<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Yatit Thakker on Medium]]></title>
        <description><![CDATA[Stories by Yatit Thakker on Medium]]></description>
        <link>https://medium.com/@ythakker?source=rss-b29fa2dd5e2d------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*UWi4GlZrPIu0lj3YEsmdgA.jpeg</url>
            <title>Stories by Yatit Thakker on Medium</title>
            <link>https://medium.com/@ythakker?source=rss-b29fa2dd5e2d------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 15:32:01 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ythakker/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Consistent Agentic AI Environments with OpenClaw and Pixi]]></title>
            <link>https://ythakker.medium.com/consistent-agentic-ai-environments-with-openclaw-and-pixi-fdfad93a5332?source=rss-b29fa2dd5e2d------2</link>
            <guid isPermaLink="false">https://medium.com/p/fdfad93a5332</guid>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[openclaw]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[docker]]></category>
            <dc:creator><![CDATA[Yatit Thakker]]></dc:creator>
            <pubDate>Sun, 08 Mar 2026 07:15:11 GMT</pubDate>
            <atom:updated>2026-03-08T07:15:11.277Z</atom:updated>
            <content:encoded><![CDATA[<p><em>When vibe coding becomes agentic engineering</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*5qVHrV1nMwUcyMaw" /><figcaption>Photo by <a href="https://unsplash.com/@enchantedtools?utm_source=medium&amp;utm_medium=referral">Enchanted Tools</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Like many software developers and agentic AI practitioners recently, I’ve been experimenting quite a bit with OpenClaw (aka ClawdBot aka MoltBot), the agentic AI framework that is taking the open-source world by storm. Unlike OpenAI, OpenClaw thrives and benefits from an open source community, and despite security risks from its early days, there have been massive improvements recently in installing it with saner defaults that prevent foot-guns from going off. Of course, I have just been running my claw on a separate Linux machine with its own workspace to see what the little lobster can do, and despite several technical challenges in setting it up, I’ve been very impressed.</p><p>The biggest challenge I’ve encountered in securing it (and continue to encounter) is having a consistent, reproducible sandbox environment for its agents to execute code within a limited workspace. Sandboxing is optional, and one might argue that I don’t need it with my Linux-based setup if I don’t give it super-user access on a machine that doesn’t host any personal information. I’ve even locked away my API keys through proxying and rate limiting. Yet, something about several different agents sharing the same workspace rubs me the wrong way because it breaks separation of context and concerns.</p><h3>Why Sandbox OpenClaw?</h3><p>In short, <a href="https://docs.openclaw.ai/gateway/sandboxing#minimal-enable-example">sandboxing an OpenClaw instance</a> allows the agent to execute code and run filesystem level tools within a containerized environment that you control. Other agent harnesses like Codex come with a sandbox enabled by default, scoped to each project. OpenClaw’s sandboxing docs generally lean toward sandboxing with Docker, which provides more flexibility at the cost of some overhead in learning or setting up a new tool. However, since I have my Claw running on a Linux instance already with Docker installed, the overhead was quite small and the benefits quite significant. Namely:</p><ol><li><strong>Context Management</strong> — If you are using OpenClaw to spawn several specialized agents, you can more tightly control the context by only giving these agents access to the files that they need through the Docker <a href="https://docs.docker.com/engine/storage/bind-mounts/">bind mounts</a> feature. Unlike the default sandboxing approach that Codex takes about one agent per GitHub project, you can link a DevOps agent to a few different GitHub repositories where that agent’s DevOps expertise is needed. Not only that, you can give it access to only specific files within these repositories (say the docker-compose.yml files) to more finely manage its context within a file system.</li><li><strong>Smaller Costs of Failure</strong> — A sandbox limits the scope of the mistake because the agent won’t delete your encryption keys or security keys by mistake. You can control which files are read-only and which files are writable by default. This prevents your AI agent from nuking your machine and you having to reinstall your OS. Although that risk is small with today’s highly capable models, it’s <a href="https://www.pcmag.com/news/meta-security-researchers-openclaw-ai-agent-accidentally-deleted-her-emails">not unheard of</a>.</li><li><strong>Consistent Agentic Harness</strong> — The neat part about having sandboxes built on containerized Docker images is that you can pre-install the tools and skills that your agent needs as it evolves over time. Use a specific Python version? Build an image with the libraries pre-installed? Use a specific Node.js version for your <a href="https://agentskills.io/home">agent skills</a>? Pre-install that also! Give your agent all the scripting tools it needs to call the APIs and MCP servers that it wants to call without it touching your main file system. This is the most non-obvious benefit and what I will expand on more in this article.</li></ol><h3>Beyond Homebrew — Custom Container Images</h3><p>OpenClaw is tightly integrated with the <a href="https://brew.sh/">Homebrew</a> package manager. So tightly integrated, in fact, that I didn’t even realize that Homebrew is very mature on Linux now until I started using it with OpenClaw. It might as well replace snap and the other fragmented package managers in the Linux ecosystem. Speaking of fragmented package manager ecosystems, Python and JavaScript / TypeScript have this problem also. Node.js comes pre-installed with NPM, which gets the job done but the problem is much more pronounced with Python, which has (among others) pip, conda, uv, poetry, and pixi.</p><p>I tend to write data-heavy scripts and backends in Python, whereas for frontend and web-server code I almost exclusively use TypeScript. Out of all of the package managers discussed above, Pixi is the only one that fit my use case. Many <a href="https://jacobtomlinson.dev/posts/2025/python-package-managers-uv-vs-pixi/">AI Engineers tend to agree</a>. More specifically, since Pixi uses the conda forge backend for its packages, you can even use it as a replacement NVM (Node version manager), allowing you to create a reproducible, isolated environment for both Node.js and Python at at the same time.</p><h3>The Pixi, Node, and Python Dockerfile for Your Agentic AI Environment</h3><pre>FROM ghcr.io/prefix-dev/pixi:latest<br><br>RUN apt update &amp;&amp; apt install -y bash ca-certificates curl git jq ripgrep ca-certificates golang-go rustc cargo coreutils unzip pkg-config libasound2-dev build-essential file<br>RUN useradd --create-home --shell /bin/bash sandbox<br>USER sandbox<br><br>WORKDIR /home/sandbox/env<br><br>RUN pixi init<br>RUN pixi add &quot;python=3.14&quot; pip ffmpeg requests pyyaml ffmpeg-python pillow schedule click &quot;nodejs=24.*&quot;<br>RUN pixi add --pypi ollama<br><br># Create the shell-hook bash script to activate the environment<br>RUN pixi shell-hook &gt; /home/sandbox/shell-hook.sh<br><br># extend the shell-hook script to run the command passed to the container<br>RUN echo &#39;exec &quot;$@&quot;&#39; &gt;&gt; /home/sandbox/shell-hook.sh<br><br>WORKDIR /home/sandbox<br><br># set the entrypoint to the shell-hook script (activate the environment and run the command)<br># no more pixi needed in the prod container<br>ENTRYPOINT [&quot;/bin/bash&quot;, &quot;/home/sandbox/shell-hook.sh&quot;]<br>CMD [&quot;sleep&quot;, &quot;infinity&quot;]</pre><p>There are a few things worth highlighting in the Dockerfile above:</p><ol><li>The above Dockerfile is a combination of the OpenClaw sandbox image and a Pixi docker image in one.</li><li>Pixi’s default image is built on Ubuntu, so using apt to set up a few basic utilities, followed by using Pixi to install Python and Node.js gives your AI agents access to the two most popular languages in their LLM’s training data.</li><li>Using Pixi as your package manager doesn’t preclude you from using pip or uv. It’s more of a replacement for conda environments, which tend to be clunky, slow, and difficult to reproduce in a DevOps setting.</li><li>Pixi has other features that make it compatible with the rest of the package managers, and you can even install “global” npm packages and pip packages within its environment. The pip package installation is demoed by the line RUN pixi add —pypi ollama, which is useful for calling a separate machine running ollama, even be the host machine if it is broadcasting its port.</li><li>The shell-hook.sh script is a very neat feature of pixi that is especially useful for agentic AI environments because it lets your AI agent drop into the pixi shell without running the command to activate the virtual environment. So, your AI agent doesn’t need to know about Pixi’s environment (unless you want it to). It will just notice that there’s npm and pip already installed, enabling it to install the packages it needs without blowing up your host machine.</li><li>As an additional security precaution, the Docker container will run the agentic harness as a non-root user, preventing installation of apt or Homebrew packages inside the container. This layer of hardening not only makes the agentic environment more secure, it also makes the agent’s work easier to verify through package install manifests like pip’s requirements.txt file or npm’s package.json.</li><li>Overall, this Dockerfile allows you to customize build images with the packages and libraries your agent needs, while making it compatible with two of the most popular package managers that your AI agent is already trained on, namely pipand npm.</li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fdfad93a5332" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI’s Biggest Weakness — Correlation vs Causation]]></title>
            <link>https://ythakker.medium.com/ais-biggest-weakness-correlation-vs-causation-447257328a2f?source=rss-b29fa2dd5e2d------2</link>
            <guid isPermaLink="false">https://medium.com/p/447257328a2f</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[statistics]]></category>
            <dc:creator><![CDATA[Yatit Thakker]]></dc:creator>
            <pubDate>Sat, 24 Jan 2026 20:53:39 GMT</pubDate>
            <atom:updated>2026-01-24T20:53:39.313Z</atom:updated>
            <content:encoded><![CDATA[<h3>AI’s Biggest Weakness — Correlation vs Causation</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*jMGvlaiZWGBN8t5t" /><figcaption>Photo by <a href="https://unsplash.com/@kageyama?utm_source=medium&amp;utm_medium=referral">Yuichi Kageyama</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>The <a href="https://en.wikipedia.org/wiki/Butterfly_effect">butterfly effect</a> is a very effective way of describing the modern world, which is increasingly powered by data. Small changes in our interactions with this data have far-reaching consequences and effects that we cannot predict in the moment that the decision is made. It’s a very seductive promise, then, that AI can make these predictions for us. By using advanced statistics, AI supposedly becomes this expert predictive machine that can draw conclusions from data that correlates with existing patterns. With ChatGPT, we can now make these predictions while our coffee brews. Easy, right? Not so fast.</p><h3>Correlation does not effect Causation</h3><p>Every breakthrough technology has its roots in a combination of a foundational branch of mathematics, combining in new and innovative ways with a foundational branch of the sciences, with some cutting edge engineering doing a lot of heavy lifting in bridging these gaps. With AI, this foundational branch of mathematics is Statistics, specifically predictive statistics. When combined with the foundational branch of Computer Science, lifted up by two decades of engineering investment in Graphical Processing Units (GPUs) through video games and bitcoin mining, we arrived at modern AI, which can run on a pro consumer grade laptop today.</p><p>Thanks for reading! Subscribe for free to receive new posts and support my work.</p><p>But really, most generative AI models today are word predictors on steroids, like the auto-correct ability on your phone. They can correlate the average pattern of words from their training data and predictive statistics. However, major changes are by their own nature anomalies. Major changes are cause by people. People with power. People with influence. People who assert their freedom.</p><p>Even though AI is an amazing machine for prediction, it cannot replicate the unique judgment that people have. The ability to make decisions, to see a path forward, and to believe in a better future are uniquely human characteristics. This is why, even the most advanced AI-driven processes that are being developed today fall flat without a decision making person who is looped into the entire process that is being driven by AI. This is a very crucial piece of the workflow, aptly named “human-in-the-loop”. Autonomous AI processes are not transparent, lead to poor decision-making, and will always be prone to hallucinations. This doesn’t even address the tendency toward average or sub-par outcomes built into the nature of predictive statistics.</p><h3>Are People Better Decision-Makers?</h3><p>This begs the question: by having a person in the loop making decisions, does the quality of decision-making get better? It’s debatable but I lean towards yes. It’s not that people individually are better decision-makers (let’s not flatter ourselves, we are not). However, with carefully structured output from AI, evaluated with the help of a sounding board or two, this collaboration between people and AI can produce much better outcomes than not.</p><p>One of the best parts about AI is that it documents its thought process very thoroughly so that when more information becomes available later on about what went into a specific decision-making process, it’s much easier to review the output from an AI agent rather than a conversation from a meeting where some people may have conflicting opinions about what steps led to a particular decision being made. AI is self-documenting, and effortlessly so.</p><h3>Biggest Source of Causation</h3><p>Ultimately, people are the biggest source of causation, and AI will not change that anytime soon. Like almost every technology that has come before, AI provides a source of leverage to amplify our efforts. It’s the execution layer, at least a small but very significant part of it. When that execution layer gets squeezed, the quality of decisions gets worse in exchange for a greater speed of decision making. The problem is that the consequences of bad decision-making are often hard to see at a particular moment in time. Even not making a decision is still a decision to not decide. My friends and I like to tell each other “fuck around and find out”. As someone who has been working with AI for 7 years and can recognize the influence it’s had on society, we are at the “find out” stage.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=447257328a2f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Rediscovering Mathematics]]></title>
            <link>https://ythakker.medium.com/rediscovering-mathematics-4c4c9f81fe01?source=rss-b29fa2dd5e2d------2</link>
            <guid isPermaLink="false">https://medium.com/p/4c4c9f81fe01</guid>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[haskell]]></category>
            <category><![CDATA[mathematics]]></category>
            <dc:creator><![CDATA[Yatit Thakker]]></dc:creator>
            <pubDate>Sun, 02 Jan 2022 23:54:33 GMT</pubDate>
            <atom:updated>2022-01-02T23:54:33.467Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="A surfboard standing vertically on a beach with emerald-green water" src="https://cdn-images-1.medium.com/max/1024/0*ZgHYfGMupdqn8lE5" /><figcaption>Photo by <a href="https://unsplash.com/@nbb_photos?utm_source=medium&amp;utm_medium=referral">Lacie Slezak</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>My Calculus teacher in high school introduced a debate in the mathematical community—is mathematics discovered or invented? Discovery implies finding something that already exists. Inventing implies creating something that wasn’t there before. The title gives away the answer, of course. Math is a universal language. We only invent new ways of expressing it.</p><h3>Programming Meets Mathematics</h3><p>Having taken two years of math in college as part of my engineering degree, then writing software code for almost 10 years, I was not expecting to discover parts of math that would challenge me with new ways of thinking. This changed when I decided to dive deeper into functional programming and learn <a href="https://en.wikipedia.org/wiki/Haskell_(programming_language)">Haskell</a>.</p><p>Although Haskell was invented over thirty years ago, its stable release was not until 2010. That’s only a year after node.js, now a ubiquitous browser runtime, was released. So, it is an older language finding new life in several areas, like finance and medical tech, where security and reliability are critical.</p><p>Although I’m very much a beginner when it comes to the ins and outs of the language, compiling a few programs and writing test cases now helped me understand what makes Haskell special and why I’ll continue learning it. Here’s the best way I could write about it and how it relates to the famous debate about whether mathematics is discovered or invented.</p><h3>Why Haskell is Unique</h3><p>Suppose there was a gray area between discovery and invention. The invention side of this gray area would have things about mathematics that are already known or obvious to prove to someone with formal study. Then, through a scientific process of trial, error, and experimentation, suppose a few computer scientists took these concepts about mathematics and designed a <a href="https://en.wikipedia.org/wiki/Turing_completeness">turing-complete</a> programming language using them. You’d get Haskell. In other words, Haskell is almost like <strong>a fabric that connects the world of discovery to the world of invention</strong>.</p><p>Functions like `map`, `filter`, and `reduce` originate from the world of functional programming. Although I make use of these and other pure functions, Haskell rips off the training wheels. There is no explicit keyword to loop through a list of items. Everything is functionally recursive. It’s a stumbling block to the uninitiated. I count myself among them. But the water is fine. I’m ready to surf the currents of discovery and revisit what I missed last time.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4c4c9f81fe01" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Trust Brokers in Web 3.0]]></title>
            <link>https://medium.com/swlh/trust-brokers-in-web-3-0-13239a4c7e20?source=rss-b29fa2dd5e2d------2</link>
            <guid isPermaLink="false">https://medium.com/p/13239a4c7e20</guid>
            <category><![CDATA[blockchain]]></category>
            <category><![CDATA[smart-contracts]]></category>
            <dc:creator><![CDATA[Yatit Thakker]]></dc:creator>
            <pubDate>Tue, 21 Dec 2021 20:42:57 GMT</pubDate>
            <atom:updated>2021-12-21T20:42:57.933Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*PJATuQ_LE2xxzJRB" /><figcaption>Photo by <a href="https://unsplash.com/@dylandgillis?utm_source=medium&amp;utm_medium=referral">Dylan Gillis</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>There is a lot of hype around Blockchain technology being the introduction of Web 3.0. As the name implies, Web 3.0 is assumed to be better than Web 2.0 in every way. While Web 3.0 is revolutionary from the perspective of <strong>financing</strong> and <strong>governance</strong>, it falls short in other areas like reliability and accessibility. But first, let’s review what is Web 2.0 and why it has been revolutionary.</p><h3>Web 2.0 Introduced Web APIs</h3><p>The biggest benefit that Web 2.0 introduced is data sharing. Before Web 2.0, data existed in proprietary silos. Accessing it required large amounts of capital. Think about early GPS devices like TomTom, which built an entire business around navigating and directions so that we didn’t have to print directions from Yahoo Maps or MapQuest.</p><p>Then Web 2.0 came along, and together with the Google Maps API, web developers could consume mapping and navigation data by only paying per “request” rather than paying huge licensing fees for accessing all of the data. This business model was prolifically applied across almost all industries, leading to game-changing products like Amazon Web Services (AWS), Google Image search, Machine Learning as-a-service and many others. It was Web 2.0 that gave birth to Software as a Service, Hardware as a Service, ride-sharing as a service, and most controversially, attention as a service.</p><p>This last product — attention as a service — also proved to be the most controversial. Social media platforms like Facebook, Snapchat, TikTok, and YouTube capitalized on people’s attention by designing interfaces that simulated addiction in order to keep people’s attention. In the meantime, the mined as much data as they could using this new currency of “attention”, leading to controversial practices like tracking your browsing history across the web, selling your data to third party advertisers, and much more.</p><p>Since they owned your data, they could (and did) sell it to the highest bidder. These advertising giants like Alphabet (Google’s parent) and Meta (Facebook’s parent) built empires worth trillions of dollars on shady practices that involved tracking people across different websites, finding out your interests and showing you ads that were eerily similar to what you were looking for. In fact, some of their advertising algorithms were so advanced that you might find an ad for something you were just talking to your friend about. It was almost as if the algorithms knew you better than you knew yourself.</p><p>Although the web became easier to use during this time, it also became much more centralized. This process only accelerated during the COVID-19 pandemic in 2020 such that within the next two years, the top five technology companies were worth more than 6 trillion dollars. That’s 6000 billion dollars. Meanwhile, many Instagram users were susceptible to addiction, low self-esteem, and other issues that had never been studied before.</p><h3><strong>Blockchain and Web 3.0</strong></h3><p>While large tech companies made hundreds of billions of dollars on centralizing their data, a new movement was emerging. <strong>Decentralization</strong>. The first aspect of society that was ripe for decentralization was finance. In a protest against the establishment that rescued failing banks with hundreds of billions of dollars of taxpayer dollars during the Great Recession, an anonymous programmer named Satoshi Nakamoto created Bitcoin with the help of other collaborators around the world. Unlike fiat currencies that inflated over time, Bitcoin was made to be finite. It was heralded as the people’s currency.</p><p>Over the next 10 years, the laws of supply and demand took over. As Bitcoin became harder and harder to mine, its price kept going up. Graphics cards, and later specialized chips, were created for Bitcoin mining. Yet, finance was only one aspect of the underlying blockchain technology. The Blockchain could decentralize not only finance, but computation itself. What if decentralized “miners” could do the work of computers? Of web servers that hosted websites and data? What if it replaced corrupt financial institutions in developing countries?</p><p>These were powerful ideas—far ahead of their time—accelerated by a global pandemic seeding desperation across the globe. Yet, there remain major hurdles to overcome. The first was accessibility. Blockchain programming and participation required large amounts of money, time, and risk compared to the Cloud. Cloud computing was cheap and mature.</p><p>The other problem was that there wasn’t a trustworthy broker. With Bitcoin, the <strong>exchange</strong> is the broker. Over time, Coinbase proved itself to be one of the more capable exchanges for buying Bitcoin and it flourished by building trust with regulators, its users, and even public investors through its IPO. This earned it the nickname of being the general store for Cryptocurrency.</p><p>With smart contracts, the Blockchain is the broker. Specifically, the “whales” of the Blockchain have to act as trustworthy brokers to build trust around their “trustless” system. How can they do that? By stewarding the Blockchain like a people’s app store. Executing smart contracts ought to be affordable, programmable, and reviewed for quality assurance. The right incentives ought to be in place for network participation among everyone involved in the smart contract review and publication process.</p><p>There’s still a ways to go before a blockchain can compete with major app publishing platforms, but having the right processes in place early can only be good for scalability. The blockchain that ultimately succeeds will not only need to have capable technology, but also very capable trust brokers that differentiate it from other platforms. Right now, Ethereum is in the lead, but who can really tell where Crypto will go?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=13239a4c7e20" width="1" height="1" alt=""><hr><p><a href="https://medium.com/swlh/trust-brokers-in-web-3-0-13239a4c7e20">Trust Brokers in Web 3.0</a> was originally published in <a href="https://medium.com/swlh">The Startup</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Triple Bottom Line: is the Next Wave of Capitalism Here?]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.datadriveninvestor.com/the-triple-bottom-line-is-the-next-wave-of-capitalism-here-f5485706b085?source=rss-b29fa2dd5e2d------2"><img src="https://cdn-images-1.medium.com/max/2600/0*VR-VM1Q6uCRDncxu" width="6016"></a></p><p class="medium-feed-snippet">In order to become more sustainable&#x2014;sustainably&#x2014;today&#x2019;s investors and environmentalists use a &#x201C;triple bottom line&#x201D; (TBL) approach to&#x2026;</p><p class="medium-feed-link"><a href="https://medium.datadriveninvestor.com/the-triple-bottom-line-is-the-next-wave-of-capitalism-here-f5485706b085?source=rss-b29fa2dd5e2d------2">Continue reading on DataDrivenInvestor »</a></p></div>]]></description>
            <link>https://medium.datadriveninvestor.com/the-triple-bottom-line-is-the-next-wave-of-capitalism-here-f5485706b085?source=rss-b29fa2dd5e2d------2</link>
            <guid isPermaLink="false">https://medium.com/p/f5485706b085</guid>
            <category><![CDATA[trends]]></category>
            <category><![CDATA[environment]]></category>
            <category><![CDATA[capitalism]]></category>
            <category><![CDATA[investing]]></category>
            <category><![CDATA[sustainability]]></category>
            <dc:creator><![CDATA[Yatit Thakker]]></dc:creator>
            <pubDate>Sun, 17 Oct 2021 19:28:12 GMT</pubDate>
            <atom:updated>2021-10-19T01:23:04.940Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[What 5 years of tech work in Silicon Valley has Taught Me]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://ythakker.medium.com/what-5-years-of-tech-work-in-silicon-valley-has-taught-me-beee771bb8f0?source=rss-b29fa2dd5e2d------2"><img src="https://cdn-images-1.medium.com/max/2600/0*T1Lel5BOR576iTm1" width="6000"></a></p><p class="medium-feed-snippet">It was a typical arrival to an atypical place. A few exits away from Sand Hill Road, in 2015 I was dropped off by my second cousin to the&#x2026;</p><p class="medium-feed-link"><a href="https://ythakker.medium.com/what-5-years-of-tech-work-in-silicon-valley-has-taught-me-beee771bb8f0?source=rss-b29fa2dd5e2d------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://ythakker.medium.com/what-5-years-of-tech-work-in-silicon-valley-has-taught-me-beee771bb8f0?source=rss-b29fa2dd5e2d------2</link>
            <guid isPermaLink="false">https://medium.com/p/beee771bb8f0</guid>
            <category><![CDATA[work-life-balance]]></category>
            <category><![CDATA[silicon-valley]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Yatit Thakker]]></dc:creator>
            <pubDate>Sun, 14 Feb 2021 21:29:51 GMT</pubDate>
            <atom:updated>2021-02-14T21:29:51.753Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[A Tale of Two Extremities]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/teachers-on-fire/a-tale-of-two-extremities-e9403a883964?source=rss-b29fa2dd5e2d------2"><img src="https://cdn-images-1.medium.com/max/2600/0*dzFSv0mFfxc3BZje" width="3236"></a></p><p class="medium-feed-snippet">It was the best of times. It was the worst of times. Some called it a modern Renaissance. Others called it a new Gilded Age. There had&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/teachers-on-fire/a-tale-of-two-extremities-e9403a883964?source=rss-b29fa2dd5e2d------2">Continue reading on Teachers on Fire Magazine »</a></p></div>]]></description>
            <link>https://medium.com/teachers-on-fire/a-tale-of-two-extremities-e9403a883964?source=rss-b29fa2dd5e2d------2</link>
            <guid isPermaLink="false">https://medium.com/p/e9403a883964</guid>
            <category><![CDATA[philosophy]]></category>
            <category><![CDATA[society]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[2020]]></category>
            <category><![CDATA[psychology]]></category>
            <dc:creator><![CDATA[Yatit Thakker]]></dc:creator>
            <pubDate>Fri, 27 Nov 2020 23:50:38 GMT</pubDate>
            <atom:updated>2020-11-27T23:50:38.930Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[The Two Sides of Google vs. US DOJ]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://ythakker.medium.com/the-two-sides-of-google-vs-us-doj-8f710b2b8854?source=rss-b29fa2dd5e2d------2"><img src="https://cdn-images-1.medium.com/max/2600/0*c9OQBknpA_TJb0KK" width="5892"></a></p><p class="medium-feed-snippet">The US Department of Justice led by Attorney General Bill Barr on Oct. 20, 2020 filed an antitrust lawsuit against Google accusing it of&#x2026;</p><p class="medium-feed-link"><a href="https://ythakker.medium.com/the-two-sides-of-google-vs-us-doj-8f710b2b8854?source=rss-b29fa2dd5e2d------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://ythakker.medium.com/the-two-sides-of-google-vs-us-doj-8f710b2b8854?source=rss-b29fa2dd5e2d------2</link>
            <guid isPermaLink="false">https://medium.com/p/8f710b2b8854</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[law]]></category>
            <category><![CDATA[technews]]></category>
            <category><![CDATA[internet]]></category>
            <dc:creator><![CDATA[Yatit Thakker]]></dc:creator>
            <pubDate>Thu, 22 Oct 2020 08:05:04 GMT</pubDate>
            <atom:updated>2020-10-22T15:14:39.121Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[How to solve the fresh water crisis with clean energy desalination]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/predict/how-to-solve-the-fresh-water-crisis-with-clean-energy-desalination-3f4bc930e768?source=rss-b29fa2dd5e2d------2"><img src="https://cdn-images-1.medium.com/max/2600/0*B38o03-tcG7BtST1" width="5464"></a></p><p class="medium-feed-snippet">Clean energy is the first step in achieving carbon neutrality. This means energy from wind, solar, hydroelectric, and even nuclear power&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/predict/how-to-solve-the-fresh-water-crisis-with-clean-energy-desalination-3f4bc930e768?source=rss-b29fa2dd5e2d------2">Continue reading on Predict »</a></p></div>]]></description>
            <link>https://medium.com/predict/how-to-solve-the-fresh-water-crisis-with-clean-energy-desalination-3f4bc930e768?source=rss-b29fa2dd5e2d------2</link>
            <guid isPermaLink="false">https://medium.com/p/3f4bc930e768</guid>
            <category><![CDATA[climate-change]]></category>
            <category><![CDATA[water]]></category>
            <category><![CDATA[environmental-issues]]></category>
            <category><![CDATA[sustainability]]></category>
            <category><![CDATA[green-energy]]></category>
            <dc:creator><![CDATA[Yatit Thakker]]></dc:creator>
            <pubDate>Fri, 02 Oct 2020 23:57:58 GMT</pubDate>
            <atom:updated>2020-10-03T22:11:24.702Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Microsoft and Apple will not hit Net Zero Emissions by 2030]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/predict/microsoft-and-apple-will-not-hit-net-zero-emissions-by-2030-e8643ad4fe45?source=rss-b29fa2dd5e2d------2"><img src="https://cdn-images-1.medium.com/max/2600/0*1PEiHimv5EY4vman" width="4000"></a></p><p class="medium-feed-snippet">It&#x2019;s that thing about software&#x2026; is it keeps running and taking up energy</p><p class="medium-feed-link"><a href="https://medium.com/predict/microsoft-and-apple-will-not-hit-net-zero-emissions-by-2030-e8643ad4fe45?source=rss-b29fa2dd5e2d------2">Continue reading on Predict »</a></p></div>]]></description>
            <link>https://medium.com/predict/microsoft-and-apple-will-not-hit-net-zero-emissions-by-2030-e8643ad4fe45?source=rss-b29fa2dd5e2d------2</link>
            <guid isPermaLink="false">https://medium.com/p/e8643ad4fe45</guid>
            <category><![CDATA[microsoft]]></category>
            <category><![CDATA[sustainability]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[apple]]></category>
            <category><![CDATA[green-energy]]></category>
            <dc:creator><![CDATA[Yatit Thakker]]></dc:creator>
            <pubDate>Wed, 22 Jul 2020 04:00:52 GMT</pubDate>
            <atom:updated>2020-07-26T05:05:14.241Z</atom:updated>
        </item>
    </channel>
</rss>