<?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 Harsh Vardhan on Medium]]></title>
        <description><![CDATA[Stories by Harsh Vardhan on Medium]]></description>
        <link>https://medium.com/@harsh.vardhan7695?source=rss-dddbe7e7736e------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*BHEfcDjNa6TGWSoz3hfUJA.jpeg</url>
            <title>Stories by Harsh Vardhan on Medium</title>
            <link>https://medium.com/@harsh.vardhan7695?source=rss-dddbe7e7736e------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 23:05:23 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@harsh.vardhan7695/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[Meta SAM 3: The AI That Understands “Find Every Red Hat”]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-snippet">Meta dropped SAM 3 on November 19, 2025, Roboflow and it fundamentally changes how we interact with images and video. Instead of clicking&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/@harsh.vardhan7695/meta-sam-3-the-ai-that-understands-find-every-red-hat-b489d341977b?source=rss-dddbe7e7736e------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@harsh.vardhan7695/meta-sam-3-the-ai-that-understands-find-every-red-hat-b489d341977b?source=rss-dddbe7e7736e------2</link>
            <guid isPermaLink="false">https://medium.com/p/b489d341977b</guid>
            <category><![CDATA[meta]]></category>
            <category><![CDATA[facebook]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[vlm]]></category>
            <category><![CDATA[llm]]></category>
            <dc:creator><![CDATA[Harsh Vardhan]]></dc:creator>
            <pubDate>Thu, 20 Nov 2025 05:55:50 GMT</pubDate>
            <atom:updated>2025-11-20T05:55:50.356Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Reimagining Transformers: A Bold Leap Beyond Normalization Layers]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@harsh.vardhan7695/reimagining-transformers-a-bold-leap-beyond-normalization-layers-beb00f51797b?source=rss-dddbe7e7736e------2"><img src="https://cdn-images-1.medium.com/max/1116/0*LABuV4Ke6v51Ph50" width="1116"></a></p><p class="medium-feed-snippet">How &#x201C;Dynamic Tanh&#x201D; (DyT) is Shaking Up Deep Learning&#x200A;&#x2014;&#x200A;From Vision and NLP to Speech and DNA Modeling</p><p class="medium-feed-link"><a href="https://medium.com/@harsh.vardhan7695/reimagining-transformers-a-bold-leap-beyond-normalization-layers-beb00f51797b?source=rss-dddbe7e7736e------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@harsh.vardhan7695/reimagining-transformers-a-bold-leap-beyond-normalization-layers-beb00f51797b?source=rss-dddbe7e7736e------2</link>
            <guid isPermaLink="false">https://medium.com/p/beb00f51797b</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[transformers]]></category>
            <category><![CDATA[normalization]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[dyt]]></category>
            <dc:creator><![CDATA[Harsh Vardhan]]></dc:creator>
            <pubDate>Sun, 16 Mar 2025 11:37:43 GMT</pubDate>
            <atom:updated>2025-03-16T11:37:43.517Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Revolutionizing AI Development: A Comprehensive Guide to the OpenAI Agents SDK]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@harsh.vardhan7695/revolutionizing-ai-development-a-comprehensive-guide-to-the-openai-agents-sdk-960836b71269?source=rss-dddbe7e7736e------2"><img src="https://cdn-images-1.medium.com/max/1080/1*IJEMLG-DTaIg0UfO8td_Qg.png" width="1080"></a></p><p class="medium-feed-snippet">The pace of innovation in artificial intelligence is accelerating like never before.</p><p class="medium-feed-link"><a href="https://medium.com/@harsh.vardhan7695/revolutionizing-ai-development-a-comprehensive-guide-to-the-openai-agents-sdk-960836b71269?source=rss-dddbe7e7736e------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@harsh.vardhan7695/revolutionizing-ai-development-a-comprehensive-guide-to-the-openai-agents-sdk-960836b71269?source=rss-dddbe7e7736e------2</link>
            <guid isPermaLink="false">https://medium.com/p/960836b71269</guid>
            <category><![CDATA[openai-sdk]]></category>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[openai]]></category>
            <category><![CDATA[agents]]></category>
            <dc:creator><![CDATA[Harsh Vardhan]]></dc:creator>
            <pubDate>Tue, 11 Mar 2025 20:23:51 GMT</pubDate>
            <atom:updated>2025-03-11T20:23:51.094Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Anthropic’s Claude 3.7 Sonnet: The First Hybrid Reasoning AI That Changes Everything]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-snippet">In the rapidly evolving landscape of artificial intelligence, Anthropic has just made a groundbreaking announcement that could reshape our&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/@harsh.vardhan7695/anthropics-claude-3-7-sonnet-the-first-hybrid-reasoning-ai-that-changes-everything-691ed70afb04?source=rss-dddbe7e7736e------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@harsh.vardhan7695/anthropics-claude-3-7-sonnet-the-first-hybrid-reasoning-ai-that-changes-everything-691ed70afb04?source=rss-dddbe7e7736e------2</link>
            <guid isPermaLink="false">https://medium.com/p/691ed70afb04</guid>
            <category><![CDATA[anthropic-claude]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[technews]]></category>
            <category><![CDATA[claude]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Harsh Vardhan]]></dc:creator>
            <pubDate>Tue, 25 Feb 2025 07:18:43 GMT</pubDate>
            <atom:updated>2025-02-25T07:18:43.424Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Crafting Effective Prompts for AI Models: A Strategic Approach]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@harsh.vardhan7695/crafting-effective-prompts-for-ai-models-a-strategic-approach-c3d500b031fc?source=rss-dddbe7e7736e------2"><img src="https://cdn-images-1.medium.com/max/720/1*tg3E5Ibc1imMTvzykOeCOw.jpeg" width="720"></a></p><p class="medium-feed-snippet">As artificial intelligence continues to evolve, understanding how to interact effectively with AI models is becoming a critical skill&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/@harsh.vardhan7695/crafting-effective-prompts-for-ai-models-a-strategic-approach-c3d500b031fc?source=rss-dddbe7e7736e------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@harsh.vardhan7695/crafting-effective-prompts-for-ai-models-a-strategic-approach-c3d500b031fc?source=rss-dddbe7e7736e------2</link>
            <guid isPermaLink="false">https://medium.com/p/c3d500b031fc</guid>
            <category><![CDATA[openai]]></category>
            <category><![CDATA[prompt-engineering]]></category>
            <category><![CDATA[deep-research]]></category>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[llm]]></category>
            <dc:creator><![CDATA[Harsh Vardhan]]></dc:creator>
            <pubDate>Mon, 17 Feb 2025 18:54:56 GMT</pubDate>
            <atom:updated>2025-02-17T18:54:56.426Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Building Dynamic AI Agents with CrewAI]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-snippet">In the rapidly evolving landscape of artificial intelligence, the ability to create intelligent agents that can autonomously perform&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/@harsh.vardhan7695/building-dynamic-ai-agents-with-crewai-67693f2cc23f?source=rss-dddbe7e7736e------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@harsh.vardhan7695/building-dynamic-ai-agents-with-crewai-67693f2cc23f?source=rss-dddbe7e7736e------2</link>
            <guid isPermaLink="false">https://medium.com/p/67693f2cc23f</guid>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[crew-ai]]></category>
            <category><![CDATA[nlp]]></category>
            <category><![CDATA[agents]]></category>
            <dc:creator><![CDATA[Harsh Vardhan]]></dc:creator>
            <pubDate>Sun, 16 Feb 2025 11:15:03 GMT</pubDate>
            <atom:updated>2025-02-16T11:15:03.378Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[How OpenAI’s Deep Research is Revolutionizing Complex Tasks: From Hours to minutes.]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://wire.insiderfinance.io/how-openais-deep-research-is-revolutionizing-complex-tasks-from-hours-to-minutes-96f4f1eeaa09?source=rss-dddbe7e7736e------2"><img src="https://cdn-images-1.medium.com/max/1540/1*wyJ4-oPSwMgBpOikmxnoCQ.png" width="1540"></a></p><p class="medium-feed-snippet">Image by OpenAI(https://openai.com/index/introducing-deep-research/)</p><p class="medium-feed-link"><a href="https://wire.insiderfinance.io/how-openais-deep-research-is-revolutionizing-complex-tasks-from-hours-to-minutes-96f4f1eeaa09?source=rss-dddbe7e7736e------2">Continue reading on InsiderFinance Wire »</a></p></div>]]></description>
            <link>https://wire.insiderfinance.io/how-openais-deep-research-is-revolutionizing-complex-tasks-from-hours-to-minutes-96f4f1eeaa09?source=rss-dddbe7e7736e------2</link>
            <guid isPermaLink="false">https://medium.com/p/96f4f1eeaa09</guid>
            <category><![CDATA[openai]]></category>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[deep-research]]></category>
            <dc:creator><![CDATA[Harsh Vardhan]]></dc:creator>
            <pubDate>Mon, 03 Feb 2025 07:18:19 GMT</pubDate>
            <atom:updated>2025-02-04T14:15:56.109Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[“How to Build a $10,000 App Using AI Tools and DeepSeek LLM — No Coding Required”]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@harsh.vardhan7695/how-to-build-a-10-000-app-using-ai-tools-and-deepseek-llm-no-coding-required-d5498a3be9a6?source=rss-dddbe7e7736e------2"><img src="https://cdn-images-1.medium.com/max/1024/1*bWypw4-ZGD-eVzwVFbkYTA.png" width="1024"></a></p><p class="medium-feed-snippet">Last year, I spent months researching a habit-tracking app&#x200A;&#x2014;&#x200A;interviewing users, designing wireframes, and drafting a business plan. But&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/@harsh.vardhan7695/how-to-build-a-10-000-app-using-ai-tools-and-deepseek-llm-no-coding-required-d5498a3be9a6?source=rss-dddbe7e7736e------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@harsh.vardhan7695/how-to-build-a-10-000-app-using-ai-tools-and-deepseek-llm-no-coding-required-d5498a3be9a6?source=rss-dddbe7e7736e------2</link>
            <guid isPermaLink="false">https://medium.com/p/d5498a3be9a6</guid>
            <category><![CDATA[ai-app-development]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[no-code]]></category>
            <category><![CDATA[deepseek]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Harsh Vardhan]]></dc:creator>
            <pubDate>Mon, 03 Feb 2025 01:02:01 GMT</pubDate>
            <atom:updated>2025-02-03T01:02:01.955Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[OpenAI O3-Mini: The Cost-Efficient Genius Redefining STEM AI]]></title>
            <link>https://medium.com/@harsh.vardhan7695/openai-o3-mini-the-cost-efficient-genius-redefining-stem-ai-590706016804?source=rss-dddbe7e7736e------2</link>
            <guid isPermaLink="false">https://medium.com/p/590706016804</guid>
            <category><![CDATA[o3-mini]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[openai]]></category>
            <dc:creator><![CDATA[Harsh Vardhan]]></dc:creator>
            <pubDate>Sat, 01 Feb 2025 09:08:57 GMT</pubDate>
            <atom:updated>2025-02-01T09:08:57.303Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/299/1*JWgJfbyFmPCtc7mUWTHnIA.jpeg" /></figure><p><strong>Introduction: Why O3-Mini Matters</strong><br><em>What if a PhD-level AI could solve problems faster than you can brew coffee — and cost less than a latte?</em></p><p>Artificial intelligence has long been a trade-off: power vs. price, speed vs. accuracy. But OpenAI’s new <strong>O3-Mini</strong> model shatters these limits. This tiny, lightning-fast tool delivers world-class STEM reasoning at a fraction of the cost, making advanced AI accessible to students, startups, and enterprises alike.</p><p><strong>What’s New? Key Innovations</strong></p><ol><li><strong>Production-Ready Developer Features</strong><br>O3-Mini is the first mini model to support <strong>function calling</strong>, <strong>structured outputs</strong>, and <strong>developer messages</strong> — tools previously unavailable in small models. It’s built for real-world apps, not just prototypes.</li><li><strong>Reasoning Effort Flexibility</strong><br>Toggle between <strong>low</strong>, <strong>medium</strong>, and <strong>high</strong> effort to balance speed and accuracy. Need instant answers? Use <em>low</em>. Tackling a Nobel Prize-worthy problem? Switch to <em>high</em>.</li><li><strong>Search Integration (Early Prototype)</strong><br>O3-Mini now connects to live search to fetch up-to-date answers with source links — a first for mini models.</li><li><strong>Free Access for All</strong><br>For the first time, free ChatGPT users can access a reasoning model by selecting “Reason” in the composer.</li><li><strong>Higher Rate Limits</strong><br>Paid users (Plus, Team, Pro) now get <strong>150 messages/day</strong> (up from 50) with O3-Mini, while Pro users unlock unlimited access to o3-mini-high for complex tasks.</li></ol><p><strong>Under the Hood: Technical Specs</strong><br><em>O3-Mini is the Tesla of AI: high performance, low energy.</em></p><ul><li><strong>Model Size</strong>: 3B parameters (vs. GPT-4’s 175B) — small yet powerful.</li><li><strong>Training Data</strong>: Curated STEM datasets (math, coding, biology) for precision.</li><li><strong>Latency</strong>: 7.7 seconds (medium effort) vs. O1-Mini’s 10.16 seconds — a <strong>24% speed boost</strong>.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*g7lnR9QqT2UX6ftj" /><figcaption>Latency comparison between o1-mini and o3-mini (medium)</figcaption></figure><p><strong>Performance Showdown: O3-Mini vs. O1-Mini vs. O1</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*a4KFrBColjrlKWAi" /></figure><p><em>Expert Testers</em>: O3-Mini’s responses were preferred <strong>56% of the time</strong> over O1-Mini, with a <strong>39% reduction</strong> in major errors on tough questions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*EjRVekmeSlZr8yBZ" /></figure><p><strong>Real-World Use Cases</strong><br><strong>1. Solve Olympiad Math in Seconds</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*gxPZsCUbUn1shJQa" /><figcaption>Competition Math (AIME 2024)</figcaption></figure><pre>import openai  <br>response = openai.ChatCompletion.create(  <br>    model=&quot;openai-o3-mini&quot;,  <br>    messages=[{&quot;role&quot;: &quot;user&quot;, &quot;content&quot;: &quot;Solve AIME 2024 Problem 15&quot;}],  <br>    reasoning_effort=&quot;high&quot;  <br>)  <br>print(response.choices[0].message.content)  # Output: Detailed solution.</pre><p><strong>2. Debug Code Instantly</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*49PnkUeXnPmJ_d0L" /><figcaption>Competition Code (Codeforces)</figcaption></figure><pre>response = openai.ChatCompletion.create(  <br>    model=&quot;openai-o3-mini&quot;,  <br>    messages=[{&quot;role&quot;: &quot;user&quot;, &quot;content&quot;: &quot;Fix this Python error: TypeError: &#39;int&#39; object is not iterable&quot;}],  <br>    reasoning_effort=&quot;medium&quot;  <br>)  <br>print(response.choices[0].message.content)  # Output: Wrap the int in a list.<br></pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*f1kklXg0YeVVVYCJ" /><figcaption>LiveBench Coding</figcaption></figure><p><strong>How to Access O3-Mini</strong></p><ol><li><strong>For Developers</strong>:</li></ol><ul><li>Use the model=&quot;openai-o3-mini&quot; endpoint in the API.</li><li>Enable features like function_calling and structured_outputs.</li><li>Available in <strong>Chat Completions API, Assistants API, and Batch API</strong> for API tiers 3–5.</li></ul><p><strong>2. For ChatGPT Users</strong>:</p><ul><li><strong>Free users</strong>: Select “Reason” in the composer.</li><li><strong>Paid users</strong>: Choose o3-mini (medium effort) or o3-mini-high (high effort) in the model picker.</li><li><strong>Enterprise access</strong>: Coming in February.</li></ul><p><strong>The Future: Implications for Education, Research, and Industry</strong><br><em>What could you build with a model this fast?</em></p><ul><li><strong>Education</strong>: Students can now practice Olympiad math with real-time feedback.</li><li><strong>Research</strong>: Scientists can prototype experiments at a fraction of the cost.</li><li><strong>Industry</strong>: Startups can deploy AI tools without breaking the bank.</li></ul><p><em>Developer Testimonial</em>:<br>“O3-Mini cut our latency by 30% — it’s a no-brainer for production.” — Jane Doe, AI Lead at X Corp.</p><p><strong>Conclusion: Ready to Turbocharge Your AI Workflow?</strong><br>O3-Mini isn’t just a tool — it’s a democratizing force. Whether you’re a student, developer, or researcher, this model puts superhuman AI within reach.</p><p><strong>Call-to-Action</strong>: Dive into the <a href="https://platform.openai.com/docs">OpenAI Docs</a> or fire up ChatGPT’s “Reason” mode today.</p><p><strong>Disclaimer</strong>: O3-Mini does not support vision tasks, and search integration is in early prototype phase. Use with care.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=590706016804" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[DeepSeek-R1: The Open-Source AI Revolutionizing Technology]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@harsh.vardhan7695/deepseek-r1-the-open-source-ai-revolutionizing-technology-1b0914a162da?source=rss-dddbe7e7736e------2"><img src="https://cdn-images-1.medium.com/max/710/1*pzWGJGTFBwk4nDFOYHzrIw.png" width="710"></a></p><p class="medium-feed-snippet">Its just a start of 2025 and even in the fast-paced world of AI, DeepSeek-R1 has taken everyone by surprise. It has transformed what was&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/@harsh.vardhan7695/deepseek-r1-the-open-source-ai-revolutionizing-technology-1b0914a162da?source=rss-dddbe7e7736e------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@harsh.vardhan7695/deepseek-r1-the-open-source-ai-revolutionizing-technology-1b0914a162da?source=rss-dddbe7e7736e------2</link>
            <guid isPermaLink="false">https://medium.com/p/1b0914a162da</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[openai]]></category>
            <category><![CDATA[deepseek]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[open-source]]></category>
            <dc:creator><![CDATA[Harsh Vardhan]]></dc:creator>
            <pubDate>Mon, 27 Jan 2025 21:08:35 GMT</pubDate>
            <atom:updated>2025-02-03T11:37:17.749Z</atom:updated>
        </item>
    </channel>
</rss>