Unveiling the Link: Global M2 Money Supply and Bitcoin Price DynamicsM2 money supply refers to a broad measure of money in circulation, including cash, checking deposits, savings deposits, money market securities, and other time deposits, which reflects the total liquid assets available in an economy.
The chart "Bitcoin & Global M2" illustrates a correlation between global M2 money supply and Bitcoin's price (XBT-USD Cross Rate) from December 12, 2019, to February 25, 2025, suggesting that Bitcoin's price tends to lag behind changes in global M2 by approximately 10 weeks.
This indicates that increases in global liquidity, as represented by M2, may drive demand for assets like Bitcoin, potentially influencing its price over time, though with a delay.
@webthreeth
Unveiling GPT-4.5: OpenAI’s Next Leap in AI InnovationOpenAI announced a research preview of GPT-4.5 on February 28, 2025 (today). This model is described as their "largest and best model for chat yet" and is being rolled out initially to ChatGPT Pro users, with plans to extend access to Plus and Team users next week, followed by Enterprise and Education users the week after.
GPT-4.5 includes features like access to up-to-date information via search, support for file and image uploads, and a canvas feature for working on writing and code. However, it does not currently support multimodal features like Voice Mode, video, or screensharing in ChatGPT.
@webthreeth
GPT-4.5: Why OpenAI’s Latest Falls Short in Math and Code
GPT-4.5, released by OpenAI on February 28, 2025, ranks below specialized models in math and coding benchmarks because it’s a general-purpose conversationalist, not a niche problem-solver like OpenAI’s o1 or o3-mini, scoring 36.7% on AIME 2024 (vs. o1’s 83/120 or o3-mini’s 102/120).
It is also lagging behind frontier coders like DeepSeek v3 or Claude Sonnet 3.7 on tests like HumanEval—its modest 5% boost over GPT-4o reflecting diminishing returns from massive compute focused on broad fluency rather than precision, with trade-offs favoring safety and polish (78% hallucination reduction on PersonQA) over the raw, iterative reasoning needed for advanced math or flawless code.
@webthreeth
GPT-4.5, released by OpenAI on February 28, 2025, ranks below specialized models in math and coding benchmarks because it’s a general-purpose conversationalist, not a niche problem-solver like OpenAI’s o1 or o3-mini, scoring 36.7% on AIME 2024 (vs. o1’s 83/120 or o3-mini’s 102/120).
It is also lagging behind frontier coders like DeepSeek v3 or Claude Sonnet 3.7 on tests like HumanEval—its modest 5% boost over GPT-4o reflecting diminishing returns from massive compute focused on broad fluency rather than precision, with trade-offs favoring safety and polish (78% hallucination reduction on PersonQA) over the raw, iterative reasoning needed for advanced math or flawless code.
@webthreeth
Elicit AI Outshines Other Tools in Deep Research: Expert PhD RatingsSeventeen external evaluators with PhDs rated various AI tools on a scale of 0 to 10 for their effectiveness in deep research tasks, with Elicit Reports receiving the highest score of 7.3, making it the most favorably perceived tool among the group.
Elicit AI, developed by Ought, is an advanced AI-powered research assistant designed to streamline and automate research. It uses language models to find relevant academic papers. It has access to over 125 million papers from the Semantic Scholar Corpus.
However, it’s best suited for empirical research in fields like biomedicine and machine learning, and it may not handle non-academic or theoretical questions well.
In comparison, ChatGPT Deep Research scored 6.1, Perplexity Deep Research received a 4.6, and Gemini Deep Research had the lowest rating at 3.9, indicating Elicit’s superior performance for deep research as of February 28, 2025, according to these expert evaluators.
@webthreeth
In my POV, I think ChatGPT Deep Research is the best, but the Elicit AI Deep Research is also very good. Especially, in referring to scholar papers, I think it is better than ChatGPT Deep Research...
Should I share a research I did with both to showcase their depth?
Please like it.
Should I share a research I did with both to showcase their depth?
Please like it.
👍4
Low Graduation of Tokens from Pump Fun : Why?
The PumpFun Stats table, updated as of February 28, 2025, provides a snapshot of token activity on the PumpFun platform across various time periods.
The data highlights a consistent trend of many tokens being launched on the platform, with significantly fewer tokens graduating in each time period, reflecting the dynamic nature of token creation and progression on PumpFun.
The low token graduation rate on the PumpFun platform can be attributed to several factors. Many tokens fail to graduate—reaching the $69,000 market cap required to move to Raydium—because they lack sustained community interest or viral appeal, which is crucial for memecoins to gain traction.
The memecoin market is highly volatile and driven by hype, which can fade quickly, leaving most tokens stagnant or losing value shortly after launch.
@webthreeth
The PumpFun Stats table, updated as of February 28, 2025, provides a snapshot of token activity on the PumpFun platform across various time periods.
The data highlights a consistent trend of many tokens being launched on the platform, with significantly fewer tokens graduating in each time period, reflecting the dynamic nature of token creation and progression on PumpFun.
The low token graduation rate on the PumpFun platform can be attributed to several factors. Many tokens fail to graduate—reaching the $69,000 market cap required to move to Raydium—because they lack sustained community interest or viral appeal, which is crucial for memecoins to gain traction.
The memecoin market is highly volatile and driven by hype, which can fade quickly, leaving most tokens stagnant or losing value shortly after launch.
@webthreeth
Zemenu, a company focused on advancing AI-driven solutions for Ethiopia, has launched an early access program for its next-generation AI assistant, Lucy, as of March 2025.
The initiative, detailed in a recent weekly update, invites a limited number of users to test Lucy’s capabilities as the company works to scale its server infrastructure and refine key areas such as accent refinement for Ethiopian linguistic nuances, performance optimization, security, and user experience.
Interested individuals can apply via the official website, https://lucy.zemenu.org, to participate in shaping Lucy into a seamless, accessible, and culturally relevant AI tool for the Ethiopian market, with Zemenu emphasizing the importance of user feedback in this developmental phase.
@webthreeth
The initiative, detailed in a recent weekly update, invites a limited number of users to test Lucy’s capabilities as the company works to scale its server infrastructure and refine key areas such as accent refinement for Ethiopian linguistic nuances, performance optimization, security, and user experience.
Interested individuals can apply via the official website, https://lucy.zemenu.org, to participate in shaping Lucy into a seamless, accessible, and culturally relevant AI tool for the Ethiopian market, with Zemenu emphasizing the importance of user feedback in this developmental phase.
@webthreeth
GPT-4.5: Paving the Way for Advanced Reasoning Models
The chart highlights GPT-4.5’s potential role as a foundational model for future reasoning systems, estimating its profound impact on performance as measured by the GPQA Diamond Benchmark Accuracy.
With a strong baseline accuracy of 69%, GPT-4.5 surpasses earlier models like GPT-4, GPT-4 Turbo, GPT-4o, and GPT-3.5, which lag behind with lower accuracies. By projecting the integration of advanced reasoning, GPT-4.5’s accuracy could reach an estimated 85%–95%, a +20-point leap that illustrates the potential power of reasoning applied to its base.
Meanwhile, further training has already yielded known reasoning results for derivative models like o3-mini (77%) and o1-mini (69%), underscoring GPT-4.5’s projected critical role in driving sophisticated AI reasoning and dramatically boosting base models’ ability to handle complex tasks with precision and insight.
@webthreeth
The chart highlights GPT-4.5’s potential role as a foundational model for future reasoning systems, estimating its profound impact on performance as measured by the GPQA Diamond Benchmark Accuracy.
With a strong baseline accuracy of 69%, GPT-4.5 surpasses earlier models like GPT-4, GPT-4 Turbo, GPT-4o, and GPT-3.5, which lag behind with lower accuracies. By projecting the integration of advanced reasoning, GPT-4.5’s accuracy could reach an estimated 85%–95%, a +20-point leap that illustrates the potential power of reasoning applied to its base.
Meanwhile, further training has already yielded known reasoning results for derivative models like o3-mini (77%) and o1-mini (69%), underscoring GPT-4.5’s projected critical role in driving sophisticated AI reasoning and dramatically boosting base models’ ability to handle complex tasks with precision and insight.
@webthreeth
President Trump has announced plans for a U.S. Crypto Reserve featuring prominent digital assets like BTC, ETH, XRP, SOL, and ADA, fulfilling his campaign pledge to position himself as a "crypto president" and advance the nation's role in the digital asset landscape.
This development, highlighted in a July 27, 2024, post by WatcherGuru on X, comes amid a White House Crypto Summit and a broader regulatory pivot favoring the crypto industry, signaling the administration's intent to cement the U.S. as a global leader in this space.
@webthreeth
This development, highlighted in a July 27, 2024, post by WatcherGuru on X, comes amid a White House Crypto Summit and a broader regulatory pivot favoring the crypto industry, signaling the administration's intent to cement the U.S. as a global leader in this space.
@webthreeth
The AI Revolution in Writing: How LLMs Are Transforming Human Communication
Over the past 18 months, human written communication has experienced an unprecedented transformation, propelled by the widespread adoption of large language models (LLMs). A study conducted by researchers from Stanford University, the University of Washington, and Emory University, published by September 2024, found that 18% of financial consumer complaints, 24% of press releases, 15% of job postings, and 14% of UN press releases exhibited signs of LLM-assisted writing—though the methodology suggests this figure may underestimate actual usage.
Detailed in the paper "Widespread Adoption of Large Language Model-Assisted Writing Across Society," the research highlights a profound societal shift, driven by the pervasive integration of LLMs like ChatGPT. This trend underscores their growing impact on communication practices, influencing businesses, consumers, and international organizations alike across diverse domains.
@webthreeth
Over the past 18 months, human written communication has experienced an unprecedented transformation, propelled by the widespread adoption of large language models (LLMs). A study conducted by researchers from Stanford University, the University of Washington, and Emory University, published by September 2024, found that 18% of financial consumer complaints, 24% of press releases, 15% of job postings, and 14% of UN press releases exhibited signs of LLM-assisted writing—though the methodology suggests this figure may underestimate actual usage.
Detailed in the paper "Widespread Adoption of Large Language Model-Assisted Writing Across Society," the research highlights a profound societal shift, driven by the pervasive integration of LLMs like ChatGPT. This trend underscores their growing impact on communication practices, influencing businesses, consumers, and international organizations alike across diverse domains.
@webthreeth
Trump’s Crypto Reserve Sparks Market Surge: Bitcoin Up 9%
After Trump revealed on social media that the reserve would include major cryptocurrencies like Bitcoin (BTC), Ethereum (ETH), XRP, Solana (SOL), and Cardano (ADA), aiming to position the U.S. as the "Crypto Capital of the World."
This decision, formalized through an executive order, has driven Bitcoin’s price to $93,125.78 with an 8.94% increase, Ethereum to $2,479.56 with an 11.31% rise, and other assets like XRP surging 30.82%, as shown in the image.
The move signals a pro-crypto shift, boosting investor confidence, promoting regulatory clarity, and accelerating institutional adoption, though questions remain about funding, management, and long-term implications, including potential congressional approval and the use of seized crypto assets.
@webthreeth
After Trump revealed on social media that the reserve would include major cryptocurrencies like Bitcoin (BTC), Ethereum (ETH), XRP, Solana (SOL), and Cardano (ADA), aiming to position the U.S. as the "Crypto Capital of the World."
This decision, formalized through an executive order, has driven Bitcoin’s price to $93,125.78 with an 8.94% increase, Ethereum to $2,479.56 with an 11.31% rise, and other assets like XRP surging 30.82%, as shown in the image.
The move signals a pro-crypto shift, boosting investor confidence, promoting regulatory clarity, and accelerating institutional adoption, though questions remain about funding, management, and long-term implications, including potential congressional approval and the use of seized crypto assets.
@webthreeth
Perplexity and Deutsche Telekom Pioneer IoT Innovation with AI Phone Integration
Perplexity, an innovative AI company, has teamed up with Deutsche Telekom to bring its Perplexity Assistant directly to users through the telecom giant’s new AI Phone, marking a significant step in integrating advanced AI into everyday mobile experiences.
This collaboration builds on Deutsche Telekom’s ambitious vision, first unveiled at the 2024 Mobile World Congress, of an app-free smartphone powered entirely by artificial intelligence—a concept now brought to life with Perplexity’s cutting-edge technology.
The partnership reflects Perplexity’s broader mission to expand its presence in the mobile ecosystem, following its earlier integration with Paytm, and aims to deliver seamless, voice-activated assistance to users across various platforms, redefining how we interact with our devices.
@webthreeth
Perplexity, an innovative AI company, has teamed up with Deutsche Telekom to bring its Perplexity Assistant directly to users through the telecom giant’s new AI Phone, marking a significant step in integrating advanced AI into everyday mobile experiences.
This collaboration builds on Deutsche Telekom’s ambitious vision, first unveiled at the 2024 Mobile World Congress, of an app-free smartphone powered entirely by artificial intelligence—a concept now brought to life with Perplexity’s cutting-edge technology.
The partnership reflects Perplexity’s broader mission to expand its presence in the mobile ecosystem, following its earlier integration with Paytm, and aims to deliver seamless, voice-activated assistance to users across various platforms, redefining how we interact with our devices.
@webthreeth
GPT-4.5 Tops ChatArena: A Win for OpenAI Amid Benchmarking Debate
The competitive landscape of AI development is heating up, with OpenAI’s GPT-4.5 recently taking the lead on ChatArena (now lmarena.ai), a widely recognized platform for comparing AI models.
This shift underscores the ongoing race among industry leaders like OpenAI, Google, and xAI to deliver cutting-edge performance, as advancements in natural language processing continue to accelerate.
@webthreeth
The competitive landscape of AI development is heating up, with OpenAI’s GPT-4.5 recently taking the lead on ChatArena (now lmarena.ai), a widely recognized platform for comparing AI models.
This shift underscores the ongoing race among industry leaders like OpenAI, Google, and xAI to deliver cutting-edge performance, as advancements in natural language processing continue to accelerate.
@webthreeth
Google Launches Revolutionary Data Science Agent in Colab, Powered by Gemini 2.0
Google has unveiled its groundbreaking Data Science Agent in Google Colab, an AI-powered tool designed to transform data analysis workflows for researchers, data scientists, and developers worldwide.
Powered by the advanced Gemini 2.0 model, the agent allows users to upload datasets (up to 1GB in CSV, JSON, or .txt format) and describe their analysis objectives in natural language, automatically generating fully functional, executable Colab notebooks
Available for free to users aged 18+ in select countries and languages, the Data Science Agent enhances Colab’s serverless notebook environment, building on Google’s ongoing integration of AI-driven features.
@webthreeth
Google has unveiled its groundbreaking Data Science Agent in Google Colab, an AI-powered tool designed to transform data analysis workflows for researchers, data scientists, and developers worldwide.
Powered by the advanced Gemini 2.0 model, the agent allows users to upload datasets (up to 1GB in CSV, JSON, or .txt format) and describe their analysis objectives in natural language, automatically generating fully functional, executable Colab notebooks
Available for free to users aged 18+ in select countries and languages, the Data Science Agent enhances Colab’s serverless notebook environment, building on Google’s ongoing integration of AI-driven features.
@webthreeth
Leyu AI: Empowering Ethiopia Through Crowdsourced Language Data
Leyu is an innovative platform developed by iCog Labs, focused on addressing the scarcity of high-quality datasets for Ethiopian languages to enhance AI development. Launched to bridge information gaps and language barriers, Leyu—meaning "to identify" in Amharic—crowdsources datasets for low-resource languages such as Amharic, Afaan Oromo, Tigrinya, Af-Somali, and Sidama, starting with speech data and aiming to expand into video and image datasets.
By leveraging smartphone accessibility, Leyu engages local communities in data creation, offering micro-work opportunities while ensuring fair compensation and data accuracy through a hybrid human-automated approach.
Source - https://lnkd.in/dmNRuWYz
@webthreeth
Opera Unveils Browser Operator: A New Era of Agentic Browsing
On March 3, 2025, Opera, the innovative Norwegian browser company, launched the Browser Operator, a native AI agent integrated into its browser, heralding what it calls "agentic browsing."
Unveiled at Mobile World Congress 2025 in Barcelona, this feature allows users to delegate tasks like shopping or booking tickets—think "buy a pack of 10 white Nike tennis socks in size 12"—with the AI autonomously navigating websites while pausing for user approval on sensitive steps like payments.
Powered by Opera’s AI Composer Engine, the Browser Operator runs locally on devices for enhanced privacy, distinguishing it from cloud-dependent competitors, and offers transparency by letting users oversee its actions.
Currently in preview, with a broader rollout planned via Opera’s AI Feature Drop program, this builds on the company’s AI legacy, such as its Aria assistant, positioning Opera as a trailblazer among major browsers.
@webthreeth
On March 3, 2025, Opera, the innovative Norwegian browser company, launched the Browser Operator, a native AI agent integrated into its browser, heralding what it calls "agentic browsing."
Unveiled at Mobile World Congress 2025 in Barcelona, this feature allows users to delegate tasks like shopping or booking tickets—think "buy a pack of 10 white Nike tennis socks in size 12"—with the AI autonomously navigating websites while pausing for user approval on sensitive steps like payments.
Powered by Opera’s AI Composer Engine, the Browser Operator runs locally on devices for enhanced privacy, distinguishing it from cloud-dependent competitors, and offers transparency by letting users oversee its actions.
Currently in preview, with a broader rollout planned via Opera’s AI Feature Drop program, this builds on the company’s AI legacy, such as its Aria assistant, positioning Opera as a trailblazer among major browsers.
@webthreeth
ChatGPT Feature Status and Access Levels as of March 4, 2025
The table highlights that the GPT-4o model offers the most comprehensive features, including unlimited access for Pro accounts, UK/EU access, support for image and PDF attachments, spreadsheet integration, canvas search, data analysis, DALL-E image generation, custom GPTs, memory, and custom instructions, though Free account access is limited.
In contrast, models like GPT-4.5 and GPT-4 have restricted access for Plus and Team accounts, with some features like custom GPTs unavailable. Reasoning models (o1, o1-mini, o3-mini-high) and agent models (GPT-4o with tasks, Deep Research, Operator) show varying levels of access and functionality, with Free accounts generally having minimal or no access.
Advanced features like Voice and Sora (video generation) are available only to paid accounts, with Sora offering up to 1080p 20s+ relaxed unlimited video for Advanced accounts, while Free accounts are excluded from these capabilities.
@webthreeth
The table highlights that the GPT-4o model offers the most comprehensive features, including unlimited access for Pro accounts, UK/EU access, support for image and PDF attachments, spreadsheet integration, canvas search, data analysis, DALL-E image generation, custom GPTs, memory, and custom instructions, though Free account access is limited.
In contrast, models like GPT-4.5 and GPT-4 have restricted access for Plus and Team accounts, with some features like custom GPTs unavailable. Reasoning models (o1, o1-mini, o3-mini-high) and agent models (GPT-4o with tasks, Deep Research, Operator) show varying levels of access and functionality, with Free accounts generally having minimal or no access.
Advanced features like Voice and Sora (video generation) are available only to paid accounts, with Sora offering up to 1080p 20s+ relaxed unlimited video for Advanced accounts, while Free accounts are excluded from these capabilities.
@webthreeth