This Week I Learned - Week #26 2025
This Week I Learned -
* Gemini CLI is an open-source AI agent that brings Gemini directly into your terminal, with MCP support for extensibility and Human in the Loop for oversight.
* Prompting only changes how the model responds. Fine-tuning changes the weights—but it’s costly, brittle, and still static. Neither gives the model new knowledge in real-time. With Retrieval-Augmented Generation (RAG), you “Retrieve” the most relevant data for the user’s query from your database, use this data to “Augment” the prompt you send to the LLM, and then let it “Generate” a response based on the user query + prompt + retrieved data. RAG doesn’t modify the model; it modifies the input. It retrieves relevant context (docs, tickets, policies) and feeds it into the prompt at inference time. No retraining. Just better answers, grounded in your own data.
* Google Apps Script Web Apps can act as powerful servers for open protocols like Model Context Protocol (MCP) and Agent2Agent (A2A) Protocol, enabling sophisticated AI agent systems.
* Dolphin is an open source OCR model from ByteDance. It reads documents like humans do, preserving natural reading order. Due to parallel parsing on each element, it has top-tier performance on complex document parsing tasks. It follows an "analyze-then-parse" paradigm. This means that the model first performs page-level layout analysis to identify figures, captions, paragraphs and more. Next it can parse each of the recognized elements in parallel.
* Artificial intelligence has created a new digital divide, fracturing the world between nations with the computing power for building cutting-edge A.I. systems and those without. The gap stems partly from a component everyone wants: a microchip known as a graphics processing unit, or GPU. The biggest beneficiaries by far are the United States, China and the European Union. Those regions host more than half of the world’s most powerful data centers, which are used for developing the most complex A.I. systems. Only 32 countries, or about 16 percent of nations, have these large facilities filled with microchips and computers, giving them what is known in industry parlance as “compute power.” - NYT
* Few of Eleven Labs offerings:
- Speech-to-speech (or voice conversion) allows you to convert one voice (source voice) into another (cloned voice) while preserving the tone and delivery of the original voice.
- Professional voice cloning produces clones that are virtually indistinguishable from the real thing, requiring a minimum of 30 minutes of clean audio to generate high-quality, lifelike voice clones. Perfect for creating professional-grade audio for videos, audiobooks, podcasts, video games, and more, where authenticity and quality are paramount.
* Infosys has 400 Gen AI projects underway.
* GitHub repo with tutorials covering 75+ AI LLM apps based on AI Agents and RAG
* Data visualizations using LLMs Prompt - Slides
* Visualization of all books that have been published with an ISBN - International Standard Book Numbers (ISBNs) are 13-digit numbers that are assigned to almost all published books. Since the first three digits are fixed (currently only 978- and 979-) and the last digit is a checksum, this means the total ISBN13-Space only has two billion slots.
* Scrollytelling - Blue oaks, named for the color their leaves take on deep in the summer season, are revered among dendrochronologists. They can live for more than 550 years, survive on as little as 10 inches of average annual rainfall and are among the most drought-adapted of any tree species in California.
* A timeline of world events between 1300-2000 CE
* India Street Lettering is an archive of the shared typographic culture that thrives in the country’s urban spaces. Built over a decade, this ongoing effort by Pooja Saxena is focused on meticulously documenting, annotating and geo-tagging public lettering made by analog means from around India.
* Last year, 68 percent of the world’s population used the internet, up from 33 percent in 2012.
* Australia has raised the age for opening social media accounts to 16
* Students participating in spelling bees, and especially the Scripps Spelling Bee, which turns 100 this year, have been deemed ‘spellebrities’. Many of these spellebrities are of Indian origin, given the event has been dominated by desis since Nupur Lala’s 1999 win.
* Ecuador's Amazon rainforest covers nearly 40% of the country.
* Around 40,000 humpback whales traverse the "Humpback Highway," a migratory corridor along Australia's east coast. These massive creatures journey from their feeding grounds in the frigid waters of Antarctica to the tropical breeding areas off the coast of Queensland.
* New York, with a population of over 2 million Jews, ranks as the city with the second-largest Jewish community in the world, following Tel Aviv.
* Africa is the only continent where the Tropic of Cancer, the equator and the Tropic of Capricorn all pass through. Algeria is Africa's largest country by area, and Nigeria is its largest by population.
* Food Twin Map visualizes the complex network of food production and distribution across the globe.
* Our food environments—the type and quality of food that pervades our schools, workplaces, and neighborhoods—influence our diets as much as our tastes do. And our food environments are shaped by our incomes, our government’s choices, and our desire for convenience, as well as active manipulation by the food industry, through things like marketing campaigns and lobbying for agricultural subsidies. - New Yorker (article requires subscription but audio transcript is free to listen)
* Recipes for the fast-food staple have spread online like open-source code.
* In FY25, Blinkit, Zepto, and Instamart together made over Rs 3,000 crore in ad revenue.
* Hyderabad has only 3,500 traffic personnel for nearly 90 lakh vehicles. Out of the 600 junctions in Hyderabad about 400 of them are only manned by police. - ToI
* "A clever person solves a problem. A wise person avoids it." - Albert Einstein
* "Like oil, data is valuable only when it flows – fuelling services, enabling innovation, and empowering individuals in motion. To unlock its true value in the digital economy, we must build systems where data moves with consent, with control, and with cryptographic integrity." - Siddharth Sharma, Digi Yatra
Comments
Post a Comment