GPT-4o Update Soars to Second Place on LM Arena Leaderboard
The update to GPT-4o significantly enhanced its ability to handle complex, multi-part user prompts with greater precision and context retention, a boon for technical users tackling intricate coding challenges and scientific queries where logical reasoning is key. OpenAI refined the model to not only excel in problem-solving but also to infuse responses with intuition and creativity—offering more human-like, imaginative insights—while cutting back on excessive emojis for a more professional tone.
Available immediately to paid ChatGPT subscribers (Plus and Pro tiers), with free users gaining access in the following weeks, this update caters to coders, writers, and researchers alike, delivering, for instance, elegantly optimized code alongside novel workarounds, highlighting OpenAI’s push to make GPT-4o a versatile, innovative tool across fields like software development and academic research.
@webthreeth
The update to GPT-4o significantly enhanced its ability to handle complex, multi-part user prompts with greater precision and context retention, a boon for technical users tackling intricate coding challenges and scientific queries where logical reasoning is key. OpenAI refined the model to not only excel in problem-solving but also to infuse responses with intuition and creativity—offering more human-like, imaginative insights—while cutting back on excessive emojis for a more professional tone.
Available immediately to paid ChatGPT subscribers (Plus and Pro tiers), with free users gaining access in the following weeks, this update caters to coders, writers, and researchers alike, delivering, for instance, elegantly optimized code alongside novel workarounds, highlighting OpenAI’s push to make GPT-4o a versatile, innovative tool across fields like software development and academic research.
@webthreeth
🔥1
I am really excited to see how the Chinese will react to the new updates from Claude (New Form of Reasoner), OpenAI (New Model Generation for Image, Audio Update, and LLM Model Update for 4o), and Gemini (AI Studio, Audio Overview).
The Chinese seem to follow open-source model. If they can generate similar prototype of products over the next 12 months and make it free, then the race for AGI will be very interesting.
It might be like the Cold War, when both countries have their own AGIs
The Chinese seem to follow open-source model. If they can generate similar prototype of products over the next 12 months and make it free, then the race for AGI will be very interesting.
It might be like the Cold War, when both countries have their own AGIs
Web 3.0 Ethiopia - DeFi & AI
Perplexity AI Unveils Deep Research Update for Next Week Perplexity AI is set to launch an enhanced version of its Deep Research feature next week, promising significant advancements in its analytical capabilities. This update will equip the tool with increased…
Perplexity's Strategic Shift: From Python to GoLang for a Next-Gen Deep Research Agent
Perplexity's CEO, Aravind Srinivas, recently announced a significant infrastructure overhaul, transitioning from Python to GoLang to support a new deep research agent capable of thinking for 30 minutes or more, with the goal of surpassing competitors like Manus in performance.
This shift, however, has led to unforeseen challenges in the core product, prompting Perplexity to temporarily slow down its rapid feature releases and prioritize long-term stability over short-term gains.
The decision reflects broader industry trends, as GoLang is often preferred for high-load projects due to its efficiency.
@webthreeth
Perplexity's CEO, Aravind Srinivas, recently announced a significant infrastructure overhaul, transitioning from Python to GoLang to support a new deep research agent capable of thinking for 30 minutes or more, with the goal of surpassing competitors like Manus in performance.
This shift, however, has led to unforeseen challenges in the core product, prompting Perplexity to temporarily slow down its rapid feature releases and prioritize long-term stability over short-term gains.
The decision reflects broader industry trends, as GoLang is often preferred for high-load projects due to its efficiency.
@webthreeth
Google Makes Gemini 2.5 Available for Free
Gemini 2.5, a powerful upgrade to its multimodal AI family, and made it available for free to users worldwide.
Gemini 2.5, which includes models like 2.5 Pro, offers significant improvements in reasoning, coding, and multimodal capabilities, allowing users to process text, images, and potentially video with greater accuracy and efficiency.
This free release democratizes access to advanced AI tools, enabling developers, researchers, and casual users to leverage features like enhanced prompt understanding and complex task handling without any cost.
@webthreeth
Gemini 2.5, a powerful upgrade to its multimodal AI family, and made it available for free to users worldwide.
Gemini 2.5, which includes models like 2.5 Pro, offers significant improvements in reasoning, coding, and multimodal capabilities, allowing users to process text, images, and potentially video with greater accuracy and efficiency.
This free release democratizes access to advanced AI tools, enabling developers, researchers, and casual users to leverage features like enhanced prompt understanding and complex task handling without any cost.
@webthreeth
When you deep it, It's actually quite funny how a general-purpose technology, capable of doing incredible things, becomes widely recognized—and suddenly, the pace of distraction becomes obvious to the world—all because it can create Japanese Studio Ghibli-style images.
It’s like working in the corporate world, where the effort you don’t particularly value ends up being the very thing that makes you valuable and in demand.
Or like that Dubai chocolate trend, which blew up simply because one girl didn’t know about pistachios and kunafa inside chocolate back in 2024—sparking everyone’s curiosity.
@webthreeth
It’s like working in the corporate world, where the effort you don’t particularly value ends up being the very thing that makes you valuable and in demand.
Or like that Dubai chocolate trend, which blew up simply because one girl didn’t know about pistachios and kunafa inside chocolate back in 2024—sparking everyone’s curiosity.
@webthreeth
Web 3.0 Ethiopia - DeFi & AI
I had an idea. So, I was thinking about uploading a ten minute discussion about the weekly updates of AI in the form of a podcast through Gemini Audio Overview. I use the Gemini Audio Overview daily to learn about new stuffs and I think people will like the…
_AI_&_Crypto_Unleashed_DeepSeek's_Brain_Boost,_Google's_Thinking.wav
20.7 MB
Greetings All,
Here is the first episode for the Weekly Updates on AI and DeFI. Entirely done on Gemini Audio Overview. Do you like it?
@webthreeth
Here is the first episode for the Weekly Updates on AI and DeFI. Entirely done on Gemini Audio Overview. Do you like it?
@webthreeth
🔥4
Alibaba Introduces QVQ-Max: A Visual Reasoning Model
Alibaba's Qwen division formally announced the release of QVQ-Max, a sophisticated visual reasoning model engineered to elevate the capabilities of artificial intelligence in processing and analyzing visual information.
This model, an evolution of the QvQ-72B-Preview introduced in December 2024, offers enhanced functionalities, including multi-image recognition, mathematical reasoning, and video comprehension, all integrated within the Qwen Chat platform. Featuring a methodical "Thinking" mechanism that elucidates its analytical process, QVQ-Max encourages user engagement for further refinement, underscoring Alibaba's dedication to advancing the frontier of AI innovation.
@webthreeth
Alibaba's Qwen division formally announced the release of QVQ-Max, a sophisticated visual reasoning model engineered to elevate the capabilities of artificial intelligence in processing and analyzing visual information.
This model, an evolution of the QvQ-72B-Preview introduced in December 2024, offers enhanced functionalities, including multi-image recognition, mathematical reasoning, and video comprehension, all integrated within the Qwen Chat platform. Featuring a methodical "Thinking" mechanism that elucidates its analytical process, QVQ-Max encourages user engagement for further refinement, underscoring Alibaba's dedication to advancing the frontier of AI innovation.
@webthreeth
Runway Launches Gen-4: An Image-to-Video AI
Runway released Gen-4 today, March 31, 2025, its newest AI model for creating media, bringing its image-to-video feature to paid subscribers and enterprise users.
Without much hype, Gen-4 improves on Gen-3 Alpha, offering better visuals, smoother motion, and easier control over keeping characters, objects, and settings consistent, all starting from one image.
Look at the video it generated from a photo.
Link - https://x.com/iamneubert/status/1906719031018582048?t=8_aSgsGtFC9GTqhadXDfYw&s=19
@webthreeth
Runway released Gen-4 today, March 31, 2025, its newest AI model for creating media, bringing its image-to-video feature to paid subscribers and enterprise users.
Without much hype, Gen-4 improves on Gen-3 Alpha, offering better visuals, smoother motion, and easier control over keeping characters, objects, and settings consistent, all starting from one image.
Look at the video it generated from a photo.
Link - https://x.com/iamneubert/status/1906719031018582048?t=8_aSgsGtFC9GTqhadXDfYw&s=19
@webthreeth
Web 3.0 Ethiopia - DeFi & AI
When you deep it, It's actually quite funny how a general-purpose technology, capable of doing incredible things, becomes widely recognized—and suddenly, the pace of distraction becomes obvious to the world—all because it can create Japanese Studio Ghibli…
Sam agrees with me.
Web 3.0 Ethiopia - DeFi & AI
Sam agrees with me.
I posted this on my LinkedIn and Telegram channel yesterday. OpenAI, just added one million users, over the past hour, as per Sam Altman post on X. This growth was equal to the one million users that happened when the GPT-3 model was released to the public in 2022.
It is actually quite the tale that Chat-GPT user acquisition strategy, didn't have as much result as in terms of added features like Deep Research and New Reasoner Models, rather it came from the ability to cartoonize and to generate anime-portraits of oneself.
It reminds me of Blue Ocean Strategy, of creating new market space from unexpected sources, although for one, I don't think OpenAI thought of this much impact for the company in terms of users growth for the platform.
I don't think any strategist around the world would expect the ability to animate oneself as the Blue Ocean for increase in users and revenue.
PS - Admins Personal Post
It is actually quite the tale that Chat-GPT user acquisition strategy, didn't have as much result as in terms of added features like Deep Research and New Reasoner Models, rather it came from the ability to cartoonize and to generate anime-portraits of oneself.
It reminds me of Blue Ocean Strategy, of creating new market space from unexpected sources, although for one, I don't think OpenAI thought of this much impact for the company in terms of users growth for the platform.
I don't think any strategist around the world would expect the ability to animate oneself as the Blue Ocean for increase in users and revenue.
PS - Admins Personal Post
What is a GPT-Wrapper?
A GPT-wrapper is an application or service that primarily uses a large language model like GPT to provide its core functionality. It often adds a user-friendly interface or specific features on top of the existing GPT model, making it easier for users to interact with and apply the AI's capabilities for particular tasks or industries.
While GPT wrappers can offer convenience and tailored experiences, they are sometimes criticized for lacking unique value beyond the underlying GPT model. These applications leverage the power of GPT to address specific user needs across various domains.
They aim to simplify complex AI interactions, making advanced language models accessible to a broader audience without requiring deep technical knowledge. Some of the examples of GPT wrappers include Jasper, Notion AI, GitHub Copilot, and ChatPRD.
@webthreeth
A GPT-wrapper is an application or service that primarily uses a large language model like GPT to provide its core functionality. It often adds a user-friendly interface or specific features on top of the existing GPT model, making it easier for users to interact with and apply the AI's capabilities for particular tasks or industries.
While GPT wrappers can offer convenience and tailored experiences, they are sometimes criticized for lacking unique value beyond the underlying GPT model. These applications leverage the power of GPT to address specific user needs across various domains.
They aim to simplify complex AI interactions, making advanced language models accessible to a broader audience without requiring deep technical knowledge. Some of the examples of GPT wrappers include Jasper, Notion AI, GitHub Copilot, and ChatPRD.
@webthreeth
👍2
SelamGPT: A Multilingual Al ChatBot for Ethiopia
SelamGPT, developed by @BanaCodes, is a conversational Al developed in Ethiopia with its support for Amharic, Tigrinya, and English.
This cutting-edge tool offers an impressive suite of features, including voice mode for seamless speech-to-speech interaction (akin to ChatGPT), voice recognition across all three languages, and a read-aloud function.
The ChatBot isn't available for the public yet. The model is developed with a foundation model (like a multilingual GPT variant), then layered on specialized modules for voice, and language processing.
@webthreeth
SelamGPT, developed by @BanaCodes, is a conversational Al developed in Ethiopia with its support for Amharic, Tigrinya, and English.
This cutting-edge tool offers an impressive suite of features, including voice mode for seamless speech-to-speech interaction (akin to ChatGPT), voice recognition across all three languages, and a read-aloud function.
The ChatBot isn't available for the public yet. The model is developed with a foundation model (like a multilingual GPT variant), then layered on specialized modules for voice, and language processing.
@webthreeth
👍3🔥2👏1
It's truly wonderful to see channels like CryptoTalk Ethiopia (@cryptotalk_et) and ET Netsa Apps—pioneers in Ethiopia's blockchain industry—give a shout-out to our channel.
Their generous support in promoting our news and insights to a wider audience, completely free of charge, means a great deal. I want to express my heartfelt thanks and deep appreciation to them. It truly means a lot.
If you're interested in staying informed about the latest developments in blockchain and crypto, I highly encourage you to follow their Telegram channels. You won’t be disappointed.
Thank you very much again. It means a lot.
Their generous support in promoting our news and insights to a wider audience, completely free of charge, means a great deal. I want to express my heartfelt thanks and deep appreciation to them. It truly means a lot.
If you're interested in staying informed about the latest developments in blockchain and crypto, I highly encourage you to follow their Telegram channels. You won’t be disappointed.
Thank you very much again. It means a lot.
👍9❤1
Manus AI Pivots to Subscription Model with New Tiers and Claude 3.7 Upgrade
Manus AI, launched by the Chinese startup Butterfly Effect under the Monica.im brand on March 6, 2025, has shifted from its initial open-source strategy for its autonomous, multi-agent AI system to a commercial, subscription-based model.
Manus introduced two paid tiers on March 31, 2025: a $39/month plan with 3,900 credits and two simultaneous tasks, and a $199/month plan offering 19,900 credits, five simultaneous tasks, and priority access during peak usage, alongside an iOS app release and an upgrade to Anthropic’s Claude 3.7 Sonnet model for its backend.
@webthreeth
Manus AI, launched by the Chinese startup Butterfly Effect under the Monica.im brand on March 6, 2025, has shifted from its initial open-source strategy for its autonomous, multi-agent AI system to a commercial, subscription-based model.
Manus introduced two paid tiers on March 31, 2025: a $39/month plan with 3,900 credits and two simultaneous tasks, and a $199/month plan offering 19,900 credits, five simultaneous tasks, and priority access during peak usage, alongside an iOS app release and an upgrade to Anthropic’s Claude 3.7 Sonnet model for its backend.
@webthreeth
Disappointed is an understatement. I really wanted to try it out, but understandable if their foundational model is running on Claude 3.7 as they are burning cash to run the model.
Etenu Jebena: An Ethiopian-based Cultural Crypto Adventure Begins!
Etenu Jebena, the pioneering Ethiopian-themed Telegram airdrop game, has officially launched. Powered by ETNetsa Coin ($ETN), this TON blockchain-based mini-game seamlessly integrates cultural heritage with modern technology, allowing players to collect $ETN points through engaging gameplay.
Inspired by Ethiopia’s storied traditions, it offers a unique and rewarding experience while serving as an innovative entry point into the blockchain ecosystem for Ethiopians and across Sub-Sahara Africa.
https://t.me/ET_Netsa_Airdrop_bot?start=r2134896128
@webthreeth
Etenu Jebena, the pioneering Ethiopian-themed Telegram airdrop game, has officially launched. Powered by ETNetsa Coin ($ETN), this TON blockchain-based mini-game seamlessly integrates cultural heritage with modern technology, allowing players to collect $ETN points through engaging gameplay.
Inspired by Ethiopia’s storied traditions, it offers a unique and rewarding experience while serving as an innovative entry point into the blockchain ecosystem for Ethiopians and across Sub-Sahara Africa.
https://t.me/ET_Netsa_Airdrop_bot?start=r2134896128
@webthreeth
👍1
OpenAI Unveils Academy Website: Free AI Education for All
OpenAI has recently launched its highly anticipated OpenAI Academy website, a free online platform designed to democratize access to artificial intelligence education and empower individuals worldwide.
The Academy offers a comprehensive range of resources, including tutorials, webinars, hands-on practice, and live events, catering to diverse audiences such as students, educators, developers, and professionals. With content spanning from foundational AI literacy to advanced topics like large language model integration, the platform aims to bridge the knowledge gap and equip users with practical skills to leverage AI in real-world scenarios.
Built in collaboration with partners like Common Sense Media and academic institutions, OpenAI Academy reflects the organization’s mission to make AI accessible globally, fostering a collaborative community dedicated to innovation.
Link - https://academy.openai.com/
@webthreeth
OpenAI has recently launched its highly anticipated OpenAI Academy website, a free online platform designed to democratize access to artificial intelligence education and empower individuals worldwide.
The Academy offers a comprehensive range of resources, including tutorials, webinars, hands-on practice, and live events, catering to diverse audiences such as students, educators, developers, and professionals. With content spanning from foundational AI literacy to advanced topics like large language model integration, the platform aims to bridge the knowledge gap and equip users with practical skills to leverage AI in real-world scenarios.
Built in collaboration with partners like Common Sense Media and academic institutions, OpenAI Academy reflects the organization’s mission to make AI accessible globally, fostering a collaborative community dedicated to innovation.
Link - https://academy.openai.com/
@webthreeth
AI-Generated Short Film on Ethiopia’s Leaders Premieres April 7
The short film “ሊገናኙ ነው” will be released on Monday, April 7 on YouTube. Created by EHUD AI Studio, it is an AI-generated political psychological thriller that explores the stories and legacies of Ethiopia’s past leaders. The film was first screened on March 23, and based on viewer feedback, the creators made improvements to the story and visuals. This updated version aims to deliver a sharper and more thought-provoking experience. You can watch it on the EHUD AI Studio YouTube channel next Monday.
https://www.youtube.com/@ehudai
@webthreeth
The short film “ሊገናኙ ነው” will be released on Monday, April 7 on YouTube. Created by EHUD AI Studio, it is an AI-generated political psychological thriller that explores the stories and legacies of Ethiopia’s past leaders. The film was first screened on March 23, and based on viewer feedback, the creators made improvements to the story and visuals. This updated version aims to deliver a sharper and more thought-provoking experience. You can watch it on the EHUD AI Studio YouTube channel next Monday.
https://www.youtube.com/@ehudai
@webthreeth
👍7
I have posted it now. Please have a look in the LinkedIn page of the creator of Ehud AI Studio.
👍1
Copilot experiments reveal Deep Research, Podcasts and Actions
Microsoft Copilot is developing a range of experimental features, including a new sidebar layout and a Labs icon under “Discover” for Pro subscribers in the US to test upcoming functionalities.
Among these, Copilot is exploring "Avatars" for voice mode, a mysterious “Generative Layout,” and “Actions” tied to "Agents" and "Computer Use," which could enable task execution and browsing akin to OpenAI’s Operator, possibly limited to the Edge browser. Additionally, a “Deep Research” feature, similar to ChatGPT’s, is in the works, taking about 10 minutes per task, alongside “Pages,” which lets users create editable documents from responses with side chat support for better organization.
A hidden “Podcasts” feature is also being tested, potentially allowing users to generate audio content from responses, with a UI like Daily Briefings and possible integration into the Discover feed, resembling NotebookLM and Gemini.
@webthreeth
Microsoft Copilot is developing a range of experimental features, including a new sidebar layout and a Labs icon under “Discover” for Pro subscribers in the US to test upcoming functionalities.
Among these, Copilot is exploring "Avatars" for voice mode, a mysterious “Generative Layout,” and “Actions” tied to "Agents" and "Computer Use," which could enable task execution and browsing akin to OpenAI’s Operator, possibly limited to the Edge browser. Additionally, a “Deep Research” feature, similar to ChatGPT’s, is in the works, taking about 10 minutes per task, alongside “Pages,” which lets users create editable documents from responses with side chat support for better organization.
A hidden “Podcasts” feature is also being tested, potentially allowing users to generate audio content from responses, with a UI like Daily Briefings and possible integration into the Discover feed, resembling NotebookLM and Gemini.
@webthreeth