This model feels like a whole new level, my God. It is literally as if another worker is working with you for a project.

Hasab AI Launches Amharic-First AI Platform in Ethiopia

Hasab AI has launched an Amharic-first AI platform designed for content creators, call centers, educators, and businesses in Ethiopia and beyond. The platform offers tools like automatic Amharic and multilingual subtitle generation, AI-powered call tagging and summarization, and high-quality Amharic-English translation, all accessible via APIs. The platform emphasizes local-language inclusion.

ሀሳብ ኤአይ (Hasab AI) በኢትዮጵያ እና ከዚያም በላይ ላሉ የይዘት ፈጣሪዎች፣ የጥሪ ማዕከላት፣ አስተማሪዎች፣ እና የንግድ ድርጅቶች ተብሎ የተዘጋጀ አማርኛን-የመጀመሪያ ያደረገ የሰው ሰራሽ ዕውቀት መድረክ አስጀምሯል። መድረኩ እንደ ራስ-ሰር የአማርኛ እና የባለብዙ ቋንቋ የትርጉም ጽሑፍ (ሰብታይትል) አመንጪ፣ በሰው ሰራሽ ዕውቀት (AI) የሚመራ የጥሪ መለያ እና ማጠቃለያ፣ እንዲሁም ከፍተኛ ጥራት ያለው የአማርኛ-እንግሊዝኛ ትርጉም ያሉ መሣሪያዎችን ያቀርባል። እነዚህ ሁሉ መሣሪያዎች በኤፒአይዎች (APIs) በኩል የሚገኙ ሲሆኑ፣ መድረኩ ለአገር ውስጥ ቋንቋ ተካታችነት ትልቅ ትኩረት ይሰጣል።

🔗 - https://www.linkedin.com/company/hasabai/

Gemini 2.5 Pro (I/O Edition) Update Shines: Topping LMArena with Advanced AI Coding Prowess

Google  DeepMind has unveiled Gemini 2.5 Pro (I/O edition), an upgraded AI model engineered for native multimodality, extended context, and sharply improved coding prowess—including the ability to generate interactive web apps in Canvas with minimal prompting. The model now tops LMArena’s WebDev Arena Leaderboard with a +147  ELO score compared to the first Gemini 2.5 Pro released model.

ጎግል ዲፕማይንድ የተሻሻለውን ሰው ሰራሽ የማሰብ ሞዴል የሆነውን ጀሚኒ 2.5 ፕሮ (አይ/ኦ ኤዲሽን) ይፋ አድርጓል። ይህ ሞዴል በተፈጥሮው በርካታ መረጃዎችን (multimodality) የማስተናገድ፣ ሰፋ ያለ አውድ የመረዳትና እጅግ የተሻሻለ የኮዲንግ ችሎታ እንዲኖረው ተደርጎ የተሰራ ሲሆን፣ ይህም በካንቫስ ላይ አነስተኛ ጥያቄዎችን (prompts) በመጠቀም በይነተገናኝ የድር መተግበሪያዎችን (interactive web apps) መፍጠር ያስችለዋል። ሞዴሉ አሁን በኤልኤምአሬና የድር ልማት (WebDev Arena Leaderboard) መሪ ሰሌዳ ላይ ከመጀመሪያው የጀሚኒ 2.5 ፕሮ ሞዴል ጋር ሲነጻጸር በ +147 የኢኤልኦ (ELO) ነጥብ ብልጫ በመያዝ አንደኛ ደረጃ ላይ ይገኛል።

https://x.com/demishassabis/status/1919779362980692364?t=SVIvZ6I0WHVleWjjM8klWA&s=19

@webthreeth

OpenAI Leads Code‑Editing Surge: 20× Accuracy Jump in Nine Months, Gemini Gains Ground

OpenAI has maintained a commanding lead in code‑cleaning accuracy through April 2025, with its flagship o3 (high) model reaching nearly 80 percent on Aider Polyglot’s multi‑language editing benchmark—just ahead Gemini 2.5 Pro, which is closing the gap at roughly three‑quarters accuracy. The overall progress is striking: in only nine months, performance has vaulted almost twenty‑fold, leaping from the single‑digit accuracy of the lightweight GPT‑4o mini (Aug 2024) to the results delivered by o3 (high).

ኦፕንኤአይ እስከ ሚያዝያ 2025 ድረስ በኮድ ማፅዳት ትክክለኛነት ላይ ከፍተኛ የበላይነት ይዞ የቀጠለ ሲሆን፣ ዋናው የo3 (ከፍተኛ) ሞዴሉ በአይደር ፖሊግሎት ሁለገብ ቋንቋ የአርትዖት መለኪያ ላይ ወደ 80 በመቶ የሚጠጋ ውጤት አስመዝግቧል — ይህም ጎግልን በትንሹ የቀደመ ሲሆን ጀሚኒ 2.5 ፕሮ ወደ ሶስት አራተኛ የሚጠጋ ትክክለኛነት አሳይቷል። በአጠቃላይ የታየው መሻሻል አስገራሚ ነው፡ በዘጠኝ ወራት ውስጥ ብቻ፣ አፈፃፀሙ ከቀላል ክብደት GPT-4o ሚኒ (ነሐሴ 2024) ነጠላ አሃዝ ትክክለኛነት ወደ o3 (ከፍተኛ) ባስመዘገበው ዘመናዊ ውጤት በመዝለል ወደ ሃያ እጥፍ ገደማ ከፍ ብሏል

Source - Epoch AI

@webthreeeth

NOViQ 1.0 Now Public: Ethiopia’s AI-Powered Learning Platform Goes Mobile and Web

NOViQ 1.0, developed by NOViQ_AI, is an AI-powered educational app and web platform designed to make learning more accessible across Ethiopia. NOViQ has made the APK and web versions are now publicly available, offering seamless cross-device synchronization across the application and Web versions so users can access classes and educational content anytime, anywhere.

ኖቪክ 1.0 (NOViQ 1.0)፣ በኖቪክ_ኤአይ (NOViQ_AI) የተሰራ፣ በመላው ኢትዮጵያ ትምህርትን የበለጠ ተደራሽ ለማድረግ የተነደፈ በሰው ሰራሽ የማሰብ ችሎታ (AI) የሚንቀሳቀስ የትምህርት መተግበሪያ እና የድር መድረክ ነው። ኖቪክ የኤፒኬ (APK) እና የድር ስሪቶቹን አሁን ለሕዝብ ይፋ አድርጓል፤ ይህም ተጠቃሚዎች ክፍሎችን እና ትምህርታዊ ይዘቶችን በማንኛውም ጊዜና ቦታ ማግኘት እንዲችሉ በመተግበሪያው እና በድር ስሪቶቹ መካከል እንከን የለሽ የሆነ የመሣሪያዎች መመሳሰልን (cross-device synchronization) ያቀርባል።

🔗 - https://noviq.et/
📱 - https://play.google.com/store/apps/details?id=com.noviq.noviq_font_end

@webthreeth

Ethiopia’s AI-Powered EFPApp Wins Best Police Application Ahead of World Police Summit

The Ethiopian Artificial Intelligence Institute (EAII) announced that its AI-powered crime reporting mobile application, EFPApp, developed for the Ethiopian Federal Police, has been selected as the “Best Police Application” in preparation for the upcoming World Police Summit at the Dubai World Trade Centre. The app, which is currently operational in Addis Ababa and Dire Dawa, enables citizens to report crimes easily and enhances public safety through an intuitive, technology-driven platform.

የኢትዮጵያ ሰው ሰራሽ ዕውቀት ኢንስቲትዩት (ኢሰዕኢ)፣ ለኢትዮጵያ ፌዴራል ፖሊስ ያዘጋጀውና በሰው ሰራሽ ዕውቀት የሚሰራ የወንጀል ሪፖርት ማቅረቢያ የሞባይል መተግበርያ የሆነው 'EFPApp'፣ በዱባይ የዓለም ንግድ ማዕከል ሊካሄድ ላለው ዓለም አቀፍ የፖሊስ ጉባዔ ዝግጅት ላይ "ምርጥ የፖሊስ መተግበርያ" ተብሎ መመረጡን አስታውቋል። በአሁኑ ወቅት በአዲስ አበባ እና በድሬዳዋ አገልግሎት እየሰጠ የሚገኘው ይህ መተግበርያ፣ ዜጎች በቀላሉ ወንጀሎችን ሪፖርት እንዲያደርጉ የሚያስችል ሲሆን፣ በቀላልና በቴክኖሎጂ የታገዘ መድረክ አማካኝነት የህዝብን ደህንነት ያሻሽላል።

@webthreeth

Gemini 2.0 Image Generation Model Updated Today

Gemini is rolling out an updated 2.0 Image Generation model, featuring enhanced visual quality, more accurate text rendering, lower block rates, and higher rate limits for a seamless creative process. Priced at $0.039 per image through APIs, this model offers a cost-effective way to produce high-quality visuals.

-----------

ጄሚኒ የተሻሻለውን 2.0 የምስል ማመንጫ ሞዴሉን በይፋ ሥራ ላይ አውሏል። ይህ አዲስ ሞዴል የሚከተሉትን የላቁ ባህሪያትን ያካተተ ነው

* የተሻሻለ የምስል ጥራት፦ እጅግ የጠሩ እና ጥራት ያላቸው ምስሎችን ያቀርባል።
* የበለጠ ትክክለኛ የጽሑፍ አቀራረብ፦ በምስል ውስጥ የሚካተቱ ጽሑፎችን በትክክል የማስቀመጥ ችሎታው ከፍ ብሏል።
* ዝቅተኛ የማገጃ መጠን፦ ምስሎችን በማመንጨት ሂደት ውስጥ የሚከሰቱ እገዳዎች ቀንሰዋል።
* ከፍተኛ የፍጥነት ገደብ፦ በአጭር ጊዜ ውስጥ በርካታ ምስሎችን የማመንጨት አቅሙ ጨምሯል።

ይህ የተሻሻለ ሞዴል በኤፒአይ (API) በኩል በአንድ ምስል $0.039 ዋጋ የሚቀርብ ሲሆን፣ ከፍተኛ ጥራት ያላቸውን ምስሎች በአነስተኛ ወጪ ለማምረት ያስችላል።

@webthreeth

Figma Unveils AI-Powered Figma Styles to Revolutionize Design and CMS.

Figma Sites is an innovative, AI-powered website builder introduced by Figma at Config 2025, enabling designers to transform their Figma designs into fully responsive, functional websites with ease. It offers templates, design system integration, and AI features like text-prompted animations. Supporting responsive layouts and live HTML previews, it’s in open beta for paid Full seat users, with a CMS planned for later 2025, competing with platforms like WordPress and Wix.

ፊግማ ሳይትስ በፊግማ በኮንፊግ 2025 የተዋወቀ አዲስ የፈጠራ ችሎታ ያለው በሰው ሰራሽ የማሰብ ችሎታ (AI) የሚሰራ ድረ-ገጽ መገንቢያ ነው። ይህ መገንቢያ ዲዛይነሮች የፊግማ ዲዛይኖቻቸውን በቀላሉ ሙሉ በሙሉ ምላሽ ሰጪ እና ተግባራዊ ወደሆኑ ድረ-ገጾች እንዲቀይሩ ያስችላቸዋል። የተለያዩ የድረ-ገጽ አብነቶችን (templates)፣ ከዲዛይን ሥርዓቶች (design system) ጋር የመዋሃድ ችሎታን እና እንደ የጽሑፍ ጥያቄን መሠረት ያደረጉ የእንቅስቃሴ ውጤቶችን (animations) የመፍጠር አይነት የሰው ሰራሽ የማሰብ ችሎታ (AI) ባህሪያትን ያቀርባል።

@webthreeth

iCog Labs Opens Applications for 2025 AI Internship Program

iCog Labs is accepting applications for its 2025 Batch-1 AI Internship Program, open to students, graduates, and self-taught individuals passionate about AI and emerging technologies. The program requires a 10-hour weekly in-person commitment and is primarily for Ethiopians. The deadline to apply is July 31.

አይኮግ ላብስ ለ2025 የመጀመሪያ ዙር (ባች-1) የሰው ሰራሽ የማሰብ ችሎታ (AI) የሥራ ልምምድ መርሃ ግብር ማመልከቻዎችን እየተቀበለ ነው። ይህ ዕድል ለተማሪዎች፣ ለተመራቂዎች እና ስለ ሰው ሰራሽ የማሰብ ችሎታና አዳዲስ ቴክኖሎጂዎች ከፍተኛ ፍላጎት ላላቸው በራሳቸው ለተማሩ ግለሰቦች ክፍት ነው። ይህ ተመራጭ መርሃ ግብር እንደ AI፣ AGI (አጠቃላይ ሰው ሰራሽ የማሰብ ችሎታ)፣ የማሽን መማር (machine learning)፣ የኮምፒውተር የቋንቋ ጥናት (computational linguistics)፣ ሮቦቲክስ፣ ብሎክቼይን (blockchain)፣ ዴቭኦፕስ (DevOps) እና ባዮኢንፎርማቲክስ (bioinformatics) ባሉ መስኮች የተግባር ልምድ ይሰጣል። ፕሮግራሙ በሳምንት ለ10 ሰዓታት በአካል መገኘትን የሚጠይቅ ሲሆን በዋናነት ለኢትዮጵያውያን የተዘጋጀ ነው። የማመልከቻ ማቅረቢያ የመጨረሻ ቀን ሐምሌ 31 ነው።

Link for Application - https://qr-code.click/i/680fd719d8374

@webthreeth

AI Fuels Surge in Solo-Founder Startups, Doubling Since 2019

AI is empowering solo entrepreneurs to automate essential tasks such as marketing and customer service, diminishing the need for large founding teams and challenging traditional venture capital expectations of what a "complete" team looks like.

This shift is reflected in a growing trend toward independent, bootstrapped startups, with 38% of Carta startups in 2024 having a single founder and no VC backing—up significantly from 20.4% in 2019. The rise of AI tools is fueling a new era of lean, founder-driven ventures.

በ2024፣ 38% የካርታ የጀማሪ ኩባንያዎች አንድ መሥራች የነበራቸው ሲሆን ምንም የቬንቸር ካፒታል (VC) ድጋፍ አልነበራቸውም፤ ይህም በ2019 ከነበረው 20.4% በእጥፍ ገደማ ከፍ ብሏል። ይህ ከፍተኛ ጭማሪ ሰው ሰራሽ የማሰብ (AI) ችሎታ ነጠላ ሥራ ፈጣሪዎች ራሳቸውን ችለው የጀማሪ ኩባንያዎችን እንዲገነቡና እንዲያሳድጉ እያስቻላቸው መሆኑን ያሳያል።

@webthreeth

ChatGPT Launches GitHub Connector for Real-Time Code Analysis

OpenAI has introduced its first beta connector for deep research within ChatGPT—GitHub. This new feature allows users to link their repositories, ask questions, and receive AI-generated reports based on real code and engineering documentation. ChatGPT’s new GitHub connector enables Team, Plus, and Pro users to pull live code, READMEs, and documentation from GitHub repos for real-time analysis.

@webthreeth

The Deep Learning Era: How Explosive Growth in FLOPs Is Shaping the Future of AI

The Deep Learning Era (Since 2010) is marked by a rapid rise in AI training compute, measured in FLOPs, which quantify the number of calculations needed to train a model. In the past five years, training compute has grown at a rate of 4.6× per year—significantly faster than the 1.5× annual growth seen before 2010. If this trend continues, AI models by 2035 could require over 100,000 times more compute, potentially unlocking unprecedented capabilities.

የጥልቅ ትምህርት ዘመን (ከ2002 አመተምህረት）በሰው ሰራሽ የማሰብ (AI) ስልጠና የኮምፒውተር አቅም ፈጣን እድገት የሚታወቅ ሲሆን ይህም ሞዴልን ለማሰልጠን የሚያስፈልጉትን የስሌቶች ብዛት በሚለካው በFLOPs ይገለጻል። ባለፉት አምስት ዓመታት የስልጠና የኮምፒውተር አቅም በዓመት 4.6 እጥፍ አድጓል—ይህም ከ2010 በፊት ከታየው የ1.5 እጥፍ ዓመታዊ እድገት በእጅጉ ፈጣን ነው። ይህ አዝማሚያ ከቀጠለ በ2035 የሰው ሰራሽ የማሰብ ሞዴሎች ከ100,000 እጥፍ በላይ የኮምፒውተር አቅም ሊያስፈልጋቸው ይችላል፤ ይህም ከዚህ በፊት ታይቶ የማይታወቅ አቅሞችን ሊከፍት ይችላል።

Source - EPOCH AI

@webthreeth

Google I/O 2025 Set to Highlight Imagen 3.5 and Veo 3 Advances

At the upcoming Google I/O 2025, Google is expected to unveil Imagen 3.5 and Veo 3—two major upgrades in its generative AI lineup. Imagen 3.5 is anticipated to bring significantly faster image generation, while Veo 3 aims to push the boundaries of AI-driven video creation. Building on the success of Veo 2, Veo 3 is poised to strengthen Google’s leadership in video quality, as previously demonstrated in DeepMind’s benchmark evaluations like MovieGenBench.

በሚቀጥለው የጉግል አይ/ኦ 2025 (Google I/O 2025) ጉግል በፈጠራ የሰው ሰራሽ የማሰብ (generative AI) መስመሩ ላይ ሁለት ዋና ዋና ማሻሻያዎችን ይፋ ያደርጋል ተብሎ ይጠበቃል፤ እነሱም ኢማጀን 3.5 (Imagen 3.5) እና ቪዮ 3 (Veo 3) ናቸው። ኢማጀን 3.5 በከፍተኛ ሁኔታ ፈጣን የምስል አመራረት እንደሚያመጣ የሚጠበቅ ሲሆን፣ ቪዮ 3 ደግሞ በሰው ሰራሽ የማሰብ ችሎታ የሚመራ የቪዲዮ ፈጠራን ወሰን ለማስፋት ያለመ ነው። በቪዮ 2 (Veo 2) ስኬት ላይ በመገንባት፣ ቪዮ 3 ቀደም ሲል በዲፕማይንድ (DeepMind) እንደ ሙቪጄንቤንች (MovieGenBench) ባሉ የጥራት መመዘኛ ግምገማዎች እንደታየው የጉግልን የቪዲዮ ጥራት መሪነት የበለጠ ያጠናክራል።

@webthreeth

Fellou Launches: The World’s First Agentic Browser

Fellou, the world’s first agentic browser, is redefining how users interact with the web by shifting from passive browsing to dynamic task execution through its Deep Action technology.

An agentic browser is an AI-powered web interface that actively performs complex tasks on behalf of the user, transforming traditional browsing into goal-driven, autonomous interaction.

ፌሎው የተሰኘው እና በዓይነቱ የመጀመሪያ የሆነው “ኤጀንቲክ” አሳሽ (Agentic Browser) ተጠቃሚዎች ከድረ-ገጾች ጋር ያላቸውን ግንኙነት በአዲስ መልክ እየቀየረ ነው። ይህንንም የሚያደርገው “ዲፕ አክሽን” (Deep Action) በተሰኘው ቴክኖሎጂው አማካኝነት ሲሆን፣ ተጠቃሚዎች ድረ-ገጾችን እንዲሁ ከማሰስ ባለፈ ውስብስብ ተግባራትን በቀጥታ እንዲያስፈጽሙ ያስችላቸዋል።

Agentic Browser) ማለት በተጠቃሚው ስም ውስብስብ ተግባራትን በንቃት የሚያከናውን፣ በሰው ሰራሽ አስተውሎት (AI) የሚንቀሳቀስ የድረ-ገጽ በይነገጽ ነው፡፡ ይህ ቴክኖሎጂ ባህላዊውን የአሳሽ አጠቃቀም ወደ ግብ-ተኮር እና ራሱን የቻለ የድረ-ገጽ ግንኙነት ይለውጣል።

@webthreeth

LUCY, which is designed by Ethiopian-based AI builder Zemenu, is almost here—your all-in-one AI assistant built to do it all, from telling jokes and generating code to supporting business decisions and answering everyday questions . As Zemenu get closer to launch, your feedback is crucial about it’s different variables including features, performance, or user experience. You can share your feedback via the Telegram group @webthreeth, message @michaeltomasg, or send a LinkedIn DM to the channel.

PS - This is a very good AI, and would love your feedbacks. Please drop them in the comment box

https://labs.zemenu.org

Exams are just around the corner for many of you—whether you're in high school, pursuing your bachelor's, or working on a master's degree. I’d love to share how I use AI tools (most of them free!) to learn faster and to do research (without plagiarism 🤪, of course), and make studying more effective.

I'll also show you how to structure your study sessions using mind maps and stay organized without needing tutors or classes.

Should I go ahead and create this?
Hit like if you think it would help you! Join the channel if you want a guidance

Tool #1 - NotebookLM

Notebook LM is one of the key tools I use for both academic and professional development. Powered by the Gemini 2.0 model, it serves as an advanced AI-driven research and learning assistant. Here’s how it helps:

👨‍🏫Audio Overview: Converts discussions or podcasts into structured learning content.

👨‍🏫Mind Map: Visually summarizes complex ideas using mind maps.

👨‍🏫Study Guide: Generates clear and concise study outlines to help learners grasp topics quickly.

👨‍🏫AI Chatbot: Provides a conversational assistant to answer questions and support learning.

👨‍🏫Briefing Doc & FAQ: Produces summarized documents and frequently asked questions in Word format for quick reference.


To start learning something new, simply create a new page on Notebook LM, upload your source materials, and use the available commands to interactively explore and master your learning objectives.

@webthreeth

Crossing the Intelligence Threshold: Why AI Still Falls Short of True Innovation

AI has shown significant progress in IQ test scores, rising from basic to near-human levels, yet it still often fails to produce truly original or high-impact results. This gap suggests that while current models can replicate knowledge, they haven’t yet crossed the critical intelligence threshold needed for genuine innovation.

According to Lotka’s Law—which shows that major scientific contributions come disproportionately from a small number of highly intelligent individuals—AI may need to advance by another 30 IQ points to shift from mimicking existing work to generating novel, groundbreaking ideas.

@webthreeth

Admin Opinion - This makes so much sense for me. That's why many people don't feel and understand the growth in AI unless you have been part of the journey since the beginning or have a field specialization in an area.

Source - Dan Hendrycks (US Center for AI Safety - Director)