User Experience

“Loop engineering” is a hot buzzphrase after mentions of it by Boris Cherny (Claude Code’s creator) and Peter Steinberger (OpenClaw's creator) went viral on social media. Loops are now a key part of how we get AI agents to iterate at length to build software. In this letter, I’d like to share my 3 key loops, shown in the image below, for building 0-to-1 products. These loops guide not just how I build software, but also how I decide what software to build. Agentic coding loop: Given a product specification and optionally a set of evals (that is, a dataset against which to measure performance), we can have an AI agent write code, test its work, and keep iterating until the code is bug-free and meets its specification. This idea of closing the loop took off around the end of last year, and it has been a game changer in enabling coding agents to work longer productively without human intervention. For example, over the weekend, I was building an app for my daughter to practice typing, and my coding agent could easily work for around an hour, using a web browser to check what it had built multiple times before getting back to me, without needing my intervention. The engineering loop executes quickly. Every few minutes, the coding agent might build and test a new version of the software. I hear frequently from developers who are finding new ways to engineer more effective engineering loops. This is an active area of invention! Developer feedback loop: In this loop, a developer examines the current product and steers the coding agent to improve it. Last year, a lot of developers (including me) were acting as the QA (quality assurance) function for our coding agents, manually finding bugs and then asking the agent to fix them. But with coding agents much more able to test their own code, the amount of time we need to spend on this function has decreased significantly. This allows us to make higher-level product decisions, such as what key features to offer, where the UI needs improvement, and so on. The developer-feedback loop operates over time intervals between tens of minutes and hours — that's how frequently a developer might review a product and give feedback. In the case of the typing app, I changed my mind a few times about the visual design, what cat costumes she can unlock as she learns (she loves cats), and the user flow for a grown-up to log in and steer the child's learning experience. When a developer has a clear vision for what to build, it is still a lot of work to translate that vision into a specification for a coding agent to implement. Further, after the developer has seen an implementation, they might update (or perhaps clarify) the spec to steer it toward what they want. If you find that the system repeatedly runs into certain problems, building a set of evals for the agent becomes useful. [Truncated for length. Full text: https://lnkd.in/gKDQ6H9s]

𝗡𝗮𝗶𝘃𝗲 𝗥𝗔𝗚 𝘄𝗼𝗿𝗸𝘀 𝗶𝗻 𝗮 𝗱𝗲𝗺𝗼. 𝗜𝘁 𝗳𝗮𝗶𝗹𝘀 𝘁𝗵𝗲 𝗺𝗼𝗺𝗲𝗻𝘁 𝗿𝗲𝗮𝗹 𝘂𝘀𝗲𝗿𝘀 𝘀𝗵𝗼𝘄 𝘂𝗽. Embed → retrieve → generate looks clean in a notebook. Real requirements break it: → Questions whose answer is spread across many documents → Industry terms that embeddings get wrong → Bad chunks the pipeline never catches → Answers that live in how things connect, not in any single chunk → PDFs full of tables and images a text-only index cannot read These 5 architectures are how serious teams stay ahead in the agentic AI era: 𝟬𝟭 𝗛𝘆𝗯𝗿𝗶𝗱 𝗥𝗔𝗚 → Dense vectors find meaning. BM25 finds exact words. → Reciprocal Rank Fusion combines both ranked lists. → A safe baseline for almost every team. 𝟬𝟮 𝗚𝗿𝗮𝗽𝗵𝗥𝗔𝗚 → Pull entities and their relationships into a knowledge graph. → Retrieve subgraphs and community summaries, not chunks. → Best when the answer lives in how things connect. 𝟬𝟯 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗥𝗔𝗚 → A planner agent picks the right tool: vector, web, or SQL. → A reasoner agent keeps trying until the answer is solid. → Retrieval becomes a plan, not a single step. 𝟬𝟰 𝗖𝗼𝗿𝗿𝗲𝗰𝘁𝗶𝘃𝗲 𝗥𝗔𝗚 (𝗖𝗥𝗔𝗚) → Grade every retrieval before you trust it. → Correct → answer. Unclear → rewrite the query. Wrong → search the web. → This is what production RAG actually looks like. 𝟬𝟱 𝗠𝘂𝗹𝘁𝗶𝗺𝗼𝗱𝗮𝗹 𝗥𝗔𝗚 → One embedding model (CLIP, ColPali) for text, images, and tables. → One vector index. One multimodal LLM. → No more separate pipelines for PDFs with charts. I built a runnable example for each of the five patterns. GitHub link in the first comment. The best teams in 2026 do not pick one. They combine them — hybrid retrieval inside an agentic loop, with a corrective grader, over a multimodal index. Naive RAG is a starting point, not a finish line. That is why most enterprise GenAI projects stall at the demo. Which of these five becomes the default RAG stack in the next 18 months — and which stays a specialized tool?

For decades, buying something has relied on a simple, visible sign of intent. You tap your card or hit “buy” online, and everyone knows exactly what you agreed to. That clear moment of yes is what’s made commerce work.   But what happens when that equation changes? What’s required is not just smarter tech, but a new model of trust, with privacy at the center.   As agents take on more tasks for us, trust can’t rely on a single click or tap. It must be recorded verified, and shared. Mastercard is working with partners like Google, Adyen, Checkout.com, Fiserv, Worldpay, IBM, Getnet, and Basis Theory to keep people in control and businesses confident agents are authorized to transact.     In a world where we delegate actions, intent must be clear and easy to verify.

🚛 WHEN TRANSPORT LEARNS TO THINK GREEN I came across a concept today that stopped me — an autonomous hydrogen truck-trailer drone designed for long-distance freight. At first, it looked like another futuristic vehicle. But then it hit me: this isn’t just transport evolving — it’s intent evolving. For decades, we’ve designed logistics around speed and scale. Now we’re finally designing around sustainability. This new concept merges autonomy, aerodynamics, and hydrogen power to do something radical: → Eliminate carbon emissions in heavy freight. → Cut operational energy costs through intelligent routing. → Reduce highway congestion with coordinated drone convoys. It’s not just engineering — it’s a shift in philosophy. A move from moving faster to moving responsibly. We often talk about “green tech” as a feature — but the real shift happens when sustainability becomes the invisible infrastructure behind innovation. It’s not an addition to progress. It is progress. What’s needed now isn’t more invention — it’s integration. We need to: ✅ Build networks where clean energy and automation reinforce each other. ✅ Redefine “efficiency” to include environmental balance. ✅ Shift from carbon offsetting to carbon prevention at design level. Because the next breakthrough won’t come from faster engines — but from systems that make waste impossible by design. That’s when technology stops being an experiment in innovation… and becomes an expression of intelligence. So here’s the question I keep returning to — 👉 Will the next era of transport be powered by fuel — or by foresight? #Innovation #Sustainability #Hydrogen #AutonomousVehicles #GreenTech #Logistics #FutureThinking

Invisible UX is coming 🔥 And it’s going to change how we design products, forever. For decades, UX design has been about guiding users through an experience. We’ve done that with visible interfaces: Menus. Buttons. Cards. Sliders. We’ve obsessed over layouts, states, and transitions. But with AI, a new kind of interface is emerging: One that’s invisible. One that’s driven by intent, not interaction. Think about it: You used to: → Open Spotify → Scroll through genres → Click into “Focus” → Pick a playlist Now you just say: “Play deep focus music.” No menus. No tapping. No UI. Just intent → output. You used to: → Search on Airbnb → Pick dates, guests, filters → Scroll through 50+ listings Now we’re entering a world where you guide with words: “Find me a cabin near Oslo with a sauna, available next weekend.” So the best UX becomes barely visible. Why does this matter? Because traditional UX gives users options. AI-native UX gives users outcomes. Old UX: “Here are 12 ways to get what you want.” New UX: “Just tell me what you want & we’ll handle the rest.” And this goes way beyond voice or chat. It’s about reducing friction. Designing systems that understand intent. Respond instantly. And get out of the way. The UI isn’t disappearing. It’s mainly dissolving into the background. So what should designers do? Rethink your role. Going forward you’ll not just lay out screens. You’ll design interactions without interfaces. That means: → Understanding how people express goals → Guiding model behavior through prompt architecture → Creating invisible guardrails for trust, speed, and clarity You are basically designing for understanding. The future of UX won’t be seen. It will be felt. Welcome to the age of invisible UX. Ready for it?

McKinsey & Company 𝗯𝗹𝘂𝗲𝗽𝗿𝗶𝗻𝘁 𝗳𝗼𝗿 𝗵𝗼𝘄 𝗯𝗮𝗻𝗸𝘀 𝗰𝗮𝗻 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝗲𝘅𝘁𝗿𝗮𝗰𝘁 𝗿𝗲𝗮𝗹 𝘃𝗮𝗹𝘂𝗲 𝗳𝗿𝗼𝗺 𝗔𝗜: ⬇️ This is a full-stack, enterprise-grade architecture — built on agents, orchestration, and rewired workflows. The AI bank stack consists out of 4 key layers: ⬇️ 𝟭. 𝗘𝗻𝗴𝗮𝗴𝗲𝗺𝗲𝗻𝘁 𝗟𝗮𝘆𝗲𝗿 This is the user layer — customers and employees. McKinsey calls for fully reimagined, intelligent, personalized experiences across all channels. → Multimodal chat (text, voice, image) → Omnichannel UX across mobile, contact center, branch → Digital twins for customer simulation and workforce training It’s all about a UI refresh and UX overhaul grounded in real AI. 𝟮. 𝗔𝗜-𝗣𝗼𝘄𝗲𝗿𝗲𝗱 𝗗𝗲𝗰𝗶𝘀𝗶𝗼𝗻 𝗠𝗮𝗸𝗶𝗻𝗴 This is the brain of the AI-first bank. And it’s not just predictive models anymore — it’s orchestrated agent ecosystems. → AI Orchestrators: Plan, reason, delegate across workflows → Domain Agents: Specialize in credit policy, fraud, risk, legal → Copilots: Embedded in workflows to guide users and automate decisions McKinsey reports 20–60% productivity gains in decision-making with this approach. 𝟯. 𝗖𝗼𝗿𝗲 𝗧𝗲𝗰𝗵 & 𝗗𝗮𝘁𝗮 The foundation layer most banks underestimate — until GenAI models stall in production. → Vector databases → LLM orchestration and FinOps → Search and retrieval engines → ML pipelines → Secure data architecture → API infrastructure The goal: make data accessible, tools reusable, and infra invisible to the business. Without this, nothing scales. 𝟰. 𝗢𝗽𝗲𝗿𝗮𝘁𝗶𝗻𝗴 𝗠𝗼𝗱𝗲𝗹 This is where the transformation wins or fails. Without rewiring the org, the tech doesn’t matter. → AI control towers to track value and set guardrails → Cross-functional teams across business, tech, and AI → Platform operating model for speed and alignment → Enterprise-wide reuse of AI capabilities If you're building isolated projects without shared assets or central coordination, you’re not transforming — you’re experimenting. 𝗪𝗵𝗮𝘁 𝘁𝗵𝗶𝘀 𝗮𝗹𝗹 𝗮𝗱𝗱𝘀 𝘂𝗽 𝘁𝗼? The banks that win won’t be the ones with the most pilots. They’ll be the ones that industrialize agents, orchestration, and rewired workflows, with full-stack coordination. Full McKinsey article: https://lnkd.in/dPaJzVK4 𝗜 𝗲𝘅𝗽𝗹𝗼𝗿𝗲 𝘁𝗵𝗲𝘀𝗲 𝗱𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁𝘀 — 𝗮𝗻𝗱 𝘄𝗵𝗮𝘁 𝘁𝗵𝗲𝘆 𝗺𝗲𝗮𝗻 𝗳𝗼𝗿 𝗿𝗲𝗮𝗹-𝘄𝗼𝗿𝗹𝗱 𝘂𝘀𝗲 𝗰𝗮𝘀𝗲𝘀 — 𝗶𝗻 𝗺𝘆 𝘄𝗲𝗲𝗸𝗹𝘆 𝗻𝗲𝘄𝘀𝗹𝗲𝘁𝘁𝗲𝗿. 𝗬𝗼𝘂 𝗰𝗮𝗻 𝘀𝘂𝗯𝘀𝗰𝗿𝗶𝗯𝗲 𝗵𝗲𝗿𝗲 𝗳𝗼𝗿 𝗳𝗿𝗲𝗲: https://lnkd.in/dbf74Y9E

This isn’t a luxury. This is a $200 wheelchair redefining what’s possible. For millions, standing wheelchairs have always been out of reach. Until now. At R2D2, IIT Madras, a team dared to ask: What if mobility wasn’t a privilege, but a right? Their answer is a simple innovation -no Big Tech: A wheelchair that lets you stand—on your terms Ingenious gas-spring technology for seamless movement: -Supports up to 242 pounds -Priced at $200 (when others cost $2,000 or more) But the true breakthrough isn’t just in the engineering. It’s in the lives transformed. → Physical freedom is restored. Stand tall when you choose. Reach the top shelf. Cook your own meals. Keep your body strong and active. → Health is protected. Standing improves circulation. Strengthens bones. Prevents pressure sores. Aids digestion. Reduces heart risks. → Social inclusion becomes reality. Converse at eye level. Join meetings—no barriers. Participate fully in community life. Experience true belonging. Ask yourself: When was the last time you had to look up just to be heard? For millions, that’s every day. This isn’t only about standing. It’s about dignity. It’s about independence. It’s about living fully. And for the first time, it’s within reach for those who need it most. When innovation meets accessibility, lives change. This is technology for humanity. Follow me, Dr. Martha Boeckenfeld for more stories of tech that matters. ♻️ Share with your network to learn more about how simple innovation can change people's live. #TechForGood #Innovation #Healthcare

Not ads. Not influencers. This is what builds a D2C brand. 8 years ago, when Varun Alagh and I launched Mamaearth, we weren’t the biggest brand. We didn’t have endless budgets or massive influencer deals. What we had was intent. We replied to every DM ourselves. Took feedback personally. And obsessed over what one customer was trying to say, not how many followed us. I myself talked to over 3,000 mothers to understand what they want in a baby product. That’s what most people miss about D2C: The consumer doesn’t just buy your product. They buy your intent. 🔹They notice when you make changes based on their reviews. 🔹They remember how fast you responded when they had a concern. 🔹They talk about your brand when you listen to them like a person, not a number. The edge in D2C isn’t speed or scale. While those are important too, what tops the list is how real your relationship with your consumer feels. If you're in the D2C space, don’t chase virality before you’ve built trust. And don’t confuse transactions with loyalty. What’s one lesson that’s shaped how you show up for your consumer? #Entrepreneurship #MondayMotivation #LeadershipLesson #D2C

On this fine Friday, allow this VC to pitch you a business idea born out of personal frustration. I’m a dedicated DAU (only because we don’t yet track Hourly Active Users), and yet these models barely know me. At this point, I feel like a traveling salesman from the 1800s, lugging a little suitcase of context from one AI to the next: “Here are my preferences. My priorities. My professional history. Please remember me.” And I do it. Because when you feed these tools the right context, they’re remarkable. But that potential - so close you can practically taste it - remains just out of reach. Instead, we’re stuck in a Groundhog Day loop of contextless first dates with memoryless machines. The intelligence is there. It’s the memory that fails you. What we need is a Personal Memory Vault: 🔐 Secure, encrypted, private by default 🔄 Portable across models - plug-and-play with any LLM, agent, or assistant 🧩 Composable and modular - you choose what to share, with whom, and for how long 📜 Versioned and auditable - full transparency into how and when your data is used We can build: 🔑 APIs that give developers secure, permissioned access to memory bundles - like professional context, health profiles, or travel preferences 📈 A user dashboard to track which apps have access, what they know, and when they’ve used it 💵 A freemium model for consumers - and licensing options for apps or agents that want to deliver memory-powered experiences Think: a personal OS that updates passively through your interactions. Plaid, but for context - storing a growing, contextual understanding of your life. Yes, it’s about making AI more useful. But it’s more about control. We need the memory layer to be model agnostic and platform independent, because Big Tech’s next move is obvious: offer “personalized memory” - and use it to lock you in tighter than ever. Sam Altman has already said he wants ChatGPT to remember your whole life. Look, I’m an OpenAI fan. But I reserve the right to change my mind the moment Google DeepMind, Anthropic, or xAI drops something better. What I don’t want is to spend my weekends migrating my digital soul like it’s an IBM mainframe in 1986. I don’t want a hundred context engines. I don’t want to be platform-loyal out of sunk-cost guilt. I just want to stop going on first dates with my own data. I’m just a girl, standing in front of an AI, asking it to remember her. 💔

I've been using Clean Architecture for 6+ years. Here’s why I think it’s amazing. 👇 The biggest pain in enterprise systems? A lack of structure. Every project reinvents the wheel. Every team builds layers differently. And knowledge doesn’t transfer between systems. But there’s a proven way to fix this. It’s called Clean Architecture. It’s not about how many projects you create. It’s not about fancy patterns. ✅ It’s about the direction of dependencies. Inner layers (domain, app) define abstractions. Outer layers (infra, presentation) implement those abstractions. Never the other way around. That’s it. That’s the rule. You can package this as: - Layers (domain, app, infra, web) - Vertical slices (grouped per feature) - Components (layers + vertical slices) They all work — if you follow the rule. What are the benefits? - Modular code - Clear separation of concerns - Easy-to-test business logic - Faster onboarding - Loosely coupled components Clean Architecture has helped me ship excellent products. And I’ll keep using it because it works. Want to simplify your development process? Grab my free Clean Architecture template here: https://lnkd.in/eDgfyWKB