<?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 RAKESH CHANDA on Medium]]></title>
        <description><![CDATA[Stories by RAKESH CHANDA on Medium]]></description>
        <link>https://medium.com/@rakeshchanda?source=rss-b0160f08a440------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*QT8yPrVhZduPs3gIMSg3_Q.jpeg</url>
            <title>Stories by RAKESH CHANDA on Medium</title>
            <link>https://medium.com/@rakeshchanda?source=rss-b0160f08a440------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 13:50:59 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@rakeshchanda/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[The Hidden Story of TypeSafe AI’s JEV Model]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@rakeshchanda/the-hidden-story-of-typesafe-ais-jev-model-2f1dcca44395?source=rss-b0160f08a440------2"><img src="https://cdn-images-1.medium.com/max/1672/1*1I_SR3u4QQniPnwZcYAfeg.png" width="1672"></a></p><p class="medium-feed-snippet">A Model that doesn&#x2019;t write, but rather decides. The whole story discussed.</p><p class="medium-feed-link"><a href="https://medium.com/@rakeshchanda/the-hidden-story-of-typesafe-ais-jev-model-2f1dcca44395?source=rss-b0160f08a440------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@rakeshchanda/the-hidden-story-of-typesafe-ais-jev-model-2f1dcca44395?source=rss-b0160f08a440------2</link>
            <guid isPermaLink="false">https://medium.com/p/2f1dcca44395</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[ai-agent]]></category>
            <dc:creator><![CDATA[RAKESH CHANDA]]></dc:creator>
            <pubDate>Thu, 24 Sep 2026 15:39:06 GMT</pubDate>
            <atom:updated>2026-09-24T15:39:06.701Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Billions of Vectors. Milliseconds. How HNSW Makes It Possible]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@rakeshchanda/billions-of-vectors-milliseconds-how-hnsw-makes-it-possible-5b88a3673584?source=rss-b0160f08a440------2"><img src="https://cdn-images-1.medium.com/max/2000/0*aNUWlARGOfy9ccvA.png" width="2000"></a></p><p class="medium-feed-snippet">Learn how HNSW navigates billions of vectors, balances latency and recall, and powers modern AI search.</p><p class="medium-feed-link"><a href="https://medium.com/@rakeshchanda/billions-of-vectors-milliseconds-how-hnsw-makes-it-possible-5b88a3673584?source=rss-b0160f08a440------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@rakeshchanda/billions-of-vectors-milliseconds-how-hnsw-makes-it-possible-5b88a3673584?source=rss-b0160f08a440------2</link>
            <guid isPermaLink="false">https://medium.com/p/5b88a3673584</guid>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[software-architecture]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[vector-database]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[RAKESH CHANDA]]></dc:creator>
            <pubDate>Thu, 03 Sep 2026 02:31:00 GMT</pubDate>
            <atom:updated>2026-09-03T02:31:00.944Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[The horse was never the problem]]></title>
            <link>https://medium.com/@rakeshchanda/the-horse-was-never-the-problem-3bac81a4f363?source=rss-b0160f08a440------2</link>
            <guid isPermaLink="false">https://medium.com/p/3bac81a4f363</guid>
            <category><![CDATA[harness-engineering]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[agentic-ai]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[software-engineering]]></category>
            <dc:creator><![CDATA[RAKESH CHANDA]]></dc:creator>
            <pubDate>Tue, 01 Sep 2026 06:50:34 GMT</pubDate>
            <atom:updated>2026-09-01T06:50:34.376Z</atom:updated>
            <content:encoded><![CDATA[<h3>Every agent failure I have debugged traced back to the same missing piece and it wasn’t the model.</h3><p>Give an agent shell access. Tell it: “clean up the test artifacts.”</p><p>Ten minutes later, it deleted a folder that wasn’t a test artifact. The pattern matched. Nothing stopped it.</p><p>The model didn’t get dumber. Nothing in its environment made “ask first” more likely than “just do it.” That environment has a name: the harness.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/658/0*8L5Jc0faZJJvfQVD.png" /></figure><h3>The horse was never the problem</h3><p>A horse can pull huge weight. Alone in field, that strength does nothing. Hitch it to a harness - collar, reins, bit and the same strength plows a straight furrow.</p><p>The harness doesn’t make the horse stronger. It aims strength that already exists.</p><p>A model already has the skill to write the fix, run the tests, open the PR. What decides if it does that safely, on <em>this</em> codebase, without touching the wrong file that’s not the model. That’s the harness: what tools it’s given, what needs approval, what’s in its context window, whether a failure gets shown to it or swallowed.</p><p>Where the analogy breaks: a horse doesn’t imagine a bridge that isn’t there and walk onto it. A model can. Its context window fills and gets summarized lose the wrong line during that, and the model never knew it. A horse’s strength is steady. A model’s competence can vanish mid-task for reasons invisible in the transcript. The harness exists to catch that too.</p><h3>What “harness” actually means</h3><p>Harness engineering is everything between raw model capability and a reliable outcome. Not the model that’s training. Not one good instruction that’s prompt engineering, and it’s just one piece.</p><p>The harness is:</p><ul><li><strong>Tools.</strong> What actions exist read a file, run a command, edit a line.</li><li><strong>The loop.</strong> Model proposes an action. Harness runs it. Result goes back in. Repeat. That loop, run enough times, is what “agent” means.</li><li><strong>Permission gates.</strong> What runs immediately, what waits for a human, what’s blocked.</li><li><strong>Context management.</strong> What’s in the window, and what gets cut when it’s full.</li><li><strong>Feedback.</strong> When a tool call fails, does the model see the real error or nothing, and repeat the same mistake?</li></ul><p>None of this touches the model’s weights. All of it decides whether the same model ships a clean PR or wrecks the repo.</p><h3>You can watch one work</h3><p>A coding agent doesn’t get raw machine access by default. It gets a fixed set of tools. Destructive git commands force-push, hard reset need explicit confirmation first. Not because the model can’t type the command. Because the harness decided that action needs a human before it runs.</p><p>Routine edits flow straight through. Risky ones stop and ask. That split isn’t a model behavior it’s a policy the harness enforces no matter what the model tries next.</p><h3>When nothing is steering</h3><p>Same cleanup task, no harness: full filesystem access, a vague instruction, no distinction between a stray .pyc file and a folder with three months of uncommitted work. The model doesn’t pause on the second one. Nothing told it the second one was different.</p><p>Or: an agent’s only tool is a flaky API call. It fails. The harness shows the model “action failed” no detail. The model retries the same call. Then again. Thirty calls later, still retrying, because nothing in its context distinguishes attempt one from attempt thirty. Not a dumb model. A harness with no error detail and no retry cap.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*OgeFYE-T79jL7aYd.png" /></figure><h3>From one call to a fleet</h3><p>One API call is barely a harness. Add a tool and a loop: a minimal agent. Add more tools shell, search, browser and the harness now decides what runs automatically, what needs a human, and how to manage a context window filling up fast.</p><p>Push further: a coordinator spins up sub-agents for separate pieces of a task, each with its own tools, reporting back conclusions, not raw work. At that scale the harness isn’t just gating actions. It’s deciding what one agent even gets to know about another.</p><p>More horses need a driver who can hold more than one set of reins. More agents need a harness that can hold more than one thread of trust.</p><h3>What the harness charges you</h3><p>Tight gates confirm every write are safe and slow. Ask a human often enough, and they stop reading the confirmations. The gate is still there. It stopped working.</p><p>Loose gates — auto-allow broad actions are fast, and occasionally catastrophic, in proportion to how broad “broad” was.</p><p>Bigger context windows mean less compacting, less risk of losing a constraint mid-task. They also cost more, and a model doesn’t attend to token one thousand the way it attends to token one. Bigger isn’t free.</p><p>Sandboxing — running actions in an isolated environment instead of straight on production costs latency and setup. It’s the difference between a bad guess deleting a test fixture and a bad guess deleting production data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*J2uFxR3hddAbdgXE.jpeg" /></figure><h3>The mental model</h3><p>The model is the horse. The harness is what turns raw pulling power into a straight furrow instead of a runaway cart. You don’t get a better plow by breeding a stronger horse. You get it by building a harness around the horse you already have.</p><p>The last thing a good farmer worries about is whether the horse is strong enough.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3bac81a4f363" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Microsoft Fabric and OneLake: Innovation or Just a Rebrand?]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/art-of-data-engineering/microsoft-fabric-and-onelake-innovation-or-just-a-rebrand-cb8b1d5f7299?source=rss-b0160f08a440------2"><img src="https://cdn-images-1.medium.com/max/875/0*KvVoTZO_WaHDdgne.png" width="875"></a></p><p class="medium-feed-snippet">Microsoft Fabric and OneLake are marketed as groundbreaking cloud solutions, offering seamless data management and analytics. But a closer&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/art-of-data-engineering/microsoft-fabric-and-onelake-innovation-or-just-a-rebrand-cb8b1d5f7299?source=rss-b0160f08a440------2">Continue reading on Art of Data Engineering »</a></p></div>]]></description>
            <link>https://medium.com/art-of-data-engineering/microsoft-fabric-and-onelake-innovation-or-just-a-rebrand-cb8b1d5f7299?source=rss-b0160f08a440------2</link>
            <guid isPermaLink="false">https://medium.com/p/cb8b1d5f7299</guid>
            <category><![CDATA[microsoft]]></category>
            <category><![CDATA[data-storage]]></category>
            <category><![CDATA[data-engineering]]></category>
            <category><![CDATA[big-data]]></category>
            <category><![CDATA[data-lakehouse]]></category>
            <dc:creator><![CDATA[RAKESH CHANDA]]></dc:creator>
            <pubDate>Mon, 16 Jun 2025 14:32:38 GMT</pubDate>
            <atom:updated>2025-06-18T10:41:15.844Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Agentic AI in Action: Building Self-Healing Data Pipelines]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/towards-data-engineering/agentic-ai-in-action-building-self-healing-data-pipelines-343bc4aeb50f?source=rss-b0160f08a440------2"><img src="https://cdn-images-1.medium.com/max/600/0*5aS7hO5ECGP66Ial.png" width="600"></a></p><p class="medium-feed-snippet">Introduction</p><p class="medium-feed-link"><a href="https://medium.com/towards-data-engineering/agentic-ai-in-action-building-self-healing-data-pipelines-343bc4aeb50f?source=rss-b0160f08a440------2">Continue reading on Towards Data Engineering »</a></p></div>]]></description>
            <link>https://medium.com/towards-data-engineering/agentic-ai-in-action-building-self-healing-data-pipelines-343bc4aeb50f?source=rss-b0160f08a440------2</link>
            <guid isPermaLink="false">https://medium.com/p/343bc4aeb50f</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[analytics]]></category>
            <category><![CDATA[data-engineering]]></category>
            <category><![CDATA[big-data]]></category>
            <category><![CDATA[data-pipeline]]></category>
            <dc:creator><![CDATA[RAKESH CHANDA]]></dc:creator>
            <pubDate>Sun, 15 Jun 2025 19:14:11 GMT</pubDate>
            <atom:updated>2025-06-16T04:38:12.075Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Microsoft Fabric for Data Engineers: A Practical Introduction]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/towards-data-engineering/microsoft-fabric-for-data-engineers-a-practical-introduction-bcc0d7841c41?source=rss-b0160f08a440------2"><img src="https://cdn-images-1.medium.com/max/1000/0*6dAoyVzV_Xcld-3r" width="1000"></a></p><p class="medium-feed-snippet">Why Microsoft Fabric Matters to Data Engineers</p><p class="medium-feed-link"><a href="https://medium.com/towards-data-engineering/microsoft-fabric-for-data-engineers-a-practical-introduction-bcc0d7841c41?source=rss-b0160f08a440------2">Continue reading on Towards Data Engineering »</a></p></div>]]></description>
            <link>https://medium.com/towards-data-engineering/microsoft-fabric-for-data-engineers-a-practical-introduction-bcc0d7841c41?source=rss-b0160f08a440------2</link>
            <guid isPermaLink="false">https://medium.com/p/bcc0d7841c41</guid>
            <category><![CDATA[spark]]></category>
            <category><![CDATA[big-data-analytics]]></category>
            <category><![CDATA[microsoft]]></category>
            <category><![CDATA[data-engineering]]></category>
            <category><![CDATA[fabric]]></category>
            <dc:creator><![CDATA[RAKESH CHANDA]]></dc:creator>
            <pubDate>Sun, 15 Jun 2025 19:00:17 GMT</pubDate>
            <atom:updated>2025-06-16T04:38:14.394Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Agentic AI in Data Engineering: From Automation to Autonomous Intelligence]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/art-of-data-engineering/agentic-ai-in-data-engineering-from-automation-to-autonomous-intelligence-bfdd5483fe93?source=rss-b0160f08a440------2"><img src="https://cdn-images-1.medium.com/max/990/0*41eaJr7BjeA5fXjr.jpeg" width="990"></a></p><p class="medium-feed-snippet">Why AI Agents Are Set to Redefine the Future of Data Infrastructure</p><p class="medium-feed-link"><a href="https://medium.com/art-of-data-engineering/agentic-ai-in-data-engineering-from-automation-to-autonomous-intelligence-bfdd5483fe93?source=rss-b0160f08a440------2">Continue reading on Art of Data Engineering »</a></p></div>]]></description>
            <link>https://medium.com/art-of-data-engineering/agentic-ai-in-data-engineering-from-automation-to-autonomous-intelligence-bfdd5483fe93?source=rss-b0160f08a440------2</link>
            <guid isPermaLink="false">https://medium.com/p/bfdd5483fe93</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[data-engineering]]></category>
            <category><![CDATA[agentic-ai]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[RAKESH CHANDA]]></dc:creator>
            <pubDate>Tue, 06 May 2025 15:19:24 GMT</pubDate>
            <atom:updated>2025-05-27T15:30:17.113Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[From ETL to AI Agents: How AI Is Transforming Data Engineering]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://blog.dataengineerthings.org/from-etl-to-ai-agents-how-ai-is-transforming-data-engineering-1e9d0f54c187?source=rss-b0160f08a440------2"><img src="https://cdn-images-1.medium.com/max/1024/0*ckdXC34JDf6benSt" width="1024"></a></p><p class="medium-feed-snippet">How AI Is Reshaping the Future of Data Engineering</p><p class="medium-feed-link"><a href="https://blog.dataengineerthings.org/from-etl-to-ai-agents-how-ai-is-transforming-data-engineering-1e9d0f54c187?source=rss-b0160f08a440------2">Continue reading on Data Engineer Things »</a></p></div>]]></description>
            <link>https://blog.dataengineerthings.org/from-etl-to-ai-agents-how-ai-is-transforming-data-engineering-1e9d0f54c187?source=rss-b0160f08a440------2</link>
            <guid isPermaLink="false">https://medium.com/p/1e9d0f54c187</guid>
            <category><![CDATA[artificalintelligence]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[data-engineering]]></category>
            <category><![CDATA[agents]]></category>
            <dc:creator><![CDATA[RAKESH CHANDA]]></dc:creator>
            <pubDate>Sun, 04 May 2025 10:01:57 GMT</pubDate>
            <atom:updated>2025-05-04T10:01:57.105Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[From ETL to AI Agents: How AI Is Transforming Data Engineering]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@rakeshchanda/from-etl-to-ai-agents-how-ai-is-transforming-data-engineering-69feb31bd5f0?source=rss-b0160f08a440------2"><img src="https://cdn-images-1.medium.com/max/1024/0*ckdXC34JDf6benSt" width="1024"></a></p><p class="medium-feed-snippet">How AI Is Reshaping the Future of Data Engineering</p><p class="medium-feed-link"><a href="https://medium.com/@rakeshchanda/from-etl-to-ai-agents-how-ai-is-transforming-data-engineering-69feb31bd5f0?source=rss-b0160f08a440------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@rakeshchanda/from-etl-to-ai-agents-how-ai-is-transforming-data-engineering-69feb31bd5f0?source=rss-b0160f08a440------2</link>
            <guid isPermaLink="false">https://medium.com/p/69feb31bd5f0</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[data-engineering]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[RAKESH CHANDA]]></dc:creator>
            <pubDate>Sat, 03 May 2025 18:44:10 GMT</pubDate>
            <atom:updated>2025-05-03T18:44:10.027Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[I Spent Hours Unraveling Database Sharding — Here’s What I Discovered]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/art-of-data-engineering/i-spent-hours-unraveling-database-sharding-heres-what-i-discovered-3d7bb8f6a572?source=rss-b0160f08a440------2"><img src="https://cdn-images-1.medium.com/max/1584/0*6vlpG3N4fSIB-PH4.png" width="1584"></a></p><p class="medium-feed-snippet">Database Sharding: How to Boost Performance by Splitting the Load</p><p class="medium-feed-link"><a href="https://medium.com/art-of-data-engineering/i-spent-hours-unraveling-database-sharding-heres-what-i-discovered-3d7bb8f6a572?source=rss-b0160f08a440------2">Continue reading on Art of Data Engineering »</a></p></div>]]></description>
            <link>https://medium.com/art-of-data-engineering/i-spent-hours-unraveling-database-sharding-heres-what-i-discovered-3d7bb8f6a572?source=rss-b0160f08a440------2</link>
            <guid isPermaLink="false">https://medium.com/p/3d7bb8f6a572</guid>
            <category><![CDATA[sql]]></category>
            <category><![CDATA[big-data]]></category>
            <category><![CDATA[database]]></category>
            <category><![CDATA[data-engineering]]></category>
            <category><![CDATA[data-analytics]]></category>
            <dc:creator><![CDATA[RAKESH CHANDA]]></dc:creator>
            <pubDate>Sun, 15 Sep 2024 10:46:51 GMT</pubDate>
            <atom:updated>2024-09-16T04:00:11.763Z</atom:updated>
        </item>
    </channel>
</rss>