This Week I Learned - Week 7 2026
This Week I Learned -
* Python turns 35 this month. It started as a hobby project in 1991.
* A sovereign cloud is a type of cloud service where data is both legally protected and physically stored within a country’s borders. Global spending on sovereign cloud infrastructure as a service (IaaS) is expected to reach $80 billion in 2026, marking a 35.6% increase from 2025, according to Gartner.
* Just 3.3% of Copilot Chat users pay for it
* Kimi K2.5 is an open-source model from Moonshot AI. Kimi K2.5 Built as a native multimodal model, K2.5 delivers state-of-the-art coding and vision capabilities and a self-directed agent swarm paradigm. Kimi K2.5 is available via Kimi.com, the Kimi App, the API, and Kimi Code. Kimi.com & Kimi App now supports 4 modes: K2.5 Instant, K2.5 Thinking, K2.5 Agent, and K2.5 Agent Swarm (Beta). Beyond text prompts, K2.5 excels at coding with vision. By reasoning over images and video, K2.5 improves image/video-to-code generation and visual debugging, lowering the barrier for users to express intent visually. K2.5 can reconstruct a website from video.
* Perplexity powers Deep Search with Kimi K2 Thinking, using chain of reasoning to dissect multi-step queries and deliver structured, deeper insights.
* Tencent CodeBuddy integrates Kimi K2 Thinking as its underlying model
* Gemini CLI is an open-source agentic coding assistant that works from your terminal, giving it access to your local filesystem, development tools, and cloud services. This lets you delegate complex workflows—from building web features to creating marketing materials—through high-level instructions while the agent autonomously plans and executes multiple steps.
* OpenClaw is a configurable agentic framework that runs on a local computer or in a virtual machine in the cloud. Users can build agents to browse and write to their local file systems or operate within predefined sandboxes. They can also give agents permission to use cloud services like email, calendar, productivity applications, speech-to-text and text-to-speech applications, and virtually any service that responds to an API. Agents can use coding tools like Claude Code, interact on social networks, scrape websites, and spend money on users’ behalfs.
* Wikimedia foundation also offers an open Kaggle dataset for noncommercial AI training.
> Have 50 conversations a day with AI
High-frequency usage forces you to find "micro-use cases" you would otherwise ignore. It moves you past the "novelty" phase into the "utility" phase where you instinctively know which tasks the AI can handle.
Don't save AI for big projects. Use it to fix grammar in a text, explain a joke, plan a menu, or debug a single line of code.
* Firebase Studio uses a fork of Code OSS (open-source VS Code). You get familiar editor features (extensions, debugging, terminal, Git integration), but everything — including emulators, previews, and AI assistance — lives in the cloud.
* Odd pack sizes make comparison shopping a hassle. I vibe-coded a Protein Cost Calculator with Firebase Studio to handle non-standard pack sizes and compare the actual protein cost (₹/g) across various brands and food groups. The tool lists one representative packaged high protein item per food category with pre-filled pack size, price & protein. These fields are fully editable so that you input local/store prices. It auto-calculates protein price per gram and the Unit Sale Price (USP).
* Ask ChatGPT a complex question and you'll get a confident, well-reasoned answer. Then type, "Are you sure?" Watch it completely reverse its position.
This isn't a quirky bug. A 2025 study found GPT, Claude, and Gemini flip their answers ~60% of the time when users push back. Not even with evidence, just doubt.
We trained AI this way. RLHF rewards agreement over accuracy. Human evaluators consistently rate agreeable answers higher than correct ones.
We built the world's most expensive yes-men and deployed them where we need pushback the most.
What you need is for the model to push back when it doesn't have enough context. It won't unless you tell it to. Here's the irony: once you instruct it to challenge your assumptions and refuse to answer without sufficient context, it will, because pushing back becomes what you asked for.
* "Industrial capitalism has a marked predilection for the creation and exploitation of what Marx called surplus value, value beyond the value of the labour required to produce commodities, and floats into the mysterious form of profit for the owners and managers of capital.
Today, human beings are the new object and horizon of capitalist extraction. The new object is sociality itself. It is our friendships, our love lives, our family connections, our classmates, our children, our fellow workers, our neighbours, as well as our digital lives, our political allies, our food and drug suppliers, and more. This is profiling on steroids. It mines our deepest affinities and our most ephemeral social ties. It is a new feat of creative destruction, which renders ideas such as privacy, intimacy and trust obsolete. It makes all our sociality a resource that can be mined without permission or limit. And in this unbounded mining of sociality, our stories are the access code. There are three recent changes in the market for selves, which together drive the mineable self.
Stories, technology platforms and digital markets are transforming humans into commodities" - Arjun Appadurai, Emeritus Professor of Media Studies at New York University
* After World War II, Nestlé launched Nescafé in Japan expecting fast growth. But for decades, coffee barely moved. Japan was a tea nation to its core. From 1945 to 1975, Nestlé couldn't crack Japan's coffee market. Nestlé hired French psychoanalyst Clotaire Rapaille in 1975.
His advice was radical: Stop trying to convert tea-drinking adults. Instead, give children their first positive experience with coffee flavor. Not through drinks, but through candy. Nestlé won Japan by understanding one principle: People don't buy what tastes good, they buy what feels familiar. And familiarity starts in childhood.
* In 1492, there were around 18 million Native Americans, but by 1900, the population dropped to about 350,000.
* Established in 1984, Vellore Institute of Technology draws 1 lakh students from all Indian states and 75 countries across the globe.
* India's food delivery market, estimated at $9 billion in FY2025, could reach $25 billion by FY30, according to brokerage firm Jefferies.
* There’s a $39 billion bounty headed to ... cities in India, which by conservative estimates are home to more than a third of the population and contribute up to 70% of GDP...funding and political decentralization since the mid-90s have helped Kerala urbanize close to 70% of the population - Bloomberg
* "I need only three things in life - books, books and books" - Tolstoy
* "Don’t ask a barber whether you need a haircut." - Warren Buffett
Comments
Post a Comment