<?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 Sudhanshu Dhamija on Medium]]></title>
        <description><![CDATA[Stories by Sudhanshu Dhamija on Medium]]></description>
        <link>https://medium.com/@sudhanshudhamija?source=rss-fcca2706aa75------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*8-LgIjGKc7O2T6mM0SfSQg@2x.jpeg</url>
            <title>Stories by Sudhanshu Dhamija on Medium</title>
            <link>https://medium.com/@sudhanshudhamija?source=rss-fcca2706aa75------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 09:45:27 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@sudhanshudhamija/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[Welcome to the Era of “Morphware”]]></title>
            <link>https://medium.com/@sudhanshudhamija/welcome-to-the-era-of-morphware-350456b65acb?source=rss-fcca2706aa75------2</link>
            <guid isPermaLink="false">https://medium.com/p/350456b65acb</guid>
            <dc:creator><![CDATA[Sudhanshu Dhamija]]></dc:creator>
            <pubDate>Mon, 02 Nov 2026 12:17:07 GMT</pubDate>
            <atom:updated>2026-11-02T12:17:07.242Z</atom:updated>
            <content:encoded><![CDATA[<p>For the last twenty years, the golden rule of building software-as-a-service (SaaS) was simple: build for the 80%, ignore the 20%.</p><blockquote>If a mid-market customer asked for a bizarre workflow – say, automatically cross-referencing incoming refund requests against local weather patterns and pinging a specific WhatsApp group if the customer was a VIP – the Product Manager’s job was to smile politely and say no. Engineering hours were scarce, technical debt was lethal, and quality, deployable code was the ultimate bottleneck. We conducted user research to find the highest common denominator, prioritized the quarterly roadmap, and forced customers to bend their business processes around our rigid software.</blockquote><blockquote>That bottleneck is evaporating and with it, the traditional SaaS playbook is about to be turned inside out.</blockquote><p>When AI makes writing secure, production-ready code nearly instantaneous and practically free, customers won’t stop buying SaaS or subscribing to consumer apps. They still need a trusted system of record, enterprise security, and core infrastructure. What they will stop tolerating is being told “that feature isn’t on our roadmap.”</p><h3>We are entering the age of Morphware – and a wild new customer behavior I call Sporketing (a mashup of bespoke forking and socketing).</h3><p>What Happens When Every User is a “Market of One”?</p><p>In the very near future, users won’t submit feature requests and wait six months hoping enough other companies complain about the same thing. Instead, businesses and consumers will demand the ability to prompt bespoke features into existence directly inside the products they buy, and socket those products together in weird, hyper-specific ways that only make sense to them.</p><p>Imagine using a travel app or a CRM where the UI and workflow engine aren’t fixed. Need a custom modal that pulls data from an obscure internal legacy database, runs a proprietary pricing formula, and triggers a custom approval flow? You don’t hire an agency or beg your account manager. You simply tell the software what you need, and the software rewrites its own edge layer for your account alone. That is Sporketing: forking a vendor’s core capability and socketing it into your own idiosyncratic reality.</p><p>The Forward-Deployed Engineer Trap</p><p>As this wave hits, the knee-jerk reaction from software vendors will be to throw human bodies at the problem. We are already seeing the early tremors of this: companies hiring armies of Palantir-style Forward-Deployed Engineers (FDEs) or technical account architects.</p><p>In this transitional phase, the FDE’s job is twofold:</p><blockquote>* Act as a human firewall – reviewing, securing, and approving the bespoke workflows that customers (or their AI agents) want to deploy inside the platform.</blockquote><blockquote>* Act as intelligence gatherers – bringing the best custom workflows back to the core product team to be hardened and shipped to everyone.</blockquote><blockquote>Here is my hot take: both of those instincts are a trap.</blockquote><p>First, dragging bespoke customer workflows back to the core product team defeats the entire premise of the AI wave.</p><h3>The beauty of a user building a hyper-specific workflow is precisely that it doesn’t work for everyone.</h3><p>If you merge a thousand niche edge cases into the main product branch, you end up with bloated, unusable Franken-software. Not every good workflow needs to be democratized; most should live and die in the single tenant account that birthed them.</p><p>Second, human FDEs don’t scale as code reviewers. If every customer is generating ten bespoke micro-features a week, an army of account engineers manually checking security permissions will quickly become the exact bottleneck AI was supposed to eliminate.</p><p>The Real Future: Headless Cores and Ephemeral Edges</p><p>So how does this actually play out? If you are building or investing in software today, the architecture of SaaS has to split into two distinct layers:</p><p>* The Immutable Core (What Vendors Sell): This is the bedrock. Vendors will no longer compete on having the prettiest buttons or the most pre-built workflow templates. They will compete on data integrity, state management, permissioning schemas, compliance, and deep domain primitives.</p><p>* The Ephemeral Edge (What Customers Generate): Around that rigid core, every SaaS product will offer a sandboxed execution environment. Customers will generate their own UI views, automation rules, and cross-app bridges on the fly.</p><p>Instead of an army of human FDEs reviewing every custom script, smart software companies will build Automated Policy Gatekeepers – cryptographic sandboxes and AI security auditors built directly into the platform. If a customer’s bespoke workflow tries to leak PII, bypass billing tiers, or drop a core database table, the sandbox blocks it automatically. The human Forward-Deployed Engineer doesn’t disappear, but their role elevates: they stop being code-approvers and become Boundary Architects, helping enterprise clients design the safe sandboxes within which their teams can build freely.</p><p>How Prioritization Changes Forever</p><p>Does this mean product research and roadmaps are dead? Not quite, but they flip from proactive guessing to reactive harvesting.</p><p>In the Morphware era, product teams won’t spend months interviewing users to guess which workflow to build next. They will simply watch the telemetry of what their customers are already “sporketing” at the edge. When a product team notices that 400 different customers have independently prompted a nearly identical custom workflow into existence in their private sandboxes, that is the signal. The core team then steps in, abstracts that pattern into a fast, native platform primitive, and pushes it down into the Immutable Core.</p><p>Software is about to stop feeling like a leased apartment where you aren’t allowed to paint the walls. It’s becoming high-grade structural steel that rearranges itself the moment you walk into the room. The companies that win the next decade won’t be the ones with the best product roadmaps – they will be the ones that build the safest playgrounds for their customers’ imagination.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=350456b65acb" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Rainy Day Holiday]]></title>
            <link>https://medium.com/@sudhanshudhamija/rainy-day-holiday-699543d863e0?source=rss-fcca2706aa75------2</link>
            <guid isPermaLink="false">https://medium.com/p/699543d863e0</guid>
            <category><![CDATA[nostalgia]]></category>
            <category><![CDATA[rain]]></category>
            <category><![CDATA[childhood-memories]]></category>
            <category><![CDATA[childhood]]></category>
            <dc:creator><![CDATA[Sudhanshu Dhamija]]></dc:creator>
            <pubDate>Sun, 01 Nov 2026 13:11:20 GMT</pubDate>
            <atom:updated>2026-11-01T13:11:20.062Z</atom:updated>
            <content:encoded><![CDATA[<p>The evening before, nine-year-old Gudu had finally conquered the sky. He had flown a kite entirely on his own – no big cousins holding the spool, no hovering uncles – and today was supposed to be his coronation. Lunch break was meant for holding court before the mere mortals of his classroom, recounting how he had slain the twin demons of self-doubt and adult supervision.</p><p>Then he woke to the roar on the tin roof.</p><p>Before he even opened his eyes, he knew the Gods had rewritten his script. In nineties Dehradun, monsoon rains didn’t just fall; they claimed the valley, as inseparable from the city’s soul as Tendulkar was from cricket.</p><p>He had barely been awake thirty minutes when the heavy black landline rang in the hallway. His aunt lifted the receiver, listened for three seconds, and smiled. She didn’t even need to say the words. A collective roar went up from Gudu and his cousins, their feet thudding against the cool floor in pure, unfiltered euphoria. In a sprawling joint family with eight adults under one roof, a cancelled school day meant twelve hours of tangled limbs, bruised knees, petty squabbles, and deafening laughter. The grown-ups exchanged weary, affectionate glances and braced themselves. The children were already gone to another world.</p><p>It was a Rainy Day Holiday. To anyone who hadn’t grown up cupped inside that green valley, the phrase meant nothing. In a gentler era before yellow school buses and anxious queues of parents in cars, getting to school meant pedaling a bicycle or walking in dripping raincoats. When the clouds burst over the foothills, the streets turned into amber rivers, and the world simply gave you permission to stay home.</p><p>Out came the shiny wellies, the fleet of creased notebook-paper boats, and broken broomsticks turned into makeshift oars. In the drowned flowerbeds of the back garden, fat yellow frogs leaped through the marigolds like kings of a sunken empire. Inevitably, the electricity surrendered with a quiet sigh, plunging the house into a cool, grey twilight. One of the mothers herded the damp, shivering pack into the dining room, piling plates high with hot breakfast and anchoring their wild energy with stories while the rain drummed against the dark windowpanes.</p><p>Two hours later, as suddenly as it had begun, the sky sealed itself shut. Dehradun performed its quiet magic: the sloped valley drank the floodwaters down into its stony gullies, and within an hour, the tarmac stood washed and clean, as if the storm had been a fever dream.</p><p>Not that the children were willing to surrender the mood. Rain or no rain, a holiday demanded crispy potato pakoras, and the kitchen obliged – platters of golden fritters disappearing alongside mugs of warm, sweet milk. Outside, the shop shutters remained rolled down, surrendering the gleaming lane to the neighborhood. Wickets were chalked onto shopfront doors, and for one glorious afternoon, uncles and nine-year-olds played side by side on an empty street that belonged entirely to them.</p><p>Gudu’s great kite triumph had nearly drowned in the day’s magic, though he tucked it safely into the back of his mind as dusk settled – saving his glory for tomorrow, right alongside the bragging rights of today’s match.</p><p>Decades later, across the chaotic sprawl of Delhi and eventually beneath the polite, drizzling skies of London, Gudu would watch the rain hit the glass and try to explain it to his wife and friends. He would tell them about the landline ringing, the flooded marigolds, and the whole street turning into a cricket pitch. They would smile and nod, hearing a charming story, never quite grasping the ache behind it – how a whole childhood could be held inside a phrase nobody else knew. Some places don’t exist on maps anymore; they live only in the smell of wet earth, the memory of Dehradun, and the faded ink of a nine-year-old’s diary.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=699543d863e0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why Switching AI Providers Should Work Like Switching Broadband in the UK]]></title>
            <link>https://medium.com/@sudhanshudhamija/why-switching-ai-providers-should-work-like-switching-broadband-in-the-uk-c0f364c0e15c?source=rss-fcca2706aa75------2</link>
            <guid isPermaLink="false">https://medium.com/p/c0f364c0e15c</guid>
            <category><![CDATA[world]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[privacy]]></category>
            <dc:creator><![CDATA[Sudhanshu Dhamija]]></dc:creator>
            <pubDate>Tue, 27 Oct 2026 16:08:11 GMT</pubDate>
            <atom:updated>2026-10-27T16:08:11.374Z</atom:updated>
            <content:encoded><![CDATA[<p>We have a fragmented memory problem in AI. Until we decouple user context from model providers, true consumer choice is an illusion.</p><p>For the past few years, my AI stack has looked like a relay race where the runners refuse to pass the baton.</p><p>I was an early, heavy user of ChatGPT. Months of brainstorming, architectural debates, and personal preferences were baked into its memory. Eventually, I migrated my core daily workflows to Claude for its reasoning and writing nuances. Meanwhile, I also use Gemini for lightweight tasks because it comes bundled with my cloud storage.</p><p>The result? Three of the most advanced intelligence engines on the planet hold three completely fractured versions of who I am.</p><p>ChatGPT knows how I approached problems two years ago. Claude knows what I am building today. Gemini gets the quick, transactional questions in between. None of them share a continuous timeline, and none of them truly know me as a whole.</p><p>Yes, the platforms will tell you that data portability exists. You can request an archive, wait for an email, download a massive .zip file of raw JSON, and upload it into a competitor’s import tool. Or you can run clever “memory extraction” prompts and copy-paste markdown files across tabs.</p><p>Let’s be honest: that isn’t interoperability. That’s digital moving day. And the friction is deliberate.</p><p>The Utility Precedent: One Touch Switch</p><p>We have already solved this exact problem in regulated utilities and telecommunications. Look at how broadband or electricity switching works in the UK today.</p><p>Over the years, I have switched my home broadband from Vodafone to BT and back to Vodafone as my needs and tariffs changed. When I decide to switch, I don’t climb a telegraph pole to rip out fiber-optic cables, and I don’t spend a weekend manually translating my household’s consumption history from one proprietary format to another.</p><p>At the tap of a button, the switch happens in the background.</p><p>Why does this work so seamlessly? Because we separated the service provider from the underlying infrastructure:</p><p>* The Shared Rail: Whether it is the physical fiber network (like Openreach) or the national Data Communications Company (DCC) that sits behind UK smart meters, the foundational layer is treated as common infrastructure. Retail providers lease access to the rail; they don’t own the pipes entering your home.</p><p>* Mandated Handoff Protocols: Under regulatory frameworks like Ofcom’s One Touch Switch, the losing and gaining providers are legally required to communicate in the background to hand off the customer without downtime or manual intervention.</p><p>* Standardized Telemetry: My smart meter doesn’t stop reading my energy usage or lose its historical baseline just because I changed the logo on my monthly bill. The new supplier simply plugs into the existing feed.</p><p>In utilities, providers compete on price, reliability, and service – not on how hard they can make it for you to leave.</p><p>The AI Trap: Context as a Hostage</p><p>Right now, AI labs are competing on two fronts: raw model intelligence (compute and weights) and accumulated user context (memory, project files, style preferences, and history).</p><p>While competing on model intelligence drives rapid innovation, using accumulated context as a moat is fundamentally anti-consumer. Every conversation you have with an LLM refines an implicit profile of how you think, work, and communicate. When that context is trapped inside a single vendor’s walled garden, switching models means lobotomizing your assistant and starting from scratch.</p><p>Retention should be a choice driven by product quality, not a compulsion driven by switching costs.</p><p>If Model A releases a breakthrough in coding tomorrow, and Model B is better at strategic synthesis on Thursday, consumers and enterprises shouldn’t have to pay a “context tax” to use the best tool for the job.</p><p>The Blueprint: The “Context Twin” and the Sovereign Vault</p><p>To fix this, we need an architectural shift in how AI memory works. We need to stop treating context as a byproduct owned by the model provider, and start treating it as a portable asset owned by the user.</p><p>Imagine an ecosystem built around a Personal Context Vault – an encrypted, user-owned data layer that houses your Context Twin.</p><p>Instead of uploading your life story into OpenAI, Anthropic, and Google separately, the relationship flips:</p><p>* Plug-and-Play Authorization: You store your Context Twin in a secure vault of your choosing. When you subscribe to Claude, ChatGPT, or Gemini, you grant that model temporary, scoped read/write access to your vault – much like OAuth lets you sign into apps without handing over your master credentials, or how Open Banking lets financial apps read your ledger via standardized APIs.</p><p>* The Zero-Friction Contract: The core agreement between the user and the AI provider is simple: as you interact with the model, the provider must continuously write structured context updates back to your vault in an open, universal schema.</p><p>* Instant Unplugging: If you decide to switch providers – or use three different models simultaneously – there is nothing to export. You revoke access from Provider A and grant access to Provider B. The new model plugs into your Context Twin and picks up the exact thread you were working on thirty seconds ago.</p><p>* Integrity Without Replication: Crucially, the vault acts as the single source of truth. Providers can read from it to personalize responses and write back new learnings via strict validation rules, but they cannot corrupt the core state or hoard a proprietary shadow copy to lock you in once you leave.</p><p>Why This Inevitably Wins</p><p>AI companies will naturally resist this at first, just as telecom giants once resisted number portability and banks resisted Open Banking. Giving up the data moat feels dangerous when models are commoditizing at breakneck speed.</p><p>But history shows that when you remove switching friction, the entire market expands.</p><p>For enterprises, a portable context layer eliminates the terrifying vendor lock-in of building company knowledge bases inside a single LLM provider. For everyday consumers, it means our digital assistants finally work for us, growing smarter across our entire lives rather than fracturing across our subscriptions.</p><p>Intelligence should be a competitive utility we lease on demand. Our memory, however, belongs to us. It’s time we built the smart meter for AI context.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c0f364c0e15c" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>