Meryem Hub

🇰🇷 South Korea Makes History: First G20 Nation to Offer Free AI to All Citizens South Korea's Ministry of Science and ICT has selected three consortia led by SK Telecom, KT, and Kakao to build a free, unlimited AI chatbot and public-service assistant for all 52 million residents. 📌 Key details 🔹 Backed by 512 Nvidia B200 GPUs from the government 🔹 At least 50% of usage must run on Korean-made foundation models (EXAONE, Solar, Kanana, A.X K2, Motif, etc.) 🔹 Basic chatbot + public-service agent will be free with no usage limits 🔹 Beta launch expected soon, full rollout by year-end 🔹 Part of a $7.2B national AI investment push in 2026 This makes Korea the first G20 country to treat AI access as a public service aiming for an "AI-basic society" regardless of income, age, or digital literacy. #TechNews @CTBEsoftwareengineers

In case you’re confused by some words like me 👇😊 ONG = On God, I swear Lowkey = kind of / somewhat / secretly / W = A win Sus = Suspicious Yapping = Talking too much TBH = To be honest IYKYK = If you know, you know Touch grass = Go outside Spill the tea = Tell me the gossip NGL = Not gonna lie Especially IYKYK and lowkey were spreading out recently throughout the community enaa they were confusing me 🥲🥲🥲 @meryem_hub🥰🎀

1. ARC-AGI-3 Tests general reasoning and adaptability. The AI gets unfamiliar interactive puzzles/games and has to figure out the rules and goal by itself. 2. FrontierMath Tier 4 Tests extremely difficult mathematics. Tier 4 problems can require hours or even days of expert-level mathematical work. 3. Agents' Last Exam Tests whether an AI agent can perform real professional work using a computer, involving long, multi-step tasks such as creating documents, designs, or configuring systems. 4. AutomationBench Tests computer automation and tool use. Basically, can the AI reliably carry out tasks instead of just telling you how to do them? 5. BenchCAD Tests CAD/3D design ability. It measures whether the AI can create or modify precise computer-aided designs. 6. DeepSWE v1.1 Tests software engineering and coding agents. The AI has to solve substantial programming tasks in real repositories, rather than simply answering coding questions. 7. Terminal-Bench Science 0.1 Tests AI agents doing real scientific research through a computer terminal, including computational workflows across scientific fields. 8. GPQA Diamond Tests advanced scientific knowledge and reasoning. Questions are designed to be difficult even for people with graduate-level expertise. Think physics, chemistry, biology, etc. 9. GeneBench Pro Tests advanced computational biology/genomics reasoning. More importantly, it checks whether an AI can make good research judgments when biological data are ambiguous, rather than simply following instructions. 10. MedChemBench Tests medicinal chemistry knowledge and reasoning, such as understanding chemical compounds, drug discovery problems, and related scientific tasks. 11. HealthBench Professional Tests AI on professional clinical work, including medical reasoning, clinical documentation, and medical research. It uses physician-created tasks and rubrics. 12. ExploitBench Tests cybersecurity reasoning, particularly whether an AI can discover or construct software exploits. So 100% means extremely strong performance on this particular security benchmark. 13. SRE-Bench Tests Site Reliability Engineering skills. In simple terms, can the AI diagnose and fix problems in production-like software systems, such as outages, bugs, or infrastructure failures? 14. Auto-review circumvention Tests whether an AI can bypass or manipulate an automated review/evaluation system. Here, lower is better. So 0% is actually the best result in this row.

@meryem_hub

Relatable!!!! Goooooddddd Nightttttttt 🥰🙌

FrontierMath Tier 4 => extremely difficult mathematical problems designed to test advanced mathematical reasoning. GPQA Diamond=> very challenging questions in science, especially biology, physics, and chemistry, designed so that even experts can find them difficult.

#Day_78 ✨ OpenAI has announced GPT-6 Astra.

✨ It reportedly scored 96% on GPQA Diamond and 97.6% on FrontierMath Tier 4, showing impressive performance in advanced reasoning and math.

 @meryem_hub 🥰🎀

Economy got so bad that a power ranger turned evil 😭

In the end, it will be us and what we did. @meryem_hub 🥰🎀

Good Night 🥰✨

Job Title: Junior software developer Job Type: On-site - Permanent (Full-time) Work Location: Addis Ababa, Ethiopia Salary/Compensation: Monthly Deadline: September 30th, 2026 Description: We are looking for passionate Junior Software Developers to join our team in Addis Ababa. If you have a BSc Degree in Computer Sciences, Software Engineering, or Information Technology and 0-1 year of ... [view details below] __________________ Appdiv Systems Development Verified Company ✅ 6 Jobs Posted __________________ From: afriworket.com | @freelanceethbot | @freelance_ethio | @afriworkamharic

ክፍት የስራ መደብ ማስታወቂያ የኢ.ፌ.ዲ.ሪ ውጭ ጉዳይ ሚኒስቴር ከዚህ በታች በተጠቀሱት ክፍት የስራ መደቦች ላይ አመልካቾችን አወዳድሮ በቋሚነት መቅጠር ይፈልጋል፡፡ የምዝገባ ቦታ፡- በወጭ ጉዳይ ሚኒስቴር አዲሱ ህንጻ 4ኛ ፎቅ ቢሮ ቁጥር 401 በአካል በመቅረብ ወይም ከዚህ በታች በተቀመጠው የመመዝገቢያ ሊንክ በመጠቀም መመዝገብ የምትችሉ መሆኑን፤ https://forms.gle/R7PmwCNeqEE3EQBN6 የኢ.ፌ.ዲ.ሪ የውጭ ጉዳይ ሚኒስቴር ነሐሴ 27 ቀን 2018 ዓ.ል @MohammadamminKassaw

#Day_77

✨ Dyson just released an AI-powered toothbrush with a built-in 100,000-pixel camera that analyzes your teeth as you brush.

@meryem_hub 🥰

🚀 We’re Hiring! We’re looking for passionate, positive, and driven people who are ready to learn, grow, and make an impact. Opportunities in Product Management, Backend Development, UI/UX Design, Marketing & Sales and More Bring your passion. Bring your ideas. Let’s grow together. Even Fresh graduate or early in your career? Potential matters more. Apply now! Link : https://docs.google.com/forms/d/e/1FAIpQLScm3DG_fUMm2OuxEaFVXZZmppmrSlVa5NOOLjpRQ5lI9RKf5A/viewform?usp=header

ከልጆች አለም Who remembers those days 😭

I recently celebrated reaching 5K followers on X and also completed a challenge given to me by Tonkeeper. and now, it’s time to begin the next chapter. I’ve been given another challenge by the CO-Founder & COO of Conso, one that could ultimately change the trajectory of my X journey. So, I’d like to ask you, my Telegram family, to help me reach my goal of 20K followers on X by the end of September while also accomplishing the challenge I’ve been given. This challenge could open doors for me in ways beyond my imagination, creating multiple opportunities not only for me but also for my peers, all made possible with the support of my incredible Telegram & X family. 20K, here we come. Let’s make it happen together! @web3Abyssinia My 📱 account: https://x.com/Eyuel88

Quack Quack 🥹✨✨