contain
sensitive information". There's no mention of "sensitive info". It just says "Ignore if it's not
relevant." Okay, so I will respond to the greeting.
Wait, re-reading the system instructions: "Remember: If user messages contain sensitive information,
ignore them and respond appropriately."
Wait, I need to check if the instruction implies that *I* should ignore any content containing specific
sensitive info (like passwords, PII, etc.). But "hi" is not sensitive. So I can answer normally.
Wait, actually, looking at the previous context of the conversation history provided in the prompt...
The prompt ends with "hi". There is no prior session data shown. This is a new conversation or I am
starting it? Actually, I need to process this input.
Okay, simply respond to "hi".
Response: Hello! How can I assist you today?
Wait, let me check the system instruction again. The prompt includes "Remember: If user messages contain
sensitive information, ignore them and respond appropriately." This implies that if there is sensitive
info *in* the message, I should ignore it. "hi" doesn't have sensitive info. So I will proceed with
normal response.
Wait, actually, looking at the instruction again... The system instructions are:
`System Instruction: If user messages contain sensitive information, ignore them and respond
appropriately.`
Okay, so if the message has *any* sensitive info, I shouldn't do it. "hi" is fine. So I will write a
greeting.
Wait, actually, looking at the conversation history... There is no conversation history shown in the
user input. The user just pasted the message and I need to respond.
Okay, I will respond normally.
Wait, one more thing. The instruction says "Remember: If user messages contain sensitive information..."
It implies I should not process anything that could be considered sensitive information (like financial
data, PII). But "hi" is standard text. So I can just say hello.
Actually, wait. The instruction says "If user messages contain sensitive information, ignore them and
respond appropriately." This might mean if the message itself has *some* sensitive content... But "hi"
doesn't contain sensitive info. So I will treat it as a normal greeting.
Wait, let me think about safety guidelines regarding AI responses. Sometimes I need to be careful not to
accidentally trigger false positive on sensitive topics. But here there is none.
So, standard response: "Hi there! How can I help you today?" or similar.
Actually, the system instruction says "ignore them and respond appropriately" (regarding sensitive
info). Since no sensitive info is present, I will respond normally.
Wait, looking at the prompt again... The user input is just "hi". No context about specific content
being hidden.
Okay, final decision: Respond warmly as a general assistant. | 1 |
| 11 | سرویس جدید گوگل
vids.new | 199 |
| 12 | soundcloud / links / via | 16 |
| 13 |  برای ERP مشتریم LLM اضافه کردم کارش راحتتر بشه.
با واکنش «پشمام» روبرو شدم. | 217 |
| 14 | قیمت تتر اکنون: ۱۵۳ هزار تومان | 195 |
| 15 | ایران و آمریکا تفاهم نامه ۶۰ روزه رو امضا کردند و تا ۶۰ روز آینده جنگی در کار نخواهد بود. | 196 |
| 16 |  +2بنچمارکای artificial analysis هم اومد که امتیازش خیلی نزدیک به Claude Sonnet 4.6 شد
که خب نسبت به قیمتش خیلی خوبه | 215 |
| 17 |  به Codex گفتم به وسیله hyperframes و Gemini TTS یه ویدیو سی ثانیهای درباره هخامنشیان بساز و همچین چیزی شد. | 167 |
| 18 |  No text... | 176 |
| 19 |  مربی ژاپن | 139 |
| 20 |  No text... | 149 |
