Web 3.0 Ethiopia - DeFi & AI

I will be back starting from tomorrow. Sorry, I was really busy with work.

https://www.tiktok.com/@economicsimplifiedet This is my new economics analysis page. Please watch the videos and give me feedback on it

Kimi AI Launches K3.1 Pro with 2.8 Trillion Parameters and Advanced 3D Capabilities Kimi.ai has announced the release of Kimi K3.1 Pro, a massive 2.8-trillion parameter Mixture-of-Experts model that activates 104 billion parameters during inference. Featuring a 1-million-token context window and native multimodal support, it is the company's most capable open-weight model to date. K3.1 Pro excels in full-project game coding across Unreal, Unity, and Godot, as well as parametric 3D modeling for Blender and CAD. ኪሚ ኤአይ (Kimi.ai) 2.8 ትሪሊዮን ጠቅላላ ፓራሜትሮች ያሉት እና ስራ ሲሰራ 104 ቢሊዮኑን ብቻ የሚጠቀመውን ኪሚ ኬ3.1 ፕሮ (Kimi K3.1 Pro) የተሰኘ አዲስ ሞዴል ይፋ አድርጓል። ይህ ሞዴል 1 ሚሊዮን ቶከን መረጃን በአንድ ጊዜ የማንበብ አቅም እና የተለያዩ የመረጃ አይነቶችን የመቀበል ችሎታ ያለው ሲሆን፣ ኩባንያው እስካሁን ካወጣቸው ለህዝብ ክፍት የሆኑ ሞዴሎች እጅግ ከፍተኛ አቅም ያለው ነው። ሞዴሉ በአንሪል (Unreal)፣ ዩኒቲ (Unity) እና ጎዶት (Godot) መድረኮች ላይ የተሟሉ የጌም ኮዲንግ ፕሮጀክቶችን ለመስራት፣ እንዲሁም በብሌንደር (Blender) እና ካድ (CAD) ላይ የ3D ዲዛይኖችን ለማዘጋጀት የላቀ ብቃት አለው። @webthreeth

Reverse Engineer Anything is a serious serious skill bruh 😭

Keyet new endatelu😂

OpenAI Unveils Dots: Always-On Personal AI Agents Powered by GPT-6 Astra OpenAI has announced the launch of Dots, a new class of personalized, always-on AI agents powered by the highly anticipated GPT-6 Astra model. Operating on dedicated cloud virtual machines, these agents are designed for persistent, background task handling without interrupting the user's workflow. Dots feature seamless voice conversations, integration with over 4,000 app connectors, and high customizability driven by continuous user feedback. ኦፕንኤአይ (OpenAI) በጉጉት በሚጠበቀው ጂፒቲ-6 አስትራ (GPT-6 Astra) ሞዴል የሚመሩ እና በቀን 24 ሰዓት ክፍት ሆነው የሚሰሩ ዶትስ (Dots) የተሰኙ አዳዲስ የግል የሰው ሰራሽ አስተውሎት ወኪሎችን በይፋ አስተዋውቋል። ለየብቻቸው በተመደቡ የክላውድ ሰርቨሮች (cloud VMs) ላይ የሚሰሩት እነዚህ ወኪሎች፣ የተጠቃሚውን ስራ ሳያስተጓጉሉ ተግባራትን ያለማቋረጥ ከበስተጀርባ ለማከናወን የተነደፉ ናቸው። ዶትስ የድምፅ ልውውጥን፣ ከ4,000 በላይ መተግበሪያዎች ጋር መገናኘትን፣ እና በተጠቃሚዎች አስተያየት መሰረት ራሳቸውን ማስተካከልን የመሳሰሉ አቅሞችን አካተዋል። @webthreeth

OpenAI Dev Day - Let's see

Eleven v4 & v4 Turbo: The Most Emotive AI Voices Yet — Ranked #1 ElevenLabs announces Eleven v4 and Eleven v4 Turbo as its most emotive and fastest text-to-speech models yet, topping Artificial Analysis rankings with an Elo of 1319 in provider voice quality. New architecture enables superior voice cloning from 10-second samples, inline tags for directing emotion, pacing, sound effects, and consistent multi-speaker dialogue across 90+ languages. https://x.com/ElevenLabs/status/2104572127617994917 @webthreeth

Claude Sonnet 5.5: Anthropic’s Faster, Cheaper Upgrade for Everyday AI Tasks Claude announces Sonnet 5.5 as the latest model in the Claude 5.5 family, delivering major gains over Sonnet 5 in agentic coding, knowledge work, computer use, and multidisciplinary reasoning across various benchmarks. It generates outputs over 30% faster, requires fewer tokens for tasks, and reduces per-task costs by up to 30% at the same pricing as Sonnet 5, excelling at practical applications like bug fixing, polished documents, slides, UI design, and fast iteration. @webthreeth

https://x.com/ManusAI/status/2104592659142869352 This shi looks insane

ManusAI Unveils Manus 2.0 and New Cue Personal Agent App ManusAI has announced the launch of Manus 2.0, a significant upgrade to its general AI agent designed to execute complex tasks and deliver tangible real-world results rather than simply answering questions. Alongside this major release, the company provided access to the new Manus Studio desktop application. ማኑስ ኤአይ (ManusAI) ጥያቄዎችን ከመመለስ ባለፈ ውስብስብ ተግባራትን በራሱ በማከናወን ተጨባጭ ውጤቶችን በማምጣት ላይ ያተኮረውን አዲሱን ማኑስ 2.0 (Manus 2.0) የኤአይ ወኪል በይፋ አስተዋውቋል። ከዚህ ትልቅ ማሻሻያ ጋር ተያይዞ፣ ኩባንያው ለዴስክቶፕ ኮምፒውተሮች የሚሆነውን ማኑስ ስቱዲዮ (Manus Studio) መተግበሪያ አቅርቧል። @webthreeth

Gemini 4.0.....Loading

Claude Opus 5.5: Frontier Performance at 40% Lower Cost Claude AI announces Opus 5.5, the first model in the Claude 5.5 family, claiming it matches prior frontier performance levels while using 40% less compute and generating outputs over 30% faster than Opus 5. The new model also improves agentic coding, computer use, knowledge work, and natural communication style. @webthreeth

Alibaba Unveils Qwen 4 Series at Apsara 2026, Outlining Trillion-Parameter Roadmap to ASI At the 2026 Apsara Conference, Alibaba’s Qwen team officially debuted its next-generation Qwen 4 lineup, introducing Qwen-4-Max, Qwen-4-Plus, Qwen-4-Flash, and the Qwen-4-27B model. The launch highlights significant architectural enhancements across both flagship and lightweight tiers, signaling Alibaba's continued push to stay at the absolute frontier of AI development. Beyond the immediate model rollouts, conference slides revealed an ambitious long-term scaling strategy designed to accelerate progress toward Artificial Superintelligence. Pre-training plans for upcoming Qwen 4.5 and Qwen 5 iterations aim for a massive 5 to 10 trillion parameters. To drive these advancements, the team is leveraging Recursive Self-Improvement across model training, real-time inference, and chip co-optimization, while also preparing a next-generation video model scheduled to launch in November. @webthreeth

Resets 🚇🚇

Grok 4.7 is here. SpaceXAI has launched Grok 4.7 as a meaningful upgrade over Grok 4.6 while keeping price and speed identical. The new model shows stronger endurance on hard tasks, more careful self-checking, and improved safeguards. Benchmarks reflect clear gains, including 64% on electrical engineering (up from 53%), 38% on multi-hour terminal work (up from 20.3%), and competitive CursorBench results at a lower cost per task than rivals. SpaceXAI Grok 4.7ን እንደ Grok 4.6 ላይ ትክክለኛ ማሻሻያ ሲያደርግ ዋጋውን እና ፍጥነቱን ተመሳሳይ አድርጎታል። አዲሱ ሞዴል በከባድ ሥራዎች ላይ ጠንካራ ጥንካሬን፣ የበለጠ ትኩረት ያለውን የራስ-ማረጋገጫን እና የተሻሉ መከላከያዎችን ያሳያል። በቤንችማርኮች ግልጽ ማሻሻያዎች ታይተዋል፤ በኤሌክትሪካል ኢንጂነሪንግ 64% (ከ53% ወደ ላይ)፣ በብዙ ሰዓታት የተርሚናል ሥራ 38% (ከ20.3% ወደ ላይ)፣ እና ከተወዳደሪዎች ዝቅተኛ የተግባር ዋጋ ጋር የተወዳደረ የCursorBench ውጤት። @webthreeth

Anthropic Rumored to Be Testing Opus 5.5 and Fable 5.2 Rumors indicate that Anthropic is currently stealth-testing Claude Opus 5.5 (claude-opus-5-5) under the codename claude-wafer-eap, with a potential release as early as Tuesday, while also conducting A/B testing of the next Fable model (Fable 5.2) via Claude Code, where users have observed strong performance in generating complex, interactive JavaScript animations and games in one shot without external dependencies. @webthreeth

Exclusive Sources : OpenAI's Codex Bot response to Grok Bot OpenAI preparing a new AI agent product tentatively called Codex Bot, built on technology from OpenClaw, the popular open-source agent framework that enables autonomous multi-agent workflows like code reviews and task automation. Originally slated for the week of September 17, 2026, the launch was postponed to the following week. @webthreeth

Projects redesigned: from folder to conversation Claude AI announced a redesigned Projects feature allowing one conversation in Claude Code to spawn and manage parallel cloud threads that continue working autonomously on tasks even after closing the app. New capabilities include briefing Claude like a chief of staff for multiple requests, shared memory across threads for context retention, an Overview panel for progress monitoring, and a Library for files and artifacts created by Claude. @webthreeth