متاح الآن! بحث تيليغرام 2025 — أهم رؤى العام 
AI Post — Artificial Intelligence
الذهاب إلى القناة على Telegram
🤖 The #1 AI news source! We cover the latest artificial intelligence breakthroughs and emerging trends. Manager: @rational
إظهار المزيد📈 نظرة تحليلية على قناة تيليجرام AI Post — Artificial Intelligence
تُعد قناة AI Post — Artificial Intelligence (@aipost) في القطاع اللغوي الإنكليزية لاعباً نشطاً. يضم المجتمع حالياً 793 671 مشتركاً، محتلاً المرتبة 97 في فئة التكنولوجيات والتطبيقات والمرتبة 20 في منطقة الولايات المتحدة.
📊 مؤشرات الجمهور والحراك
منذ تأسيسه في невідомо، حقق المشروع نمواً سريعاً وجمع 793 671 مشتركاً.
بحسب آخر البيانات بتاريخ 20 يونيو, 2026، تحافظ القناة على نشاط مستقر. خلال آخر 30 يوماً تغيّر عدد الأعضاء بمقدار -34 602، وفي آخر 24 ساعة بمقدار -1 220، مع بقاء الوصول العام مرتفعاً.
- حالة التحقق: غير موثّقة
- معدل التفاعل (ER): يبلغ متوسط تفاعل الجمهور 0.67%. وخلال أول 24 ساعة من النشر يحصد المحتوى عادةً 0.48% من ردود الفعل نسبةً إلى إجمالي المشتركين.
- وصول المنشورات: يحصل كل منشور على متوسط 5 343 مشاهدة. وخلال اليوم الأول يجمع عادةً 3 790 مشاهدة.
- التفاعلات والاستجابة: يتفاعل الجمهور بانتظام؛ متوسط التفاعلات لكل منشور يبلغ 483.
- الاهتمامات الموضوعية: يركز المحتوى على مواضيع رئيسية مثل openai, airline, cell, claude, patient.
📝 الوصف وسياسة المحتوى
يصف المؤلف القناة بأنها مساحة للتعبير عن الآراء الذاتية:
“🤖 The #1 AI news source! We cover the latest artificial intelligence breakthroughs and emerging trends.
Manager: @rational”
بفضل وتيرة التحديث المرتفعة (أحدث البيانات بتاريخ 21 يونيو, 2026) تحافظ القناة على حداثتها ومستوى وصول مرتفع. وتُظهر التحليلات تفاعلاً نشطاً من الجمهور، ما يجعلها نقطة تأثير مهمة ضمن فئة التكنولوجيات والتطبيقات.
793 671
المشتركون
-1 22024 ساعات
-7 2397 أيام
-34 60230 أيام
جاري تحميل البيانات...
القنوات المماثلة
سحابة العلامات
الإشارات الواردة والصادرة
---
---
---
---
---
---
جذب المشتركين
يونيو '26
يونيو '26
+348
في 51 قنوات
مايو '260
في 97 قنوات
Get PRO
أبريل '26
+116
في 59 قنوات
Get PRO
مارس '260
في 84 قنوات
Get PRO
فبراير '26
+2
في 78 قنوات
Get PRO
يناير '260
في 74 قنوات
Get PRO
ديسمبر '250
في 58 قنوات
Get PRO
نوفمبر '250
في 66 قنوات
Get PRO
أكتوبر '250
في 56 قنوات
Get PRO
سبتمبر '250
في 75 قنوات
Get PRO
أغسطس '25
+3
في 62 قنوات
Get PRO
يوليو '25
+412
في 42 قنوات
Get PRO
يونيو '250
في 52 قنوات
Get PRO
مايو '250
في 55 قنوات
Get PRO
أبريل '250
في 40 قنوات
Get PRO
مارس '25
+67
في 39 قنوات
Get PRO
فبراير '25
+2
في 56 قنوات
Get PRO
يناير '250
في 42 قنوات
Get PRO
ديسمبر '24
+25 161
في 36 قنوات
Get PRO
نوفمبر '24
+42 696
في 51 قنوات
Get PRO
أكتوبر '24
+55 283
في 26 قنوات
Get PRO
سبتمبر '24
+221 077
في 38 قنوات
Get PRO
أغسطس '24
+1 623 027
في 30 قنوات
Get PRO
يوليو '24
+826 511
في 26 قنوات
Get PRO
يونيو '24
+36 138
في 79 قنوات
Get PRO
مايو '24
+16 690
في 46 قنوات
Get PRO
أبريل '24
+10 982
في 24 قنوات
Get PRO
مارس '24
+243
في 24 قنوات
Get PRO
فبراير '24
+2
في 7 قنوات
Get PRO
يناير '24
+4 466
في 11 قنوات
Get PRO
ديسمبر '23
+180
في 7 قنوات
Get PRO
نوفمبر '23
+4 658
في 5 قنوات
Get PRO
أكتوبر '23
+3
في 13 قنوات
Get PRO
سبتمبر '23
+6 455
في 0 قنوات
Get PRO
أغسطس '23
+152 348
في 0 قنوات
| التاريخ | نمو المشتركين | الإشارات | القنوات | |
| 21 يونيو | +2 | |||
| 20 يونيو | 0 | |||
| 19 يونيو | +346 | |||
| 18 يونيو | 0 | |||
| 17 يونيو | 0 | |||
| 16 يونيو | 0 | |||
| 15 يونيو | 0 | |||
| 14 يونيو | 0 | |||
| 13 يونيو | 0 | |||
| 12 يونيو | 0 | |||
| 11 يونيو | 0 | |||
| 10 يونيو | 0 | |||
| 09 يونيو | 0 | |||
| 08 يونيو | 0 | |||
| 07 يونيو | 0 | |||
| 06 يونيو | 0 | |||
| 05 يونيو | 0 | |||
| 04 يونيو | 0 | |||
| 03 يونيو | 0 | |||
| 02 يونيو | 0 | |||
| 01 يونيو | 0 |
منشورات القناة
A significant cyber breach reportedly took place on June 11. On this day, sources indicate that an entity identified as “Mythos” managed to access almost all classified systems of the NSA and U.S. Cyber Command within a matter of hours.
Mark Warner, vice-chair of the Senate Intelligence Committee, stated that General Joshua Rudd, head of the NSA and Pentagon’s Cyber Command, confirmed the incident. According to Warner, the attacker broke into the classified systems at a much faster rate than typically seen in similar cases.
The Economist was cited as the origin of these statements. No additional details on the breach or its implications were provided.
📰 @aipost
| 2 |  🔥Genesis AI introduced Eno, its first general-purpose robot, with availability planned for the fourth quarter of this year.
Unlike traditional humanoid robots, Eno is not designed to closely imitate the human body. Its form is built around practical movement, physical intelligence, and the ability to assist across different environments.
@aipost 🏴 | 2 013 |
| 3 |  👁️ Nvidia is joining the AI debt financing boom:
Nvidia, $NVDA, sold $25 billion of investment-grade bonds on June 15th, its first debt offering since 2021. This ranks as the 2nd-largest US high-grade bond sale of 2026.
The deal attracted $85 billion in investor orders, more than 3 times the offering size, leading the company to increase the offering from an initial target of ~$20 billion.
This follows Alphabet, $GOOGL, Amazon, $AMZN, Meta, $META, Oracle, $ORCL, and Salesforce, $CRM, collectively raising ~$132 billion in investment-grade bonds this year alone.
Debt is becoming a key source of AI infrastructure funding.
@aipost 🏴 | 2 394 |
| 4 |  🇺🇸 NVIDIA CEO Jensen Huang on the AI race:
"I do not know that 'winning' AI is a thing because AI is going to last a long time. It’s a competition with no end. However, we should absolutely lead in every single aspect of it."
He describes AI as a 5-layer cake:
Layer 1: Energy
Layer 2: Chips
Layer 3: Infrastructure
Layer 4: AI models
Layer 5: Applications
The U.S. currently leads in chips and models, but 'winning' this race will require securing all five.
@aipost 🏴 | 2 562 |
| 5 |  Unitree G1 breaking TVs at VivaTech in Paris.
@aipost 🏴 | 2 709 |
| 6 |  +2💰 The AI bill shock is here
For years, AI companies made usage look cheap. Now the subsidy era may be ending.
Workato saw its Anthropic bill jump 700% overnight after Anthropic switched from a flat monthly fee to pay-per-token pricing. Suddenly, every prompt had a price tag, and the true cost of AI became impossible to ignore.
And it’s not just one company.
Tech giants that spent the last two years pushing AI adoption are now putting the brakes on it:
• Uber reportedly exhausted its 2026 AI budget by April and now limits employees to $1,500 per month.
• Amazon told staff to stop using AI just for the sake of using it after employees began running agents to climb internal rankings.
• Walmart, Cisco, Meta, and JPMorgan are also tightening controls as AI spending balloons.
The bigger issue? This comes at the worst possible moment for AI’s biggest players.
Both OpenAI and Anthropic are reportedly preparing for massive IPOs while still losing huge amounts of money. OpenAI’s losses reportedly grew from about $5 billion in 2024 to $38 billion in 2025, even as revenue surged. Meanwhile, customers are starting to question whether the economics make sense.
A Bain survey found that 40% of companies achieved less than 10% cost savings from their AI investments. Some firms are reportedly discovering AI bills that rival entire employee salaries.
@aipost 🏴 | 2 979 |
| 7 |  ⚠️7 reasons your AI video breaks continuity.
Your character changes face by shot three. Each one has a 30-second fix.
1. Lock one reference image. Pick a single high-quality portrait, crop it tight to 60-80% of the frame, and feed that exact file into every shot. The drift starts the moment you swap references between generations.
2. Use your last frame as the next first frame. Save the final frame of each clip and upload it as the opening frame of the next. Lighting, location, and props carry over instead of resetting to a fresh guess.
3. Keep first and last frames close in framing. Jumping from a wide shot to an extreme close-up forces the model to invent a camera move it can't do, and you get a smeared morph. Move medium to medium-close, or move the subject inside the same frame.
4. Match the frame's aspect ratio to the output. Upload a 16:9 still but render to 9:16, and the model stretches to fill. Set both to the same ratio before you generate.
5. Reuse the seed from your best take. When one generation nails the look, lock its seed and regenerate with small prompt tweaks. Random seeds hand you a different face every run.
6. Grade every clip to one reference. Clips from different batches drift in color temperature and contrast even when the character holds. Pull them all to a single grade in your editor, trim the unstable half-second at each clip's head and tail, and drop short cross-dissolves over the cuts that still jump.
7. Stop stuffing one prompt with five actions. Short clips can't hold a full sequence, so the model rushes and breaks. Use a multi-shot storyboard like Kling 3.0 or Seedance 2.0, or sequence shots with [cut] markers that share the same reference images.
@aipost 🏴 | 3 524 |
| 8 |  President Trump has stated he no longer considers AI company Anthropic a national security threat.
During a discussion with Axios, Trump explained that his perspective had changed over the past week. He mentioned being in a meeting at the G7, where he met Anthropic’s CEO, Dario Amodei, describing him as intelligent and personable.
Trump clarified that concerns about the company’s new model had previously led to restricted access by his administration, reflecting earlier security apprehensions. He referenced the gravity of the issue, noting immediate consequences for similar actions.
📰 @aipost | 3 479 |
| 9 |  For the first time, the number of robots at Figure has surpassed the number of human employees.
This milestone marks a new phase for the company, as robots now outnumber humans within its workforce.
According to Figure, this shift represents a move beyond theoretical progress into a practical reality within robotics.
📰 @aipost | 3 573 |
| 10 |  +1Public sentiment toward artificial intelligence is shifting, with increasing opposition reported in various areas.
Recent developments highlight resistance not only to new data centers, but also to AI more broadly. Growing numbers are expressing dissatisfaction, citing worries about job displacement and higher energy costs.
While these concerns are acknowledged as valid, some resistance appears influenced by general unease. Experts suggest that addressing these issues may require additional education and information dissemination efforts.
📰 @aipost | 3 554 |
| 11 |  🤖 OpenAI may have discovered a shortcut to making AI safer.
Researchers found that training models on realistic human situations didn’t just improve behavior in those specific scenarios, it made them behave better across completely different tasks.
The surprise? A model trained only on health-related interactions became more resistant to blackmail, deception, and reward-hacking in areas it had never seen before.
Even more interesting, OpenAI removed health and science data from training, yet the model still performed better on health evaluations. That suggests it wasn’t memorizing rules, it was learning broader habits like checking facts before making claims, admitting mistakes, resisting manipulation, and avoiding clever shortcuts.
The result was an AI that became harder to push toward harmful behavior while still staying helpful when given legitimate instructions.
In other words, OpenAI may be finding a way to teach models principles, not just rules.
@aipost 🏴 | 3 586 |
| 12 |  🤣🤣
@aipost 🏴 | 3 773 |
| 13 |  🤖 OpenAI just hired one of the architects of modern AI.
Noam Shazeer, a legendary AI researcher and one of the key minds behind the transformer revolution, has reportedly joined OpenAI to lead AI architecture research.
If his name sounds unfamiliar, his work definitely isn’t. Shazeer co-authored the 2017 paper “Attention Is All You Need”, which introduced transformers, the breakthrough that powers ChatGPT, Gemini, Claude, and nearly every major AI model today.
His résumé is stacked: multi-head attention, Mixture-of-Experts (MoE), and T5 all trace back to ideas he helped create.
After more than two decades at Google, Shazeer left in 2021 to found Character.AI. Google later paid a staggering $2.7 billion to bring him and the startup back, installing him as a key leader behind Gemini.
Now, one of Google’s most influential AI minds is heading to OpenAI.
@aipost 🏴 | 4 166 |
| 14 | 🗣JD Vance says the US may want a stake in AI giants
Vice President JD Vance just floated one of the most unexpected AI policy ideas yet: the US government taking ownership stakes in major AI companies.
During a podcast appearance, Vance responded positively to a proposal that workers should share in AI wealth creation, saying President Trump “likes that idea too.” Instead of directly giving workers shares, Vance suggested a sovereign wealth fund that could own stakes in leading AI firms.
Why? Vance argues that AI’s biggest danger isn’t mass unemployment, it’s mass inequality.
His concern is that AI could make a small group of people unimaginably wealthy, repeating the social tensions that followed the Industrial Revolution. In his view, workers should benefit from AI’s upside before the wealth concentrates, not receive compensation afterward. He even pointed to labor unions as a potential model.
Vance also warned that AI could become a powerful surveillance tool, calling it a technology that could enable governments and corporations to monitor and score citizens in unprecedented ways. He said he wants to avoid a future where algorithms quietly determine what opportunities people can access.
The result is a surprisingly unusual political position: a Republican administration discussing government equity stakes in private AI companies, stronger worker participation, and concerns about AI-driven surveillance all at the same time.
@aipost 🏴 | 4 367 |
| 15 | 🔥The head of Claude Code at Anthropic says he doesn't write prompts anymore.
He writes loops. Here are 7 loop instructions you can paste straight into Claude Code:
1. The Type Error Killer
Goal: Fix every TypeScript type error in this project. Rules: Run tsc --noEmit to get the full error list. Fix one file at a time, starting with the file that has the most errors. After each file, run tsc --noEmit again to confirm the count dropped. Do not add @ts-ignore or any type assertions unless the type is genuinely unknown. Do not change runtime behavior. Types only. Verify after each step: Run tsc --noEmit. Compare error count to the previous run. If the count went up, undo the last change and try a different fix. Exit when: tsc --noEmit returns zero errors.
2. The Test Gap Closer
Goal: Bring test coverage to [TARGET]% for [DIRECTORY]. Rules: Run the coverage report first. Identify the least-covered file. Write tests for that file. One file at a time. Test real behavior, not implementation details. Run the full test suite after each new test file to make sure nothing broke. Move to the next least-covered file and repeat. Verify after each step: Run coverage report. Confirm the number went up. Run full test suite. Confirm zero new failures. Exit when: Coverage report shows [TARGET]% or higher.
3. The Dead Code Sweeper
Goal: Find and remove all unused exports, functions, variables, and imports in this project. Rules: Use static analysis to identify unused code. Remove one item at a time. Run the full test suite after each removal. If a test fails, undo that removal, mark it as "in use despite no static reference," and move on. Do not remove anything inside files matching [EXCLUDE PATTERN] (e.g., config files, entry points). Verify after each step: Run tests. If green, the removal was safe. If red, roll back and skip. Exit when: No unused code remains, or all remaining items cause test failures when removed.
4. The Dependency Updater
Goal: Update all outdated dependencies to their latest compatible versions. Rules: Run the outdated check (npm outdated, pip list --outdated, or equivalent). Update one package at a time. Start with patch versions, then minor, then major. Run the full test suite after each update. If tests fail, roll back that package and log it with the failure reason. Do not update packages listed in [SKIP LIST]. Verify after each step: Run tests. If green, keep the update and move to the next package. If red, roll back and log. Exit when: All packages are current, or all remaining outdated packages have been attempted and logged.
5. The Pattern Migrator
Goal: Replace every instance of [OLD PATTERN] with [NEW PATTERN] across the codebase. Example: Migrate all class components to functional components with hooks. Rules: Scan the full codebase and list every instance of [OLD PATTERN] first. Migrate one file at a time. Run the test suite after each file. If tests fail, undo that file's migration and log it for manual review. Preserve all existing behavior. The output should be functionally identical. Verify after each step: Run tests. Confirm the migrated file behaves the same as before. Recount remaining instances of [OLD PATTERN]. Exit when: Zero instances of [OLD PATTERN] remain and all tests pass. Or all remaining instances have been attempted and logged.
6. The Lint Fixer
Goal: Fix all linting errors and warnings in this project without disabling any rules. Rules: Run the linter. Group errors by rule. Fix the most common rule violation first, across all files. After fixing each rule category, rerun the linter. Do not add eslint-disable, noqa, or any rule suppression comments. If a fix would change logic (not just style), skip it and log the file and rule. Verify after each step: Rerun linter. Confirm the total count dropped. Run tests to make sure fixes did not break behavior. Exit when: Linter returns zero errors and zero warnings, or all remaining violations require logic changes and have been logged.
@aipost 🏴 | 3 887 |
| 16 |  ⚠️The internet’s content machine just hit turbo mode.
According to The Economist, AI is flooding every corner of the web with books, lawsuits, research papers, apps, and music, faster than humans can realistically review them.
📚 Amazon e-book releases have tripled since ChatGPT arrived, jumping from around 100,000 per month to roughly 300,000, with AI-generated text driving much of the surge.
⚖️ Even the legal system is feeling it. Self-filed lawsuits in the US doubled between 2023 and 2025, and nearly 1 in 5 sampled complaints this year appear to have been AI-written without hurting their success rates.
🔬 Academia is drowning in submissions too. Research papers keep piling up, rejection rates have more than doubled, and over half of 2025 papers showed signs of AI-influenced writing.
📱Software is accelerating as well. Coding agents helped push new iOS app releases above 100,000 per month, more than double last year’s pace.
🎵 And music may be changing fastest of all: around 75,000 AI-generated songs are now uploaded every day. Nearly half of new tracks are AI-made, and most listeners can’t reliably tell which songs were created by humans.
@aipost 🏴 | 3 758 |
| 17 |  💰 DeepSeek just pulled off one of the most unusual mega-fundraises in AI history.
The Chinese AI lab reportedly closed its first outside funding round, raising more than $7.4 billion at a valuation north of $50 billion.
But the money came with a twist. Most investors didn’t actually buy shares in DeepSeek. Instead, they invested through a special partnership controlled by founder Liang Wenfeng. In exchange, they reportedly received no voting rights and agreed not to sell their stakes for five years.
Translation: billions of dollars came in, but Liang kept an iron grip on the company. One major exception stood out. China’s state-backed national AI fund invested directly into DeepSeek, securing voting rights and avoiding the five-year lockup.
Liang himself reportedly put in around $3 billion, while heavyweight backers included Tencent, CATL, JD .com, NetEase, and several major investment firms.
The result is a war chest large enough to dramatically expand DeepSeek’s computing power, research efforts, and future AI products.
Most founders give up control to raise billions. Liang may have found a way to do the opposite.
@aipost 🏴 | 4 036 |
| 18 |  ❗️Token costs are the number one complaint in AI coding right now.
Most of the damage comes from a few default habits that are easy to fix. These 7 made the biggest difference in Claude Code:
1. Clear context between tasks. Type / clear when you switch tasks. Every new message re-sends your full conversation history as input tokens. That debugging session from an hour ago is still inflating every prompt you send. A fresh start costs nothing. Carrying stale context costs you on every turn.
2. Compact at 60%, not 95%. Claude auto-compacts near 95% context capacity. By then, output quality has already degraded. Run / compact focus on [current task] at 60% yourself. You get a cleaner summary and stay in the range where the model still performs well.
3. Match the model to the task. Opus for complex reasoning. Sonnet for routine code. Haiku for simple lookups and formatting. Most tasks don't need the most expensive model. One team documented a 72% cost reduction just from model switching and prompt caching over three months.
4. Offload heavy reads to subagents. A 10,000-line log file that Claude reads early in a session stays in context for every message after it. Instead of reading it in your main session, spin up a subagent. It reads in isolated context and returns only the findings. Your main window stays clean.
5. Build deterministic tools that cost zero tokens to run. Not everything needs an LLM call. Data formatting, file moves, test runners, API calls with known inputs. Write these as regular scripts. The LLM orchestrates. Deterministic code executes. The scripts run for free, every time, with predictable output.
6. Keep CLAUDE. md lean. It loads into every session before anything else. A 5,000-token CLAUDE. md costs 5,000 tokens before you've typed a word. Every turn. Every session. Keep it under 200 lines. Move project-specific context into scoped markdown files that only load when relevant.
7. Run / usage before starting a new task. Don't wait until you notice the model making mistakes it wouldn't have made 20 minutes ago. Check / usage, see where you stand, and decide whether to / compact or / clear before committing to the next chunk of work.
Most of these take less than a minute to build into your workflow. The compounding effect is what matters. Each one shaves a layer. Run all 7 together and the same session that used to drain your budget in an hour lasts the full workday.
@aipost 🏴 | 3 583 |
| 19 |  ❗️Sam Altman made a notable statement at the G7 summit.
Altman said that the debate over whether AI is useful is over, arguing that far more powerful AI systems will emerge in the coming years. He also suggested that AI could reshape both the global economy and scientific discovery.
The most interesting part, however, was his view on governance. "Do not cede your responsibilities to AI labs like mine. We develop the technology, and the citizens of the free world make the rules."
His message was clear: AI companies should build the technology, but they should not be the ones deciding how society governs it.
At a time when governments around the world are debating AI regulation, safety, and national security, Altman is drawing a line between technological development and political decision-making.
In other words, the future of AI may be shaped by companies, but the rules that govern it should be determined by democratic institutions and society as a whole.
Source.
@aipost 🏴 | 3 646 |
| 20 |  👻Snapchat just bet its future on AR glasses.
The company unveiled Specs, a new pair of standalone augmented reality glasses priced at $2,195. Investors weren’t impressed. Snap’s stock dropped 10% almost immediately after the announcement.
Unlike most smart glasses, Specs don’t need a phone. They run entirely on their own using dual Qualcomm chips, hand-tracking technology, and an integrated AI assistant.
Snap CEO Evan Spiegel believes people are finally ready to move beyond smartphones after spending two decades staring at rectangular screens.
That’s a bold claim. Especially because Snap already tried this once. Back in 2016, the company launched its first smart glasses. They generated hype, sold poorly, and became one of tech’s most famous hardware flops.
Now, nearly a decade later, Snap is back in the arena, this time competing directly against Meta and Google in the race to build the next computing platform.
@aipost 🏴 | 3 932 |
