<?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 Muhammad Yasir on Medium]]></title>
        <description><![CDATA[Stories by Muhammad Yasir on Medium]]></description>
        <link>https://medium.com/@yasirdesigner96?source=rss-2bec0f05e691------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*duupjSGO_WrIfFPMi-CWEQ.png</url>
            <title>Stories by Muhammad Yasir on Medium</title>
            <link>https://medium.com/@yasirdesigner96?source=rss-2bec0f05e691------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 23:05:07 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@yasirdesigner96/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[User Goals, Needs, and Context: Where AI Product Design Starts]]></title>
            <link>https://yasirdesigner96.medium.com/user-goals-needs-and-context-where-ai-product-design-starts-e40e754a9eec?source=rss-2bec0f05e691------2</link>
            <guid isPermaLink="false">https://medium.com/p/e40e754a9eec</guid>
            <category><![CDATA[ux-research]]></category>
            <category><![CDATA[ui-design]]></category>
            <category><![CDATA[ux-design]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[user-needs]]></category>
            <dc:creator><![CDATA[Muhammad Yasir]]></dc:creator>
            <pubDate>Sat, 03 Oct 2026 12:02:01 GMT</pubDate>
            <atom:updated>2026-10-03T12:02:01.662Z</atom:updated>
            <content:encoded><![CDATA[<h4>AI product design starts with understanding what people want to achieve, what they need along the way, and the situation in which they use a product. Those details give you a reason for every design decision.</h4><figure><img alt="This image explains the user’s goals, needs and context with AI." src="https://cdn-images-1.medium.com/max/1024/1*6SLSfUXuyWgRUkTtd-qaiA.jpeg" /><figcaption>This image explains the user’s goals, needs, and context with AI</figcaption></figure><p>Think about applying for your dream job. You find an opening, click Apply, and discover a four-page form asking for your education, experience, and work history.</p><p>You fill it out because you want the job. The form is something you have to get through.</p><p>That’s the difference between a task and a goal. It’s also a useful place to start before asking AI to design anything.</p><h3>User goals: understand the outcome behind the action</h3><p>A user goal is the outcome someone wants to achieve. A task is an action they take to reach it.</p><p>When someone opens Foodpanda, ordering food is part of the process. Their goal might be to get dinner delivered before their next meeting or feed their family without cooking that evening.</p><p>The distinction changes what you pay attention to.</p><p>If you focus on checkout, you might make the payment steps shorter. If you understand that the person needs food before a meeting, you’ll also look at delivery estimates and whether they can make a confident choice before paying.</p><p>Both matter. The goal explains why.</p><p>The job application example works the same way. Making the form easier to fill out helps, but the applicant may also need to understand the role, prepare the right documents, and know that their application reached the company.</p><p><strong>During an interview, ask people about a recent experience. What were they trying to achieve? What did they do? Where did they struggle?</strong></p><p>Give them room to explain. You’re trying to understand the reason behind the action.</p><h3>User needs: find what helps people reach the goal</h3><p>User needs describe what must be in place for someone to achieve their goal.</p><p>Take a person paying an electricity bill through a banking app. They need to check the amount, confirm the bill details, complete the payment, and understand what happened afterwards.</p><p>A successful transaction matters. So does being able to recognise that it succeeded.</p><p>Users may describe these needs by asking for a feature. Someone might ask for a button to download a receipt. Before designing the button, find out why they want it.</p><p>Do they need a record for themselves? Are they sharing proof with a family member? Do they feel unsure that the payment went through?</p><p>Nielsen Norman Group’s guidance on <a href="https://www.nngroup.com/articles/user-need-statements/">user need statements</a> makes this distinction useful: describe the need behind the proposed solution.</p><p>For a first review of your design, consider five areas:</p><ul><li><strong>Functional needs: What must the product help the person do?</strong></li><li><strong>Emotional needs: What would help them feel confident or reassured?</strong></li><li><strong>Cognitive needs: What information must they understand?</strong></li><li><strong>Accessibility needs: What would help them use the product with their abilities and assistive technology?</strong></li><li><strong>Social needs: What must they share or coordinate with other people?</strong></li></ul><p>Use these as prompts to examine the experience. They aren’t a fixed list that explains every user.</p><p>Watch what people do, too. Imagine someone finishes a payment and immediately takes a screenshot. That could point to a need for proof, but you still need to ask what they’ll use the screenshot for.</p><p>A pause at the payment button deserves the same care. The person might be worried about trust, checking the amount, or trying to understand a fee.</p><p>The pause is something you observed. The reason is something you need to investigate.</p><h3>Context of use: design for the situation</h3><p>Context of use means the conditions surrounding someone when they use your product.</p><p><strong>Where are they? Which device are they using? How much time do they have? Is their connection stable? Is something distracting them?</strong></p><p>Picture the same person using a banking app at home and then outdoors. At home, they have time to read the details. Outdoors, sunlight makes the screen harder to see, and they’re trying to finish before leaving.</p><p>Their goal has not changed. Their situation has.</p><p>That gives you specific design questions. Can they read the payment amount? Can they check the recipient easily? If the connection drops, will they know whether the payment succeeded?</p><p>Write down these conditions before designing. They help you decide what to test and what information deserves attention on the screen.</p><h3>Where AI helps with product design</h3><p>Once you have research material, AI can help organise it into something easier to review. It can suggest themes, summarise notes, and draft possible need statements.</p><p>The findings still need checking. Nielsen Norman Group’s <a href="https://www.nngroup.com/articles/ai-powered-tools-limitations/">evaluation of AI research tools</a> highlights the risk of generated recommendations that aren’t grounded in the supplied research.</p><p>Suppose you have interview notes about expensive online purchases. Several people describe uncertainty after placing an order.</p><p>You could ask AI to identify possible goals and needs, with the relevant passage beside each suggestion. Then read those passages yourself.</p><p>Knowing when an order will arrive might support a wider goal, such as receiving a laptop before starting a new job. A useful analysis should keep those relationships visible.</p><p>The same approach works with needs. AI can group comments about unclear payment status, confusing instructions, or difficult-to-read information. You can then check whether the comments describe the same problem before turning them into a design decision.</p><p>Keep the evidence close.</p><h3>Give AI the context behind the screen</h3><p>A request such as <em>Design a banking app</em> leaves many decisions open. It does not explain who uses the app, what they need, or what makes the task difficult.</p><p>A more useful brief describes the situation:</p><blockquote><em>Explore a mobile bill-payment flow for people who may use the app outdoors with an unstable connection. Make the amount and recipient easy to check. Show clear payment status and a receipt they can revisit. List the assumptions that need testing.</em></blockquote><p>This is an example brief. For a real project, those details should come from research, with assumptions clearly identified.</p><p>Nielsen Norman Group’s guidance on <a href="https://www.nngroup.com/articles/ux-context-design/">using UX knowledge as AI context</a> recommends giving AI access to user research and the reasoning behind design decisions.</p><p>That’s how I’d approach the next design task: write down the user’s goal, the needs supporting it, and the conditions of use. Give AI that context, then review the proposed screen against the same evidence.</p><p>You’ll have a clearer basis for deciding what belongs in the design.</p><p>If you want to see this in detail, it is all in the video; you can check it out there.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2Feq2KZ48a1Ew%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3Deq2KZ48a1Ew&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2Feq2KZ48a1Ew%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/a391d5a96266106c3253439b90fbe6a3/href">https://medium.com/media/a391d5a96266106c3253439b90fbe6a3/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e40e754a9eec" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Human-Centered Design and User-Centered Design: From Everyday Life to AI]]></title>
            <link>https://yasirdesigner96.medium.com/human-centered-design-and-user-centered-design-from-everyday-life-to-ai-8941c9204e14?source=rss-2bec0f05e691------2</link>
            <guid isPermaLink="false">https://medium.com/p/8941c9204e14</guid>
            <category><![CDATA[human-centered-design]]></category>
            <category><![CDATA[ai-ux-design]]></category>
            <category><![CDATA[everyday-life]]></category>
            <category><![CDATA[user-centered-design]]></category>
            <category><![CDATA[ux-design]]></category>
            <dc:creator><![CDATA[Muhammad Yasir]]></dc:creator>
            <pubDate>Sat, 26 Sep 2026 16:23:56 GMT</pubDate>
            <atom:updated>2026-09-26T16:23:56.766Z</atom:updated>
            <content:encoded><![CDATA[<h4><em>Understanding people starts with noticing the small things they do every day.</em></h4><figure><img alt="Iceberg diagram showing an AI layer above water with product conventions, layout variations, UI patterns, visual style, and interface options. Below water, human-centered design covers environment, people, culture, systems, accessibility, and social impact. User-centered design covers goals, needs, emotions, behaviour, and pain points." src="https://cdn-images-1.medium.com/max/1024/1*rgF8XR9cBJJuGGSTJ--5yQ.jpeg" /><figcaption>The visible interface rests on a deeper understanding of people. This simplified view shows HCD and UCD as overlapping foundations for designing with AI.</figcaption></figure><p>Imagine leaving a supermarket with shopping bags in both hands. You reach the exit, and the door opens automatically. You walk through without stopping.</p><p>You probably don’t think about the door again.</p><p>But imagine having to pull it open. You’d stop, move the bags onto one wrist, reach for the handle, and hold the door while squeezing through.</p><p>Same person. Same goal. Different amount of effort.</p><p>Some useful design lessons are hiding in moments like this. Before we talk about human-centered design, user-centered design, or AI, let’s spend a little time noticing them.</p><h3>Start with the everyday things you barely notice</h3><p>Think about a lift. You press a floor button, it lights up, and you know the lift has registered your choice. When you reach the floor, the doors open.</p><p>Or picture a payment app. After you send money, a confirmation appears. You can stop wondering whether the payment went through.</p><p>These small responses answer questions at the moment you need an answer. Did the system register my action? Has the task finished? What should I do next?</p><p>Now imagine those responses disappearing. The lift button doesn’t light up. The payment screen stays unchanged after you tap send.</p><p>You might press again. You might wait. You might ask someone for help.</p><p>The task hasn’t changed, but you’re doing extra work to understand what’s happening.</p><p>That’s what I want you to start noticing as a designer: the effort a situation asks from someone. Sometimes that effort is physical, like opening a heavy door. Sometimes it’s mental, like figuring out whether a payment succeeded.</p><blockquote>For a small exercise, choose one everyday task and observe someone completing it, with their permission. Notice where they hesitate, repeat an action, or find a workaround. Ask what they expected to happen.</blockquote><p>Hold off on designing a solution. First, describe the difficulty clearly.</p><p>That habit gives us a practical starting point for understanding HCD and UCD.</p><h3>Human-centered design: understand people in their situation</h3><p>Human-centered design, or HCD, starts with understanding people, what they’re trying to do, and the conditions around them. It brings that understanding into design decisions and improves solutions through evaluation. <a href="https://www.nist.gov/itl/iad/human-centered-technologies/human-factors-human-centered-design">NIST’s overview of HCD</a> describes these principles.</p><p>Let’s return to the supermarket entrance.</p><p>An automatic door could help someone carrying bags. But we still need to consider someone approaching slowly with a walking aid, a parent pushing a stroller, or a worker moving a trolley.</p><p>Does the door detect them in time? Does it stay open long enough? What happens if it stops working?</p><p>The entrance needs to work in those situations too. Calling it automatic doesn’t answer those questions.</p><figure><img alt="A shopper carrying bags, an older adult using a walking frame, and a worker pushing a trolley of boxes approach an open automatic supermarket entrance." src="https://cdn-images-1.medium.com/max/1024/1*y7ppwetNOMl2NAjKyfQmpw.png" /><figcaption>One entrance, different needs. Human-centered design asks how the same space works for everyone who uses it.</figcaption></figure><p>Now imagine you’ve been asked to improve the arrival experience at a hospital.</p><p>You could begin by following what happens from the entrance to reception. A patient might struggle to find the right department. A receptionist might spend time repeating directions. A visitor might need a route without stairs.</p><p>At this point, we haven’t decided to build an app. The solution could involve clearer signs, different instructions, changes to the reception process, or a digital service.</p><p>We need to understand the difficulty before choosing.</p><p>HCD encourages that attention to people and context. It also asks us to consider who else is affected by our decisions.</p><h3>User-centered design: understand the person completing the task</h3><p>User-centered design, or UCD, focuses design work on users’ needs, goals, and tasks. To make that concrete, imagine you’re designing a banking app for someone who wants to send money.</p><p>You need to understand the decisions they make along the way.</p><p>Can they find the payment option? Identify the correct recipient? Check the amount before sending? Understand whether the transfer is pending or complete?</p><p>Each question points to something you can explore and test.</p><p>Consider a hypothetical usability test. You expect people to struggle with finding the payment button. Instead, they find it immediately but hesitate before confirming.</p><p>The recipient information is too limited. They aren’t sure they’ve selected the right person.</p><p>A larger payment button wouldn’t resolve that uncertainty. You’d need to explore how to make the recipient clearer, then test whether the change helps.</p><p>That’s the practical value of involving users: their experience can change your understanding of the problem.</p><h3>How HCD and UCD fit together</h3><p>It’s tempting to treat HCD and UCD as two separate methods with a firm boundary between them. In practice, they overlap.</p><p><a href="https://cdn.standards.iteh.ai/samples/77520/33cc3cf17e5e4f1894ee872ed4f54fb6/ISO-9241-210-2019.pdf">ISO 9241–210</a> notes that the terms are often used interchangeably. It uses <em>human-centered</em> to emphasise effects on stakeholders beyond those typically considered users.</p><p>For learning, I find it useful to shift between two views:</p><ol><li>Look closely at the person trying to complete a task.</li><li>Look around that task at the other people, conditions, and consequences involved.</li></ol><p>In the banking example, you could study how the sender checks a payment. Then widen your attention to the recipient and the support staff who handle problems.</p><p>These are connected parts of the experience.</p><p>Both approaches involve learning and revising. Research, prototypes, and testing help you do that, but completing those activities doesn’t automatically mean the design works. Iteration and user involvement are central to the <a href="https://www.nist.gov/itl/iad/human-centered-technologies/human-factors-human-centered-design">human-centered approach</a>.</p><p>Choose your next activity based on what you still need to understand. Sometimes that means making a rough prototype. Sometimes it means going back to observe the task more carefully.</p><h3>Now bring that thinking to AI</h3><p>Keep the same banking example, but imagine adding an AI assistant.</p><p>In this hypothetical feature, someone types a request to send money to a person they’ve paid before. The assistant prepares the transfer, but selects the wrong contact with a similar name.</p><p>The design now needs to help the sender notice and correct that mistake before money moves.</p><p>I’d want the proposed recipient and amount clearly visible, an easy way to change them, and explicit confirmation before sending. I’d also want to investigate what happens when the person cannot resolve the problem themselves.</p><p>That connects the direct interaction with the wider service. What information would support staff need? How would the recipient be affected? Who helps the customer recover?</p><p>Google’s <a href="https://pair.withgoogle.com/guidebook-v2/patterns">People + AI Guidebook</a> recommends planning for incorrect predictions, considering their consequences, and giving people ways to take control when automation fails.</p><p>This is where the foundations become useful. HCD and UCD give you questions to bring to an AI feature before you become attached to what it can generate.</p><p>What is the person trying to accomplish? Where would automation help? What do they need to check? What happens if the system gets something wrong?</p><p>If you use an AI tool to create a prototype, include those situations. Show the mistaken selection. Build the correction path. Test whether someone understands what will happen next.</p><p>A working demonstration gives you something to investigate. You still need evidence that it helps people complete the task.</p><h3>Continue with the full lesson</h3><p>That’s why I’m starting this course with everyday situations and design foundations. I want you to practise noticing effort, understanding its cause, and testing whether your idea actually helps.</p><p>The automatic door is a small place to begin. Pay attention to what it makes easier, and who might still struggle.</p><p>The full video, <em>AI Product Design Lecture 2: Human-Centered Design &amp; User-Centered Design</em>, is on YouTube.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2Ff66CR3RMNDc%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3Df66CR3RMNDc&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2Ff66CR3RMNDc%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/e9795f156cacd37588dd09f96db5a230/href">https://medium.com/media/e9795f156cacd37588dd09f96db5a230/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8941c9204e14" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[UI, UX, and Product Design: What’s the Difference?]]></title>
            <link>https://yasirdesigner96.medium.com/ui-ux-and-product-design-whats-the-difference-b8d57c60e4ad?source=rss-2bec0f05e691------2</link>
            <guid isPermaLink="false">https://medium.com/p/b8d57c60e4ad</guid>
            <category><![CDATA[ux-design]]></category>
            <category><![CDATA[design]]></category>
            <category><![CDATA[product-design]]></category>
            <category><![CDATA[user-experience]]></category>
            <category><![CDATA[ui-design]]></category>
            <dc:creator><![CDATA[Muhammad Yasir]]></dc:creator>
            <pubDate>Tue, 15 Sep 2026 12:08:28 GMT</pubDate>
            <atom:updated>2026-09-15T12:08:28.393Z</atom:updated>
            <content:encoded><![CDATA[<h4>A beginner-friendly guide to understanding what designers actually solve</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*R7p2o3CFH0AT9fcZz-JpOQ.jpeg" /><figcaption>Created by AI judged by human</figcaption></figure><p>Design is the intentional process of helping people achieve a goal or solve a problem. Visual appeal matters, but a beautiful product can still fail if people can’t understand or use it.</p><p>This distinction becomes clearer when we separate three connected fields: UI design, UX design, and product design.</p><h3>Design is problem-solving</h3><p>People often associate design with decoration. They think about colours, typography, illustrations, and polished screens.</p><p>Those elements are part of design, but they aren’t its full purpose.</p><p>Consider a door with a stylish handle and a clean finish. If you can’t tell whether to push or pull it, the design has failed at its main job. The door may look good, but it doesn’t clearly communicate how someone should use it.</p><p>This type of confusing door is commonly called a Norman door, named after designer and author Don Norman. It shows a simple principle: people shouldn’t have to stop and solve a puzzle before using an everyday object.</p><figure><img alt="A person struggling to open a glass door, illustrating how unclear design can confuse users." src="https://cdn-images-1.medium.com/max/1024/0*TbzXK0DVyXU92skp.jpg" /><figcaption><em>Screenshot from “Your Gateway to UX Design: Norman Doors” by William Hudson, Interaction Design Foundation.</em></figcaption></figure><p>Good design makes the intended action clear.</p><h3>What is UX design?</h3><p>User experience, or UX, covers a person’s complete experience with a product, service, or company.</p><p>According to Nielsen Norman Group, user experience includes the interactions, perceptions, and emotions that form a person’s relationship with a product or service. It begins before someone uses the main interface and can continue long after a purchase. <a href="https://www.nngroup.com/articles/what-is-user-experience/">Nielsen Norman Group explains </a>UX more broadly.</p><p>Imagine that you’re buying a smartphone from an online store. Your experience includes:</p><ul><li>Finding the phone</li><li>Understanding its price and specifications</li><li>Reading customer reviews</li><li>Adding it to your basket</li><li>Making the payment</li><li>Tracking the order</li><li>Receiving the phone</li><li>Contacting support or requesting a return</li></ul><p>The checkout screen is only one part of this experience. Search results, product information, delivery updates, customer support, and return policies also shape how you feel about the store.</p><p>UX designers study this complete experience. They try to find where people become confused, frustrated, uncertain, or unable to complete their goals.</p><p>Their questions might include:</p><ul><li>Can customers find the right product?</li><li>Do they understand the price?</li><li>Does the checkout process feel safe?</li><li>Can they recover from a payment error?</li><li>Do they know what will happen after ordering?</li><li>Can they get help when something goes wrong?</li></ul><p>UX design focuses on how an experience works and feels from the user’s point of view.</p><h3>What is UI design?</h3><p>User interface design, or UI design, focuses on the parts of a digital product that people see and interact with directly.</p><p>These parts can include:</p><ul><li>Buttons</li><li>Text fields</li><li>Icons</li><li>Colours</li><li>Typography</li><li>Spacing</li><li>Menus</li><li>Screen layouts</li><li>Navigation controls</li></ul><p>A UI designer decides how these elements should look, communicate, and respond. A primary button, for example, should be easy to notice and should make its purpose clear.</p><p>UI design has a narrower scope than UX design. It shapes the interface through which much of the experience happens, but it doesn’t cover every part of that experience.</p><p>A payment button may look clear and polished. The wider experience can still be poor if the payment fails without an explanation or the customer can’t find a way to try again.</p><p>That is why UI and UX need to work together.</p><h3>What is product design?</h3><p>Product design considers the whole product: what it should do, how it should work, who needs it, and whether it creates value for users and the business.</p><p>Product designers connect user needs with product choices and business goals. Their work may include research, problem definition, interaction design, prototyping, testing, feature planning, and collaboration with engineers and product managers.</p><p>Figma describes product design as work that connects user research, interaction design, and business strategy. It also considers whether the team can build and support the proposed solution. <a href="https://www.figma.com/resource-library/what-is-product-design/">Read Figma’s guide to product design</a>.</p><p>Before designing a feature, a product designer may ask:</p><ul><li>Who has this problem?</li><li>How important is the problem?</li><li>Why is it happening?</li><li>What should we build?</li><li>How should the solution work?</li><li>Will people use it?</li><li>Can the engineering team build it?</li><li>Does it support the business?</li><li>How will we know whether it worked?</li></ul><p>These questions stop teams from spending months polishing a feature that people don’t need.</p><h3>UI vs. UX vs. product design</h3><p>The simplest way to remember the difference is:</p><p><strong>UI design asks:</strong> How should the interface look and communicate?</p><p><strong>UX design asks:</strong> How should the experience work and feel?</p><p><strong>Product design asks:</strong> Why are we solving this problem, what should we build, how should it work, and will it create value?</p><p>The three fields overlap. Their exact responsibilities can also change between companies. Some organisations use “UX designer” and “product designer” for very similar roles.</p><p>The difference is mainly one of scope.</p><p>UI design concentrates on the interface. UX design considers the person’s experience across a task or relationship. Product design connects that experience to product direction, technical limits, and business needs.</p><h3>An example from a ride-hailing app</h3><p>Imagine someone opens a ride-hailing app because they need to get home.</p><p>The person’s goal sounds simple, but each design discipline looks at it from a different angle.</p><h4>The UI design perspective</h4><p>A UI designer may consider:</p><ul><li>Where the ride request button should appear</li><li>How the map should look</li><li>Which car icon to use</li><li>How to display the pickup point</li><li>What information belongs in the bottom sheet</li><li>Which typography and colours make the screen clear</li></ul><p>These decisions affect what the interface communicates and how easily the user can interact with it.</p><h4>The UX design perspective</h4><p>A UX designer may ask:</p><ul><li>Can the rider enter the correct destination?</li><li>Can the rider understand the estimated price?</li><li>Is the pickup point clear?</li><li>What happens if the driver cancels?</li><li>Can the rider contact the driver?</li><li>Does the app explain delays?</li><li>Can the rider recover from an error?</li></ul><p>These questions cover the flow from opening the app to reaching the destination.</p><h4>The product design perspective</h4><p>A product designer may examine wider questions:</p><ul><li>Why are drivers cancelling rides?</li><li>Is a new feature the right response?</li><li>What do riders need at this stage?</li><li>What do drivers need?</li><li>How might a pricing decision affect both groups?</li><li>Can the team build the proposed solution?</li><li>Which business goal does the solution support?</li><li>What evidence would show that the change improved completed rides?</li></ul><p>This view connects interface choices and user behaviour with technical feasibility and business results.</p><p>A new button might improve the interface without solving the real problem. If drivers cancel because pickup locations are unclear, the team may need to improve location accuracy or communication instead.</p><p>Finding the right problem comes first.</p><h3>Why beginners should learn these differences</h3><p>New designers often start with tools. They learn how to create frames, choose colours, build components, and make prototypes.</p><p>Those skills are useful. But tools can’t decide which problem deserves attention.</p><p>A strong designer learns to move between different levels of thinking:</p><ol><li>Understand the person and the problem.</li><li>Study the complete experience.</li><li>Decide what the product should do.</li><li>Design the interaction and interface.</li><li>Test whether the solution works.</li><li>Learn from the result and improve it.</li></ol><p>AI can speed up parts of this work. It can help organise research notes, generate early ideas, compare possible flows, or create a first draft of an interface.</p><p>AI still needs direction. If the problem is unclear, it can help you produce the wrong solution much faster.</p><p>That is why the foundation matters.</p><h3>A practical way to review your next design</h3><p>When you work on a screen or feature, pause and ask three sets of questions.</p><h4>Check the interface</h4><ul><li>Are the actions easy to find?</li><li>Are labels clear?</li><li>Does the visual hierarchy show what matters?</li><li>Can people understand the screen without extra instructions?</li></ul><h4>Check the experience</h4><ul><li>Can users complete their goal?</li><li>Where might they become confused?</li><li>What happens when something goes wrong?</li><li>What happens before and after this screen?</li></ul><h4>Check the product decision</h4><ul><li>Which user problem are we solving?</li><li>What evidence shows that the problem exists?</li><li>Why should the team solve it now?</li><li>Can the team build and maintain the solution?</li><li>How will the solution help users and the business?</li><li>What will we measure after release?</li></ul><p>If you can answer all three groups, you’re thinking beyond the screen. You’re beginning to think like a product designer.</p><h3>Keep learning</h3><p>These resources can help you build a stronger foundation:</p><ul><li><a href="https://www.nngroup.com/articles/what-is-user-experience/">What Is User Experience? by Nielsen Norman Group</a> explains why UX extends beyond one screen or interaction.</li><li><a href="https://www.figma.com/resource-library/what-is-product-design/">What Is Product Design? by Figma</a> covers the relationship between user research, interaction design, product decisions, and business strategy.</li><li><a href="https://www.figma.com/resource-library/what-is-ux-design/">What Is UX Design? by Figma</a> introduces the responsibilities and process of UX design.</li></ul><p>The next topic in this course is human-centred design and user-centred design. We’ll look at what these approaches share, where they differ, and how they shape product decisions.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2F0xL2jzDL8NQ%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3D0xL2jzDL8NQ&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2F0xL2jzDL8NQ%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/4ce9bfeda1d1d5afadd6dca9dc504800/href">https://medium.com/media/4ce9bfeda1d1d5afadd6dca9dc504800/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b8d57c60e4ad" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI UX Design: What Happens When AI Gets It Wrong?]]></title>
            <link>https://yasirdesigner96.medium.com/ai-ux-design-what-happens-when-ai-gets-it-wrong-45e3af7ed309?source=rss-2bec0f05e691------2</link>
            <guid isPermaLink="false">https://medium.com/p/45e3af7ed309</guid>
            <category><![CDATA[user-experience]]></category>
            <category><![CDATA[generative-ai-tools]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ux-design]]></category>
            <category><![CDATA[product-design]]></category>
            <dc:creator><![CDATA[Muhammad Yasir]]></dc:creator>
            <pubDate>Mon, 07 Sep 2026 19:00:36 GMT</pubDate>
            <atom:updated>2026-09-07T19:00:36.710Z</atom:updated>
            <content:encoded><![CDATA[<h4>Good AI UX design helps people understand what AI can do, judge its answers, and recover from mistakes. The experience needs to support all three.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_BO73W_CKIff283BGDM3YA.png" /><figcaption>AI UX should help people check information, correct mistakes, and recover.</figcaption></figure><p>The question at the centre of my <a href="https://youtube.com/shorts/Lc7Fu21iXbU?si=sn_ziVt3gkd8EfQO">Day 6 reel</a> was simple: <strong>“When AI gives the wrong answer, what should the user be able to do?”</strong></p><p>A prompt box gives someone a place to start. A loading state tells them something is happening. But once the answer appears, a different set of design questions begins.</p><p>Can the user check it? Correct it? Ask for a different approach? Continue their work if the answer is unhelpful?</p><p><a href="https://pair.withgoogle.com/guidebook-v2/chapters">Google PAIR</a>, <a href="https://www.microsoft.com/en-us/haxtoolkit/">Microsoft HAX</a>, <a href="https://developer.apple.com/design/human-interface-guidelines/generative-ai?changes=_4">Apple’s generative AI guidelines</a>, and <a href="https://www.ibm.com/design/research/">IBM’s explainability research</a> help us think through these questions. Here are four lessons from those resources, with documented ChatGPT examples where they make the design clearer.</p><h3>1. Make sure AI solves a real problem</h3><p>Before deciding where the AI button belongs, explain why the feature belongs in the product.</p><p><a href="https://developer.apple.com/design/human-interface-guidelines/generative-ai?changes=_4">Apple</a> recommends using generative AI where it offers clear value, such as saving time, improving communication, or supporting creativity. <a href="https://pair.withgoogle.com/guidebook-v2/chapters">Google PAIR </a>also begins with user needs and defining success.</p><p>ChatGPT provides a familiar example. Its documented uses include drafting and revising text, summarizing files, and comparing options. Each activity connects AI to a task someone wants to complete. <a href="https://learn.chatgpt.com/docs/use-chatgpt">OpenAI’s guide to using ChatGPT</a></p><p>For designers, the next question is whether the feature makes that task easier overall.</p><p>A generated draft still needs review. A summary still needs to preserve important information. A comparison needs a clear basis.</p><p>When evaluating an AI feature, include the effort needed to check and correct its output. That gives you a more useful picture than measuring how quickly it generates a response.</p><blockquote><strong>Ask during design: </strong>What can the user accomplish more easily because this feature exists?</blockquote><h3>2. Help users understand what the AI can, and cannot, do</h3><p>A <strong>mental model</strong> is a person’s understanding of how something works.</p><p>In an AI product, that understanding affects what people ask for and how much they rely on the answer.</p><p>Google PAIR recommends introducing capabilities and limitations so users can form realistic expectations. <a href="https://pair.withgoogle.com/guidebook-v2/chapter/mental-models/">PAIR: Mental Models</a></p><p>Consider ChatGPT’s web search. When ChatGPT uses web search on the web, search results and citations appear in the conversation. That gives the user information about the sources behind the response. <a href="https://learn.chatgpt.com/docs/web-search">OpenAI’s web search documentation</a></p><figure><img alt="ChatGPT web-search response about AI UX design, showing source citations and links to resources including Google PAIR and Microsoft HAX." src="https://cdn-images-1.medium.com/max/1024/1*-t4FbPfoXShAV8X5_Tp_CQ.png" /><figcaption>ChatGPT displays citations alongside its response, giving users a way to inspect sources and check whether they support the claims.</figcaption></figure><p>The design lesson is to make relevant system behavior visible. Someone checking a factual answer needs to understand whether sources are available and how to inspect them.</p><p>A citation gives them a route to evidence. They still need to check whether the source supports the statement.</p><p><a href="https://www.microsoft.com/en-us/haxtoolkit/">Microsoft HAX</a> extends this thinking across four stages of an experience:</p><figure><img alt="Four stages of AI interaction: explain capabilities and performance at the start, provide relevant guidance during use, support correction when AI is wrong, and communicate system changes over time." src="https://cdn-images-1.medium.com/max/1024/1*G4th-1Rog0ieVoO5BEG1IQ.png" /><figcaption>A simplified summary of Microsoft HAX’s four interaction stages: getting started, during use, when AI is wrong, and over time.</figcaption></figure><p>Expectations need attention throughout the product, especially when capabilities change. <a href="https://www.microsoft.com/en-us/haxtoolkit/ai-guidelines/">Microsoft HAX Guidelines</a></p><blockquote><strong>Ask during design:</strong> What might a user assume here, and does the product support that assumption?</blockquote><h3>3. Make correction part of the main experience</h3><p>An AI answer can look finished while still being wrong.</p><p>Google PAIR describes errors that neither the system nor the user recognizes. This means recovery design must consider mistakes inside the output, alongside visible failures such as a response that never loads. <a href="https://pair.withgoogle.com/guidebook-v2/chapter/errors-failing/">PAIR: Errors + Graceful Failure</a></p><p>ChatGPT’s conversational interaction offers one practical form of control: users can review a result, add context, correct the direction, and ask for changes through follow-up messages. OpenAI documents this as part of the normal workflow. <a href="https://learn.chatgpt.com/docs/use-chatgpt">OpenAI’s guide to using ChatGPT</a></p><p>For designers, this shows why the first response should not be treated as the end of the interaction.</p><p>But the recovery path also needs to match the problem:</p><figure><img alt="Four AI problems and next steps: simplify unclear wording, add missing context and request a revision, inspect evidence for questionable facts, and change direction or use another method for unsuitable output." src="https://cdn-images-1.medium.com/max/1024/1*-jqIQ4SnBuNpD-wuWzgN5w.png" /><figcaption>Match the recovery path to the problem. Revising wording, adding context, and checking facts require different actions.</figcaption></figure><p>These are design recommendations. Asking for a revision alone does not establish that the revised answer is accurate.</p><h4>Help people answer the question that matters</h4><p><a href="https://research.ibm.com/publications/questioning-the-ai-informing-design-practices-for-explainable-ai-user-experiences">IBM’s <em>Questioning the AI</em> research</a> explores how to design explanations around users’ information needs. Its question bank helps designers identify what people need to understand about an AI system.</p><p>That suggests a useful approach: start with the user’s question before choosing an explanation.</p><blockquote>“Where did this information come from?” needs access to evidence.</blockquote><blockquote>“Can I change this?” needs a clear control.</blockquote><blockquote>“What happens next?” needs an explanation of the action and its consequences.</blockquote><blockquote><strong>Ask during design:</strong> Can the user fix the problem without losing useful work or starting over?</blockquote><h3>4. Separate different answers from incorrect answers</h3><p><a href="https://developer.apple.com/design/human-interface-guidelines/generative-ai?changes=_4">Generative AI</a> can produce different results from the same input. Apple explicitly identifies this variability as a design consideration.</p><p>Different results can be useful when someone is exploring wording or ideas. They become a problem when facts change without justification.</p><p>ChatGPT supports asking for alternative approaches and refining a response through follow-up instructions. This makes exploration part of the interaction. <a href="https://learn.chatgpt.com/docs/use-chatgpt">OpenAI’s guide to using ChatGPT</a></p><p>For a designer, the important distinction is between <strong>changing the presentation</strong> and <strong>correcting the information</strong>.</p><p>A shorter explanation may be preferable without being more accurate. A more confident answer may still need checking.</p><p>Design controls accordingly. Where users work through multiple versions, consider how they will compare results, keep useful changes, and return to earlier work. These are decisions to make for your product, rather than capabilities to assume every AI interface provides.</p><blockquote><strong>Ask during design:</strong> Is the user exploring a preference, correcting a mistake, or trying to establish a fact?</blockquote><h3>Test what happens after the answer appears</h3><p>A successful generation tells you that the system produced something. Testing should also examine whether people can use that output responsibly and finish their task.</p><p>Include imperfect results in your prototype tests. Then observe:</p><ul><li>Do users notice important errors?</li><li>Can they find information that helps them check the result?</li><li>Can they correct a problem without losing their progress?</li><li>Do they understand what a retry or revision will change?</li><li>Can they continue when the AI is unhelpful?</li></ul><p>Bias and privacy also need specific tests relevant to the product. Apple recommends testing across diverse people to identify stereotypes and designing AI experiences with privacy in mind. <a href="https://developer.apple.com/design/human-interface-guidelines/generative-ai?changes=_4">Apple’s generative AI guidelines</a></p><p>Interface design is one part of this work. Designers need to collaborate with engineering, research, and other relevant teams to understand failures and decide how the product should respond.</p><h3>The question to bring into your next design review</h3><p><strong>“If this answer is wrong, what can the user do next?”</strong></p><p>Look beyond whether an edit button or warning exists. Follow the task through: noticing the problem, checking the information, making a correction, and continuing.</p><p>That is where principles such as trust and control become concrete product behavior.</p><blockquote>A useful AI experience helps people understand the answer, decide what to do with it, and recover when it falls short.</blockquote><p><em>This article expands Day 6 of my Professional UX/UI Design Resources series. Watch the companion YouTube reel for a short introduction to these four principles.</em></p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FLc7Fu21iXbU%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fshorts%2FLc7Fu21iXbU%3Fsi%3D4kEx_M0kgi_du9qw&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FLc7Fu21iXbU%2Fhq2.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="640" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/5083e0c720e54d36f52425b655c29458/href">https://medium.com/media/5083e0c720e54d36f52425b655c29458/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=45e3af7ed309" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[6 Free Anthropic AI Courses That Changed How I Use AI]]></title>
            <link>https://yasirdesigner96.medium.com/6-free-anthropic-ai-courses-that-changed-how-i-use-ai-08a57c72811b?source=rss-2bec0f05e691------2</link>
            <guid isPermaLink="false">https://medium.com/p/08a57c72811b</guid>
            <category><![CDATA[claude-ai]]></category>
            <category><![CDATA[ux-design]]></category>
            <category><![CDATA[product-design]]></category>
            <category><![CDATA[ai-course]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Muhammad Yasir]]></dc:creator>
            <pubDate>Sat, 15 Aug 2026 15:36:30 GMT</pubDate>
            <atom:updated>2026-08-15T15:36:30.914Z</atom:updated>
            <content:encoded><![CDATA[<h4><em>I completed all six courses in two days. They helped me understand AI beyond prompt formulas, from hallucinations to Agent Skills and subagents.</em></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*U_7d8vVM1_Sj3YGfCmu-yg.png" /></figure><p>The six free Anthropic AI courses I completed taught me one central lesson: <strong>using AI properly requires more than writing better prompts</strong>.</p><p>A good prompt can improve an answer. It cannot guarantee that the answer is accurate, supported by evidence, or suitable for the task.</p><p>I already use Claude Code, Claude Design, Claude Cowork, and VS Code. However, these courses helped me understand the concepts behind those tools. They clarified what AI can do, why it sometimes fails, and what the human must still review.</p><h3>Which Anthropic AI courses should beginners take?</h3><p>Beginners should start with AI Fluency and Capabilities and Limitations, then progress to Builders, Agent Skills, Subagents, and Small Businesses.</p><p>Here is the learning order I followed:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7-AO5diw4jzW1JAncfqWug.png" /><figcaption>The six-course Anthropic learning path I followed, from AI Fluency and AI limitations to Agent Skills, Subagents, and practical business applications.</figcaption></figure><p>Anthropic’s catalog lists approximately four hours of video across these six courses. Exercises, quizzes, and practice can increase the completion time. <a href="https://claude.com/resources/courses">View the current Anthropic course catalog</a></p><h3>Why aren’t better prompts enough?</h3><p>Better prompts improve communication with AI, but they cannot remove hallucinations or replace human judgment.</p><p>AI doesn’t think like a person. A language model predicts likely continuations based on its training, instructions, available context, and the text already generated.</p><p>Anthropic’s <strong>AI Capabilities and Limitations</strong> course explains four properties behind model behaviour:</p><ul><li>Next-token prediction</li><li>Knowledge</li><li>Working memory</li><li>Steerability</li></ul><p>These properties help explain why an AI answer can sound confident while still being incorrect. <a href="https://anthropic.skilljar.com/ai-capabilities-and-limitations">View AI Capabilities and Limitations</a></p><p>Anthropic’s interpretability research says language-model training requires models to keep predicting the next word. Training can reduce hallucinations, but it cannot remove them completely. <a href="https://www.anthropic.com/news/tracing-thoughts-language-model">Read Anthropic’s hallucination research</a></p><blockquote>OpenAI researchers have also argued that common evaluation methods can reward guessing instead of admitting uncertainty. The result can be a plausible answer that isn’t true. <a href="https://openai.com/index/why-language-models-hallucinate/">Read the OpenAI research</a></blockquote><p>For a product designer, a hallucination could appear as an invented interview quote, an unsupported accessibility claim, or a competitor feature that doesn’t exist.</p><p>The solution isn’t to stop using AI. The solution is to know what must be checked.</p><h3>What do the six Anthropic courses teach?</h3><h4>1. AI Fluency: Framework &amp; Foundations</h4><p><strong>AI Fluency: Framework &amp; Foundations teaches a system for deciding when, why, and how to work with AI.</strong></p><p>Anthropic developed the framework with Professors Joseph Feller and Rick Dakan. It organizes human-AI collaboration around four competencies, called the <strong>4Ds</strong>:</p><ul><li><strong>Delegation:</strong> Decide which work AI should complete.</li><li><strong>Description:</strong> Give AI the context and instructions it needs.</li><li><strong>Discernment:</strong> Evaluate the accuracy and usefulness of its response.</li><li><strong>Diligence:</strong> Use AI responsibly and take ownership of the result.</li></ul><p>Most prompting advice focuses only on Description. The AI Fluency Framework shows that Description is only one part of working with AI.</p><blockquote>Product designers also need to decide whether AI should perform a task, support the task, or remain outside the process. <a href="https://anthropic.skilljar.com/ai-fluency-framework-foundations">View AI Fluency: Framework &amp; Foundations</a></blockquote><h4>2. AI Capabilities and Limitations</h4><p><strong>AI Capabilities and Limitations gives beginners a working model of how generative AI behaves and why unexpected outputs occur.</strong></p><p>This 15-minute course was one of the most important parts of the sequence for me. It connects AI behaviour with practical concepts such as knowledge, working memory, and steerability.</p><p>Understanding those limitations changes how you evaluate an answer.</p><blockquote>For example, a designer can ask Claude to analyze interview transcripts. Claude can help group repeated themes, but the designer should compare every conclusion with the original transcripts.</blockquote><p>AI can support research synthesis. It shouldn’t become the source of the research.</p><h4>3. AI Fluency for Builders</h4><p><strong>AI Fluency for Builders applies the 4D Framework to the full journey from a customer problem to a shipped product.</strong></p><p>The course is designed for product people, engineers, and makers. Its curriculum includes Delegation, Description, discernment for code, and discernment for user experience.</p><p>This makes it the most directly relevant course for product designers.</p><p>The framework can support several parts of product work:</p><ul><li>Framing a customer problem</li><li>Exploring ideas and alternative solutions</li><li>Reviewing UX writing</li><li>Questioning product assumptions</li><li>Evaluating a prototype</li><li>Preparing product documentation</li></ul><blockquote>The human still owns the decision. AI can produce options, organize information, and challenge assumptions. <a href="https://anthropic.skilljar.com/ai-fluency-for-builders">View AI Fluency for Builders</a></blockquote><h4>4. Introduction to Agent Skills</h4><p><strong>Introduction to Agent Skills teaches you to turn repeated instructions into reusable procedures for Claude Code.</strong></p><p>A prompt normally handles one request. An Agent Skill stores instructions that Claude can apply whenever a matching task appears.</p><p>The course includes six lectures covering Skill creation, configuration, sharing, and troubleshooting. It also explains how Skills differ from prompts, subagents, hooks, and project instructions. <a href="https://anthropic.skilljar.com/introduction-to-agent-skills">View Introduction to Agent Skills</a></p><p>A product designer could create reusable Skills for:</p><ul><li>UX-writing standards</li><li>Heuristic evaluations</li><li>Accessibility reviews</li><li>Research-synthesis rules</li><li>Design-system documentation</li><li>Product requirement reviews</li></ul><p>Agent Skills require more setup than an ordinary Claude conversation. The benefit is consistency across repeated tasks.</p><h4>5. Introduction to Subagents</h4><p><strong>Introduction to Subagents explains how separate AI assistants can complete focused tasks and report their findings to a main agent.</strong></p><p>A subagent receives its own context, instructions, and tool permissions. This separation can help when one project contains several distinct tasks.</p><p>For example, a product designer could configure:</p><ul><li>One subagent to review accessibility</li><li>One subagent to challenge research conclusions</li><li>One subagent to examine UX copy</li><li>One subagent to find inconsistencies between requirements and screens</li></ul><p>These are possible product-design applications, not examples promised by the course.</p><p>The official course focuses on Claude Code and includes four lectures about creating and using subagents. <a href="https://anthropic.skilljar.com/introduction-to-subagents">View Introduction to Subagents</a></p><h4>6. AI Fluency for Small Businesses</h4><p><strong>AI Fluency for Small Businesses connects AI use with business decisions, responsibilities, and everyday company workflows.</strong></p><p>Product designers need to understand more than interface design. A design decision can affect customer support, operations, acquisition, pricing, and retention.</p><p>The small-business course adds that wider context.</p><blockquote>A useful question is no longer just, “Can AI help us build this feature?” The better question is, “Does this feature solve a customer problem and support the business?”</blockquote><p><a href="https://anthropic.skilljar.com/ai-fluency-for-small-businesses">View AI Fluency for Small Businesses</a></p><p>Want the quick visual version? Watch my YouTube Reel about these six Anthropic courses, and subscribe for more practical AI and product design resources.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FGJE49ZvzhfU%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fshorts%2FGJE49ZvzhfU%3Fsi%3DhoLfrXMTjTjF3fcm&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FGJE49ZvzhfU%2Fhq2.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="640" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/ae5fd1aa0a6dd22a291d65bd951de16f/href">https://medium.com/media/ae5fd1aa0a6dd22a291d65bd951de16f/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=08a57c72811b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Use AI as a Design Partner, Not as Your Design Judgment]]></title>
            <link>https://yasirdesigner96.medium.com/use-ai-as-a-design-partner-not-as-your-design-judgment-4eee8d6767d6?source=rss-2bec0f05e691------2</link>
            <guid isPermaLink="false">https://medium.com/p/4eee8d6767d6</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ux-design]]></category>
            <category><![CDATA[design-thinking]]></category>
            <category><![CDATA[ux-research]]></category>
            <category><![CDATA[product-design]]></category>
            <dc:creator><![CDATA[Muhammad Yasir]]></dc:creator>
            <pubDate>Mon, 10 Aug 2026 11:00:20 GMT</pubDate>
            <atom:updated>2026-08-10T11:00:20.211Z</atom:updated>
            <content:encoded><![CDATA[<h4><em>AI can help product designers move faster, but speed means little if the final product does not solve a real user problem.</em></h4><figure><img alt="A product designer selects one interface from multiple AI-generated options and connects the chosen design to a diverse group of users." src="https://cdn-images-1.medium.com/max/1024/1*uH14yeIThQKSoY2ypAPF3Q.png" /><figcaption><em>AI can generate possibilities, but designers remain responsible for deciding which ideas should reach users.</em></figcaption></figure><p>AI should support product designers, not make product decisions for them. A tool can generate dozens of screens in minutes, but it cannot take responsibility for whether those screens solve the right problem.</p><p>The debate around AI and design has become unnecessarily binary. One group insists that AI will never replace designers. Another claims that designers are already unnecessary because developers, product managers, and clients can generate interfaces themselves.</p><p>I believe both sides are asking the wrong question.</p><p>The important question is not whether AI can produce a design. It is whether that design is based on real user needs, informed by context, tested with people, and connected to a business goal.</p><h3>The confusing reality for product designers</h3><p>In my experience, many Pakistani organizations still do not have a mature design process. In some companies, even creating and validating a design in Figma is not treated as a necessary stage of product development.</p><blockquote>I once spoke with an engineer who described a familiar situation. A client generated a design using AI and sent it directly to the development team. The engineers built it, but when the product manager reviewed the result, the response was essentially: “What have you built?”</blockquote><p>That question revealed the real problem.</p><p>The team had an interface, but it did not have a shared understanding of the user problem, product requirements, or reasons behind the design decisions. AI had helped produce an output, but nobody had created alignment.</p><p>This is where designers face a strange contradiction.</p><blockquote>When designers use AI extensively, some organizations assume that AI is doing all their work. When designers avoid AI, the same organizations warn them that they will soon be replaced.</blockquote><p>Designers are judged on whether they use AI or not.</p><p>The better way to evaluate a designer is to examine the thinking behind the work. Can the designer explain the user problem, research insights, constraints, alternatives, tradeoffs, and testing evidence? If the answer is yes, AI is simply one of the tools used in the process.</p><h3>A generated interface is not automatically a good product</h3><p>Developers sometimes tell designers, “We can replace you. We can generate the design with AI and build it quickly through vibe coding.”</p><p>They may be right about producing something quickly. AI can generate an interface, and vibe coding can turn that interface into a working prototype faster than traditional workflows.</p><p>But I always return to one question:</p><blockquote><strong>Will the AI-generated and vibe-coded product solve the user’s problem?</strong></blockquote><p>A business problem is not solved merely because a product has been built. The business benefits only when people understand the product, use it, trust it, and receive meaningful value from it.</p><p>AI can generate a checkout screen, but it does not automatically know why customers abandon the checkout process. It can generate an onboarding flow, but it does not automatically understand what makes a new user confused or anxious. It can create a polished dashboard, but visual polish cannot tell a team whether the dashboard supports an important user decision.</p><p>Building faster is valuable only when the team is building the right thing.</p><h3>AI should expand a designer’s thinking</h3><p>AI is extremely helpful when it has a clear role. I use it as a design partner, a tool that helps me organize information, explore possibilities, and challenge my thinking.</p><p>For example, AI can help product designers:</p><ul><li>Organize notes from user interviews.</li><li>Identify possible patterns across multiple interviews.</li><li>Turn observations into early product ideas.</li><li>Generate several design directions from validated user needs.</li><li>Compare alternative flows and interaction models.</li><li>Identify missing states, edge cases, or unanswered questions.</li><li>Review a user interface for hierarchy, clarity, consistency, and accessibility.</li><li>Prepare a prototype for testing more quickly.</li></ul><p>These activities can save time, but the designer must still examine the output.</p><blockquote>AI may group two interview comments into the same theme even though they came from different contexts. It may present a weak assumption as a confident recommendation. It may generate a beautiful interface that ignores technical, cultural, or behavioral constraints.</blockquote><p>The designer’s responsibility is to determine what is relevant, what is misleading, and what requires more research.</p><h3>Research must come before artificial certainty</h3><p>AI can help analyze user research, but it should not invent the research.</p><figure><img alt="Circular product design workflow showing six stages: user research, AI synthesis, designer judgment, prototype, user testing, and refinement." src="https://cdn-images-1.medium.com/max/1024/1*i9WWtnWcvPlUWuykZfLYQg.png" /><figcaption>A responsible AI-assisted design process combines user research, AI synthesis, designer judgment, prototyping, user testing, and continuous refinement.</figcaption></figure><p>There is an important difference between asking AI to find patterns in genuine interview notes and asking it to imagine what users might say. The first activity supports research. The second creates artificial certainty.</p><p><strong>A responsible AI-assisted design process could look like this:</strong></p><ol><li>Conduct interviews or observe real users.</li><li>Collect genuine statements, behaviors, frustrations, and goals.</li><li>Ask AI to organize the evidence and suggest possible patterns.</li><li>Review those patterns against the original research.</li><li>Generate several ideas based on the validated patterns.</li><li>Select promising ideas using human judgment and product constraints.</li><li>Test the resulting design with multiple users.</li><li>Refine the product based on what the tests reveal.</li></ol><p>The final design does not emerge when AI generates a screen. It emerges after the team has tested its assumptions and learned from real people.</p><h3>AI can be a mentor, but it cannot own the decision</h3><p>Designers can also use AI as a mentor, especially when reviewing user interfaces.</p><blockquote>I may ask AI to critique information hierarchy, explore alternative layouts, identify unclear labels, or suggest questions that I should ask during usability testing. This gives me another perspective and helps me notice possibilities that I may have missed.</blockquote><p>However, I do not blindly apply every recommendation.</p><p>AI tools regularly remind users that their responses may contain mistakes and should be checked. That may be one of the most ignored warnings on the internet.</p><p>AI can provide an answer without understanding the consequences of that answer. A designer cannot avoid accountability by saying, “The AI suggested it.”</p><p>The person making the final decision must still own the result.</p><h3>How I use AI in design and content creation</h3><p>As a product designer, I use AI for research organization, pattern exploration, ideation, design variations, and interface critique.</p><blockquote>As a content creator, I use it for script ideas, early hooks, structure, visual storyboards, and planning. It helps me move from an unorganized thought to something I can examine and improve.</blockquote><p>But the final decision remains mine.</p><p>I do not publish an idea simply because AI generated it. I consider whether it reflects my experience, sounds like my voice, and offers something useful to the audience.</p><blockquote>This matters because AI makes it easy for everyone to produce similar content. If every LinkedIn post uses the same structure and every video begins with the same type of generated hook, the content may be polished without being memorable.</blockquote><p>The same principle applies to articles. Generic AI writing gives readers and search engines little reason to choose one article over hundreds of similar ones. Human experience, clear judgment, and a defensible point of view create the difference.</p><h3>What I would never delegate completely to AI</h3><p>AI can assist with many activities, but I would not use it to replace responsibility in areas such as:</p><ul><li>Claiming to understand users without conducting research.</li><li>Making the final product decision.</li><li>Copying another designer’s work.</li><li>Generating fake interviews, findings, or user quotations.</li><li>Presenting unverified information as fact.</li><li>Replacing professional judgment and accountability.</li><li>Publishing generic social posts without adding a personal perspective.</li><li>Using generated video hooks without considering the audience or message.</li><li>Treating AI output as unquestioned medical or other high-stakes judgment.</li></ul><p>The medical comparison is useful because assistance and accountability are different roles. AI may help a professional organize information or notice a possibility, but the qualified person must evaluate the context and remain accountable for the decision.</p><p>Product designers should apply the same standard to their work.</p><h3>Give AI a clear job description</h3><p>AI is best used for planning, organizing, exploring, and challenging, not for blindly approving final work.</p><blockquote>Product designers do not need to reject AI to protect their careers. They need to become better at using AI while strengthening the abilities that AI cannot responsibly own: research, judgment, empathy, prioritization, facilitation, testing, and accountability.</blockquote><p>For working and aspiring product designers, my advice is simple:</p><p>Learn AI. Use it in your daily work. Let it help you explore more ideas, organize research, and move faster.</p><p>But do not depend on it to think for you.</p><p>A strong designer does not compete with AI on the number of screens produced. A strong designer uses AI to explore possibilities, then applies human judgment to decide which possibility deserves to become a product.</p><blockquote><strong>Use AI as your design partner. Never allow it to become your design judgment.</strong></blockquote><p>Here is the short video version of this article:</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2F9WkiMjbTmk0%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fshorts%2F9WkiMjbTmk0%3Fsi%3DR3vyXAhcCkyDQPtF&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2F9WkiMjbTmk0%2Fhq2.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="640" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/05babea7393d536ae7dc348826d875ef/href">https://medium.com/media/05babea7393d536ae7dc348826d875ef/href</a></iframe><p><strong>Visit my YouTube channel for more videos about product design, UX, and AI.</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4eee8d6767d6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[GPT-5.6 Sol, Terra, and Luna: I Tested All Three Against Real Work]]></title>
            <link>https://yasirdesigner96.medium.com/gpt-5-6-sol-terra-and-luna-i-tested-all-three-against-real-work-9274a5482239?source=rss-2bec0f05e691------2</link>
            <guid isPermaLink="false">https://medium.com/p/9274a5482239</guid>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[ai-in-business]]></category>
            <category><![CDATA[app-development]]></category>
            <category><![CDATA[generative-ai]]></category>
            <category><![CDATA[ux-design]]></category>
            <dc:creator><![CDATA[Muhammad Yasir]]></dc:creator>
            <pubDate>Mon, 20 Jul 2026 13:11:12 GMT</pubDate>
            <atom:updated>2026-07-20T13:11:12.545Z</atom:updated>
            <content:encoded><![CDATA[<h4>GPT-5.6 ships as three specialized models: <strong>Sol</strong> for building software, <strong>Terra</strong> for everyday tasks, and <strong>Luna</strong> for deep document analysis. I ran each one on the job it was built for: a playable UX game, fintech onboarding microcopy, and 15 months of startup financials. This is what actually held up.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MEUkC36rp0tni9NVpl_iSQ.png" /><figcaption>Thumbnail reading “GPT-5.6 Sol Built This Game,” showing the sun, Earth, and moon orbs beside the UX Match match-3 puzzle board</figcaption></figure><p>Most coverage of a model release stops at the benchmark chart. I did the opposite; I gave each variant a real, messy task and watched where it earned its keep. The full walkthrough is in the video below; this article is the written record with the prompts and results.</p><p>Video Link: <a href="https://youtu.be/fT4EDQgNKIg?si=IzVeE7PPE0MIUiiA">https://youtu.be/fT4EDQgNKIg?si=IzVeE7PPE0MIUiiA</a></p><h3>What are GPT-5.6 Sol, Terra, and Luna?</h3><p><a href="https://openai.com/index/gpt-5-6/">GPT-5.6 </a>is a three-model family: <strong>Sol</strong> handles code generation and app building, <strong>Terra</strong> handles fast everyday work, and <strong>Luna</strong> handles long-document reasoning and pattern-finding. Each is tuned for a different job rather than being one general model.</p><p>Picking the wrong variant is the most common mistake. Sending a 15-month financial spreadsheet to Terra wastes its speed advantage; sending onboarding microcopy to Luna wastes its long-context depth. The split only pays off when the task matches the model.</p><ul><li><strong>GPT-5.6 Sol: C</strong>ode and product builds (games, apps, functional prototypes).</li><li><strong>GPT-5.6 Terra: </strong>Everyday tasks that reward speed (copy, flows, quick drafts).</li><li><strong>GPT-5.6 Luna: </strong>Long-document analysis, financial data, pattern detection.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_lxAdBJwSTQ3M1HfxJyUyQ.png" /><figcaption>Three GPT-5.6 model cards: Sol (sun) for expensive complex work, Terra (Earth) a balanced medium model, Luna (Mars) for high-volume finance and long-document work.</figcaption></figure><h3>GPT-5.6 vs Fable 5: the three benchmark charts</h3><p>Across the three benchmarks I compared, <strong>GPT-5.6 trails Fable 5 by one point on raw quality but wins clearly on speed and cost per task.</strong> The one-point gap is the headline nobody should over-read.</p><p>Here is what each chart showed:</p><p><strong>Quality: </strong>GPT-5.6 lands <strong>one point below </strong><a href="https://www.anthropic.com/news/redeploying-fable-5"><strong>Fable 5</strong></a>. A single point rarely changes the output you ship; it changes the leaderboard.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IXNePImvANduGUF9D87EAA.png" /><figcaption><em>Bar chart of AI model Intelligence Index scores: Claude Fable 5 leads at 60, GPT-5.6 Sol close behind at 59, Grok 4.5 at 54, GLM 5.2 at 51, Gemini 3.5 Flash at 50.</em></figcaption></figure><p><strong>Speed: </strong>GPT-5.6 runs <strong>noticeably faster than Fable 5</strong>, which is what you feel on every real task, not just the first.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Z82A3DB4EJnLthjp19F-qw.png" /><figcaption>Bar chart of AI model output speed in tokens per second: GLM 5.2 fastest at 206.2, Gemini 3.5 Flash at 161.6, Grok 4.5 at 120.7, GPT-5.6 Sol at 68.9, Claude Fable 5 slowest at 60.4.</figcaption></figure><p><strong>Cost per task</strong>: GPT-5.6 comes in <strong>lower on cost per task</strong>, so the speed advantage compounds at volume.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*v8WvKe5IbkiYzuwl1_gqSg.png" /><figcaption>Bar chart of AI model API cost per 1M tokens, input and output: Claude Fable 5 most expensive at $10/$50, GPT-5.6 Sol at $5/$30, Grok 4.5 at $2/$6, GLM 5.2 is the cheapest at $1.4/$4.4, Gemini 3.5 Flash at $1.5/$9.</figcaption></figure><p>The practical read: if you are running many tasks a day, a one-point quality gap is a bad reason to pay more and wait longer. Fable 5 wins the benchmark; GPT-5.6 wins the workflow.</p><h3>Test 1: GPT-5.6 Luna on 15 months of startup financials</h3><p><strong>Luna analyzed 15 months of SaaS financial data and returned insights tied to real numbers, not generic advice.</strong> I asked it to act as a senior financial analyst and judge the startup’s health, unit economics, and path to profitability.</p><p><strong>The prompt forced specificity on purpose:</strong></p><blockquote>I’m sharing 15 months of financial data for an early-stage SaaS startup. Act as a senior financial analyst and analyze this data carefully. Every insight must use real numbers from the data, no generic advice.</blockquote><blockquote>First, give me an honest health assessment of this startup. Look at revenue growth, burn rate, cash balance, and runway together. Find the two most dangerous months where the company almost ran out of cash and explain what went wrong.</blockquote><blockquote>Then analyze the unit economics. Is CAC vs LTV healthy? Which months had the worst churn and why? Based on the current growth trend, predict the exact month this startup reaches profitability.</blockquote><blockquote>Finally give three specific recommendations, each one directly tied to a pattern you found in this data.</blockquote><p>Where Luna earned its slot in the lineup:</p><ul><li>It pulled the <strong>two lowest-runway months</strong> out of the sheet instead of hand-waving about “watching burn.”</li><li>It reads <strong>revenue growth, burn rate, cash balance, and runway together</strong> rather than in isolation.</li><li>It held the <strong>full 15-month context</strong> long enough to spot churn patterns and tie each recommendation back to a specific trend in the data.</li></ul><p>This is the task on which the one-point quality gap matters least; the value came from pattern-finding across a long document, and Luna’s long-context handling is exactly what that needs.</p><h3>Test 2: GPT-5.6 Terra on fintech onboarding microcopy</h3><p><strong>Terra wrote complete microcopy for a five-screen fintech onboarding flow, button labels, helper text, error messages, and empty states, in one fast pass.</strong> This is the everyday-task category, and speed is the whole point.</p><p><strong>The prompt:</strong></p><blockquote>“I am working on a fintech mobile app onboarding flow, 5 screens. Write microcopy for each screen: button labels, helper text, error messages, and empty state messages. Keep tone friendly and simple.”</blockquote><p>Microcopy is deceptively hard; it has to be short, on-brand, and cover edge cases like errors and empty states that most drafts skip. Terra handled all five screens without needing the deep reasoning budget Luna spends, which is why it’s the right tool here: the job rewards a fast, clean first draft over slow perfection.</p><h3>Test 3: GPT-5.6 Sol built a playable UX learning game</h3><p><strong>Sol built UX Match, a Candy Crush-style match-3 game that teaches UX concepts, with five playable levels, as a fully working React and Vite app.</strong> This was the hardest test and the clearest showcase of the Sol variant.</p><p>UX Match is a match-3 puzzle where players swap tiles representing design elements, wireframes, color swatches, typography, personas, accessibility, prototypes, and every valid match pops a one-line UX tip. It’s built to teach designers while they play.</p><p><strong>The prompt:</strong></p><blockquote>UX Match is a Candy Crush-style match-3 puzzle game that teaches UX/UI design concepts through gameplay. Instead of candies, players match tiles representing design elements, wireframes, color swatches, typography, personas, accessibility, and prototypes. Every match shows a real one-line UX tip related to the matched tile. Built for designers to learn while playing, and to demonstrate AI builder capability live on localhost.</blockquote><blockquote>The game includes a match-3 grid, cascade and combo scoring, special tiles, move and score goals per level, UX tip popups after every match, five levels with increasing difficulty, and a reset or retry option at any point without losing overall progress.</blockquote><blockquote>Player lands on home screen and taps Play to load Level 1. During gameplay, player selects a tile, selects an adjacent tile, and swaps, valid matches clear tiles and cascade new ones while score updates. When the target score is reached, a win modal appears with a Next Level button. When moves run out before reaching the goal, a lose modal appears with a Retry button.</blockquote><blockquote>Home screen shows the game title, short instructions, and a Play button. Game board screen shows the 8x8 grid with a top bar displaying score, moves remaining, current goal, and a pause button. UX Tip Toast appears after each match and auto-dismisses in 2 to 3 seconds. Win Modal shows score summary and 2 to 3 tips learned. Lose Modal shows an out of moves message with a Retry button. Pause Menu includes Resume, Restart Level, and Back to Home options.</blockquote><blockquote>Background is soft neutral #F7F7F5. Tile colors are distinct and flat, slate blue for wireframes, coral for color swatches, amber for typography, teal for personas, purple for accessibility, green for prototypes. Typography uses Inter or system-ui, bold for scores and headers, regular for tips. Tiles are rounded with soft shadows and flat SVG or emoji- style icons. Matched tiles pulse and glow before disappearing. New tiles drop in with a smooth animation. Score increments animate as a count-up.</blockquote><blockquote>Board state is an 8x8 array of 6 tile types with no pre-existing matches at start. Match detection finds horizontal or vertical runs of 3 or more identical tiles. Matching 4 in a line creates a striped tile that clears a full row or column when matched again. Matching 5 creates a bomb tile that clears all tiles of that type on the board. Move counter decrements only on a valid swap that produces a match, invalid swaps snap back with no move deducted. Score is 10 points per tile cleared with a 1.5x multiplier per cascade chain level. Level configs are stored as a simple array, Level 1 is 20 moves and 500 points, Level 5 is 12 moves and 2000 points. UX tips are a hardcoded array per tile type, selected randomly, never repeating the same tip twice in a row.</blockquote><blockquote>If no valid moves remain on the board, auto-shuffle all tiles and show a brief Reshuffling message. If a player swaps two tiles with no resulting match, animate the swap back to original position with no move cost. If multiple simultaneous matches result from one swap, resolve all matches, cascade, and sum combo scores correctly. If the board is fully cleared during a cascade, refill all empty cells from the top with new random tiles. If a player taps or swaps rapidly, disable all input while an animation or cascade is resolving to prevent state bugs…</blockquote><blockquote>Build as a full React and Vite app with Tailwind CSS. File structure must be clean, /src/ components, /src/hooks, /src/data, and /src/App.jsx. Use React state only, useState or useReducer, no external state libraries. No user accounts, backend, database, sound effects, leaderboard, multiplayer, monetization, or mobile packaging in this version. Fully local and browser only. Must run with npm install then npm run dev. Every feature must be fully functional not a visual shell. Matches, cascades, scoring, and level progression must all work end to end.What Sol had to get right, end to end:</blockquote><ul><li><strong>An 8×8 board</strong> with no pre-existing matches at the start and match detection for runs of three or more.</li><li><strong>Special tiles </strong>a striped tile from a 4-match that clears a row or column, and a bomb tile from a 5-match that clears every tile of that type.</li><li><strong>Cascade scoring </strong>10 points per tile cleared with a 1.5× multiplier per cascade chain level, combos summed correctly.</li><li><strong>Five levels</strong> of rising difficulty (Level 1: 20 moves / 500 points → Level 5: 12 moves / 2000 points), plus win and lose modals with Next Level and Retry.</li><li><strong>Edge cases</strong> auto-shuffle when no valid moves remain, invalid swaps snap back with no move cost, and input is locked during animations to prevent state bugs.</li><li><strong>Clean architecture, </strong>React and Vite with Tailwind, a /src/components, /src/hooks, /src/data structure, React state only, running from npm install then npm run dev.</li></ul><p>The bar I set was “every feature fully functional, not a visual shell”: matches, cascades, scoring, and level progression all working. That’s the real test of a code model: not whether it renders a grid, but whether the game plays.</p><p><strong>Play UX Match:</strong> <a href="https://ux-match.vercel.app/">https://ux-match.vercel.app/</a></p><h3>Which GPT-5.6 model should you use?</h3><p><strong>Use Sol to build software, Terra for fast everyday tasks, and Luna for long-document analysis.</strong> Match the variant to the job, and the one-point gap versus Fable 5 stops mattering.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*77PsC7TwuWozFz6klE4JRw.png" /><figcaption>Table matching GPT-5.6 variants to tasks: Sol for building apps and games, Terra for copywriting and quick drafts, Luna for financial data and long documents.</figcaption></figure><h3>Watch the full test</h3><p>I walk through all three models, the benchmark charts, and every prompt live in the video, including the UX Match game running on localhost. <a href="https://www.youtube.com/@yasir.design"><strong>YouTube video link</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9274a5482239" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Codex for Designers: Build Real UI Without Figma First]]></title>
            <link>https://yasirdesigner96.medium.com/codex-for-designers-build-real-ui-without-figma-first-ab89565998d0?source=rss-2bec0f05e691------2</link>
            <guid isPermaLink="false">https://medium.com/p/ab89565998d0</guid>
            <category><![CDATA[openai-codex]]></category>
            <category><![CDATA[ai-tools]]></category>
            <category><![CDATA[design-systems]]></category>
            <category><![CDATA[ui-design]]></category>
            <category><![CDATA[product-design]]></category>
            <dc:creator><![CDATA[Muhammad Yasir]]></dc:creator>
            <pubDate>Mon, 06 Jul 2026 14:01:28 GMT</pubDate>
            <atom:updated>2026-07-06T14:01:28.424Z</atom:updated>
            <content:encoded><![CDATA[<h4>Yes, a designer can use Codex to build a real UI from the terminal without first touching Figma. I gave Codex one prompt, and it generated a full mental-health app interface in about 30 minutes. Then I built a gym landing page the same way.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hzyRpyFBH-jjFQMOt5FZLw.png" /></figure><p><em>Here’s the part most designers don’t want to hear:</em></p><blockquote>The ones who can’t prompt an agent are going to fall behind the ones who can. Not someday. This year. The tool that used to belong to developers is now the fastest way I’ve found to go from an idea to a working screen.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JftiRsCNlyRhY9m2DTRkkA.png" /><figcaption>Mental health and Gym landing page UI generated by Codex for designers</figcaption></figure><h3>What is Codex, and why should a designer care?</h3><p>Codex is OpenAI’s coding agent. It lives in your terminal (there’s also an app and a VS Code extension); it reads and writes files, and it runs commands. Developers use it to ship code. But nothing about it is dev-only, and that’s the opening for designers.</p><p>You care because Codex closes the gap between <em>“what the screen should feel like” and “a screen you can actually open.”</em> A designer’s real skill is taste, hierarchy, and knowing what good looks like. Codex handles the mechanical part — the layout code, the spacing, the components. You bring the judgment. It brings the speed.</p><p>Three pieces make Codex genuinely useful for design work.</p><h4>The Codex design stack: Skills, MCP, and Plugins</h4><p><strong>Skills:</strong> Markdown files (a SKILL.md) that teach Codex a repeatable way to do something. Codex loads the full instructions only when a task matches.</p><p><strong>MCP servers: C</strong>onnections that give Codex tools and context from outside your project. <strong>Mobbin’s MCP, for example, feeds Codex real UI patterns to reference while it builds.</strong></p><p><strong>Plugins:</strong> Installable bundles that package skills, app connectors, and MCP servers together.</p><p>Learn those three, and you’ve learned Codex. Everything else is detail. Full docs live in <a href="https://openai.com/academy/skills/">OpenAI’s guides for skills</a>, <a href="https://developers.openai.com/codex/mcp">MCP</a>, and <a href="https://developers.openai.com/codex/plugins">plugins</a> if you want to go deeper.</p><h3>How do you set up Codex for design work?</h3><p>The setup is three moves. I walk through each one on camera in the video, but here’s the shape of it so you know what you’re getting into.</p><h4>1. Connect the Mobbin MCP for live inspiration</h4><p>Mobbin is a library of real app and website UI. Wired in as an MCP server, Codex can pull those patterns while it designs. So instead of describing “a clean onboarding flow” into the void, Codex references how strong products actually structure it. You add MCP servers with codex mcp, and the config is shared across the CLI and the IDE extension, so you set it once.</p><p><strong>Remember:</strong> inspiration, not imitation. <strong>I tell Codex explicitly to use Mobbin for layout quality, hierarchy, and conversion structure, and to never copy a real app.</strong></p><h4>2. Create a skill, so Codex designs your way</h4><p>A skill is just a folder with a SKILL.MD file. Write down how you want Codex to approach design, your spacing rules, your component preferences, your “never do this” list, and it applies them without you re-explaining every time. Codex ships with a built-in creator that asks what the skill does and when it should trigger, then scaffolds the file for you. Front-load the description with clear trigger words so Codex knows when to reach for it.</p><h4>3. Install plugins for the rest</h4><p>Plugins fold together what used to be three separate setups, MCP config, skill install, and app connector, into one click.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*i1_GkT2CnaH3Sdz2PUblag.png" /><figcaption>Codex plugins &amp; skills</figcaption></figure><h3>What’s a design-system .md file, and why does it beat a Figma library here</h3><p>Before either build, I hand Codex one plain text file. It defines the color theme and the design system I want to use. That’s it.</p><p><strong>You can download the design.MD files from</strong><a href="https://getdesign.md/"><strong> getdesign.md</strong></a></p><p>A Figma library is a beautiful reference for a human to translate. A .md design system is an instruction that the machine executes.</p><h3>I designed a mental-health app UI with one prompt. Here’s what happened.</h3><p>I attached the design-system file, told Codex to use Mobbin for inspiration, and asked for a health and fitness app: <strong>sleep, exercise, and daily wellness tracking, with a meditation section and three main areas, Explore, a Meditation Guide, and an AI Coach. Explore surfaces, sleep, exercise, and wellness cards, plus recommended meditation content. The Meditation Guide starts sessions, walks through breathing steps, and plays audio. The AI Coach reviews activity and gives simple guidance.</strong></p><p>One prompt. About 30 minutes. A coherent, three-section app UI came back, real structure, real hierarchy, and the design-system colors applied throughout.</p><h3>Then I built a gym landing page the same way</h3><p>Same method, different job. I asked for a landing page for “Energy Club Gym,” <strong>a bold, premium, energetic site with a dark theme, strong typography, and high-contrast sections. Hero, headline, CTAs, workout programs, membership plans, a trainer section, class schedule, testimonials, features, contact, footer. Reusable components for cards, buttons, pricing, and trainer profiles. Responsive for desktop and mobile. Inspiration from Mobbin landing pages, copy nothing.</strong></p><p>What came back looked like a real product. Rounded cards, bold shadows, a proper conversion structure top to bottom. The reusable-component instruction paid off; the pricing plans and trainer cards were consistent because I’d asked for a system.</p><h3>What actually runs the show: the prompt behind the LLM</h3><p>Codex is powerful, but it’s not a mind reader; it’s a very fast, very literal collaborator. Vague prompt, vague UI. Precise prompt plus a real design-system file plus good references, and it’s genuinely impressive.</p><p>In the video, I break down what’s really happening underneath, what the design.md file is doing, what all those .md files are for, and the prompt logic behind the LLM that turns your words into a screen. Understand that layer, and you stop getting lucky and start getting consistent. That’s the difference between a designer who plays with AI and one who ships with it.</p><h3>Where Codex wins and where it doesn’t</h3><p>Codex wins at speed to a first working screen, at applying a system consistently, and at solving the blank-canvas problem. It’s the fastest zero-to-one I’ve used for UI.</p><p>Where it doesn’t win: taste, restraint, and the last 20%. It’ll happily give you something confident and slightly wrong, and if you can’t tell, you’ll ship slightly wrong. Which loops back to the uncomfortable point. The value isn’t in Codex. It’s the designer driving it.</p><blockquote>So learn to prompt agents. Not because Figma is dead, it isn’t, but because the designers who can direct an agent are about to move at a speed the ones who can’t simply can’t match.</blockquote><p>Want the full setup on screen? I walk through connecting Mobbin’s MCP, creating skills, installing plugins, and building step by step in the video. Watch it, then go build your first screen.</p><p>Video Link: <a href="https://youtu.be/XXsbozWNhK4?si=3-Pazj0ibndRovEx">Codex Tutorial</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BiA0TVybwu-1o7KADIvuAg.png" /><figcaption>YouTube Video</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ab89565998d0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Think Beyond When Designing a Product: UX/UI Case Study of Ticket Booking App]]></title>
            <link>https://yasirdesigner96.medium.com/think-beyond-when-designing-a-product-ux-ui-case-study-of-ticket-booking-app-7c996289cd74?source=rss-2bec0f05e691------2</link>
            <guid isPermaLink="false">https://medium.com/p/7c996289cd74</guid>
            <category><![CDATA[ui-design]]></category>
            <category><![CDATA[apps]]></category>
            <category><![CDATA[ticket-booking]]></category>
            <category><![CDATA[ui]]></category>
            <category><![CDATA[ux-design]]></category>
            <dc:creator><![CDATA[Muhammad Yasir]]></dc:creator>
            <pubDate>Thu, 10 Oct 2024 14:35:12 GMT</pubDate>
            <atom:updated>2024-10-10T14:35:12.057Z</atom:updated>
            <content:encoded><![CDATA[<h4>See how strategic UX/UI design increased conversions and made booking tickets faster and easier for users.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8V5Jh7Xucnv_41iB5N1kVw.png" /><figcaption>Cover image</figcaption></figure><p>A client approaches us with a request to design a ticket-booking app. At first glance, it seems like a straightforward project. But for us, it’s never just about creating something visually appealing. Our approach goes deeper than that. Before we even touch design tools, we start with a conversation.</p><p>We sit down with the client, not to discuss colours or fonts, but to talk about <strong>how their business works</strong>. We ask questions like, “What’s the biggest pain point for your customers?” and “How do you want your app to help users?” It’s about understanding <strong>what drives their business</strong> and how the app can be more than just a booking tool it needs to <strong>solve real problems</strong>.</p><p>Once we understand the business side, we focus on the users. Who are they? What do they need? In the case of a ticket booking app, we know users value <strong>speed</strong> and <strong>ease</strong>. No one wants to jump through hoops to book a ticket, and every extra second spent on a confusing interface means a potential lost customer.</p><p>That’s why, when we design, we think beyond just what looks good. We create seamless experiences, removing obstacles between users and their goals. It’s about making the process so intuitive that users don’t even think about it they book and go.</p><p>This story is at the heart of every project we take on. It’s not just design for the sake of design. It’s a process that considers the business, the users, and how to make both work together effortlessly.</p><h3>The Role of Communication in Designing for Success:</h3><p>Good design starts with clear communication. It’s not just about aesthetics; it’s about understanding the bigger picture. When we begin any project, especially something as user-centric as a ticket booking app, the most critical step is talking with our clients. We want to know how their business works, what their customers struggle with, and where the app fits into their overall strategy.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*6VjqAMLAFWZqELB0" /><figcaption>Unsplash</figcaption></figure><p>This communication helps us align the design with real business goals. For example, if the client’s main concern is reducing booking abandonment rates, our focus shifts to creating a faster, more intuitive booking flow. Without open and continuous communication, the design would risk being just a pretty interface rather than a functional tool that drives results.</p><p>Through clear discussions, we learn about the <strong>key performance indicators (KPIs)</strong> that matter most to the business, whether it’s <strong>higher conversions</strong>, <strong>increased user retention</strong>, or <strong>enhanced customer satisfaction</strong>. Once we understand these priorities, we can design with purpose, ensuring every decision supports the client’s business objectives while keeping the user experience front and center.</p><h3>Meeting Tight Deadlines with the Design Sprint Process</h3><p>When a project has a tight deadline, like delivering a ticket booking app in one month, we rely on the <strong>Design Sprint Process</strong>. This structured approach helps us work quickly without compromising on quality.</p><p>You can learn the design sprint from the <a href="https://designsprintkit.withgoogle.com/methodology/overview">Google Sprint</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HqHDU57MlIA_tRj-Q7VITw.png" /><figcaption>Designed by me</figcaption></figure><p>When building a product, there are several key factors to consider <strong>understanding user needs</strong> and <strong>gathering ongoing feedback</strong> are two of the most critical steps in the process.</p><h4>1. User Interviews: The Foundation of Product Design</h4><p>One of the first steps in designing a product is conducting <strong>user interviews</strong>. These interviews help us understand the users’ problems, goals, and behaviours. By directly speaking with potential users, we can gather insights on what challenges they face and how the product can solve these issues.</p><p>For example, if we’re designing a ticket booking app, user interviews may reveal frustrations with confusing layouts or slow checkout processes. These insights allow us to design a solution that addresses these pain points from the very beginning, ensuring that the product is built around the user’s needs.</p><h4>2. User Feedback: Iterating for Improvement</h4><p>After the initial design phase, gathering <strong>user feedback</strong> is essential for refining the product. Once we have a working prototype or an early version of the product, we present it to users to see how they interact with it. Their feedback helps identify areas for improvement, such as navigation issues or confusing features.</p><p>For instance, in a ticket booking app, user feedback may show that certain buttons are hard to find, or that the payment process feels too complicated. By acting on this feedback, we can make necessary adjustments before final development, ensuring a smooth and enjoyable user experience.</p><p>By prioritizing <strong>user interviews</strong> and <strong>continuous feedback</strong>, we ensure that the product evolves based on real user needs, resulting in a more successful, user-centered design.</p><blockquote>When designing a product, understanding <strong>human psychology</strong> and <strong>digital psychology</strong> plays a critical role in creating a seamless user experience.</blockquote><h4>3. Human Psychology: How People Think and Behave</h4><p>Human psychology helps us understand how people process information, make decisions, and react to visual stimuli. When designing, we consider basic principles like <strong>cognitive load</strong>, which refers to how much mental effort is required to complete a task. If a design is too complex or overwhelming, users may feel frustrated and abandon the process.</p><p>Another key factor is <strong>habit formation</strong>. People are creatures of habit, and they tend to prefer familiar patterns and experiences. By following common design conventions (like placing the “Buy” button in expected locations), we create a sense of comfort and trust.</p><p>Learn UX Psychology from<a href="https://growth.design/"> Growth. Design</a> and<a href="https://builtformars.com/"> build for Mars</a></p><h3>The Importance of a Design System:</h3><p>A <strong>design system</strong> is a set of standards, guidelines, and reusable components that streamline the design and development process. When creating a product, especially at scale, a well-defined design system ensures consistency, efficiency, and scalability.</p><p>For developers, a design system acts as a blueprint that defines how components should be built and implemented. It saves time and reduces ambiguity in translating design to code.</p><ul><li><strong>Reusable Code Components:</strong> Developers can reuse the same components across different parts of the app, avoiding redundancy and speeding up development. Once a component is coded, it can be used multiple times, ensuring consistency in both design and functionality.</li><li><strong>Clear Communication:</strong> A design system provides clear specifications, reducing back-and-forth between designers and developers. Developers know exactly how elements should behave, which cuts down on errors and misunderstandings.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5F1BzZPGjWDypW2xsMEJrg.png" /><figcaption>Design System</figcaption></figure><h3>Getting the Figma File Ready for Developers</h3><p>When we finish the design in Figma, we create a special <strong>“Ready for Development”</strong>. This page helps developers by organizing everything they need, making their job easier and faster.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/247/1*Ibny9LQABo4ciwdWjpdJ7A.png" /><figcaption>Figma</figcaption></figure><h4>1. Organizing the Figma File</h4><p>On this page, we include:</p><ul><li><strong>Design Components:</strong> Things like buttons, icons, and fonts are clearly labelled and easy to find.</li><li><strong>Style Guides:</strong> We list important details like colors, font sizes, and spacing, so everything looks consistent in the app.</li><li><strong>Interactive Elements:</strong> We show how things like hover effects or animations should work, so developers know how elements should behave.</li></ul><h4>2. Meeting with Developers</h4><p>Before the developers start working, we have a <strong>meeting</strong> to go through the Figma file together. In this meeting, we explain:</p><ul><li><strong>Key Design Elements:</strong> We point out important parts of the design and how they should work.</li><li><strong>Animations and Interactions:</strong> We explain how any animations or transitions should behave in the app.</li><li><strong>Using the Design System:</strong> We show developers how to use the design system to speed up their work.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/646/1*NJf1GWnJPhKje-OKXBpkFg.png" /><figcaption>Ready for development</figcaption></figure><h3>Why Visual Appeal Matters in Design</h3><p>Humans are naturally <strong>visual creatures</strong>. We process images much faster than text, which is why a visually appealing design is so powerful. When something looks good, it catches attention, builds trust, and makes the experience more enjoyable.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wVS7qyfCNaszIPIOEpaX1w.png" /><figcaption>Onboarding</figcaption></figure><p>For example, <strong>Dieter Rams</strong>, in his book <em>“Ten Principles for Good Design”</em>, talks about the importance of simplicity and functionality. His designs, which are visually clean and minimal, make it easy for users to understand and interact with a product. Similarly, <strong>Don Norman</strong> in <em>“The Design of Everyday Things”</em> explains how good design makes things not only look good but also <strong>work effortlessly</strong>.</p><blockquote>But visual appeal is not just about looks it’s about clarity. A well-organized design helps everyone involved in the project. For developers, a visually structured design makes it easier to understand how elements should behave. They don’t need to guess where things go or how they work because it’s clear from the design itself.</blockquote><h3>Follow Me for More Insights</h3><p>Stay connected and follow me on social media for more design tips and insights:</p><ul><li><a href="https://dribbble.com/yasirdesigner96">Dribble</a></li><li><a href="https://www.linkedin.com/feed/">LinkedIn</a></li></ul><p>Check out the <strong>full case study</strong> on Behance and view the <strong>complete design</strong> on Figma Community:</p><ul><li><a href="https://www.behance.net/muhammadyasir43"><strong>Behance</strong></a></li><li><a href="https://www.figma.com/design/Vh0qAcIw7szkMtlGrkrydP/Ticket-event-Booking-App-(Community)?node-id=30-115&amp;t=zHDXolR95XLGXcM7-1"><strong>Figma Community</strong></a></li></ul><p>If you’re a designer and want to collaborate, feel free to reach out. Let’s create something amazing together!</p><ul><li><strong>Contact me:</strong> m.yasir.k.2001@gmail.com</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7c996289cd74" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Build Your Own GPT Assistant in ChatGPT-4 Without Coding: A Step-by-Step Guide]]></title>
            <link>https://medium.com/etechviral/how-to-build-your-own-gpt-assistant-in-chatgpt-4-without-coding-a-step-by-step-guide-5e7441448ecc?source=rss-2bec0f05e691------2</link>
            <guid isPermaLink="false">https://medium.com/p/5e7441448ecc</guid>
            <category><![CDATA[psychiatry]]></category>
            <category><![CDATA[openai]]></category>
            <category><![CDATA[chatgpt4]]></category>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[chatbots]]></category>
            <dc:creator><![CDATA[Muhammad Yasir]]></dc:creator>
            <pubDate>Sun, 25 Aug 2024 10:49:59 GMT</pubDate>
            <atom:updated>2025-06-25T07:01:17.985Z</atom:updated>
            <content:encoded><![CDATA[<h4><em>The new feature announced by ChatGPT allows users to customize their own ChatGPT and share it with others without any coding, revolutionizing the way AI is used. This feature is a game-changer in the AI market, enabling easy creation and sharing of customized AI models.</em></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*g948bO6SI2jshcL8R8YGzQ.png" /></figure><p>Creating personalized chatbots with the ability to access and modify features can revolutionize user experience and content creation. Understanding the limitations and configurations of chatbot features is crucial for effective utilization in various tasks.</p><p>With the introduction of GPT-4 by OpenAI, creating your own customized GPT-powered assistant has become easier than ever. If you have a ChatGPT Plus subscription, you can now build a personalized GPT assistant directly within the ChatGPT interface — no coding required! This guide will walk you through the process of setting up and customizing your own GPT, tailored to your specific needs and preferences.</p><h3>What are Custom GPTs in ChatGPT-4?</h3><p>Custom GPTs are personalized versions of OpenAI’s language model, designed to cater to your specific needs. Unlike the default ChatGPT, these custom models can be tailored with unique instructions, behaviours, and integrated tools, making them perfect for specialized tasks or personal preferences.</p><h3>Benefits of Building Your Own GPT</h3><ul><li>Personalization: Create a GPT that understands your unique requirements, preferences, and style.</li><li>Flexibility: Adapt the model for different roles, such as a Psychiatrist, Psychotherapist, UX Designer, customer support, or even an AI Doctor.</li><li>No Coding Required: With a ChatGPT Plus subscription, building and customizing your GPT is straightforward and requires no technical expertise.</li></ul><h3>Subscription Requirements</h3><p>To access the custom GPT feature, you need a ChatGPT Plus subscription. This premium service unlocks ChatGPT-4’s full potential, allowing you to create and tailor your own GPT models.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*j8QVpHQnH5plDX07xrcmVw.png" /><figcaption>ChatGPT</figcaption></figure><h3>Accessing the Custom GPT Feature</h3><ol><li>Log into ChatGPT: Ensure you are logged into your account.</li><li>Navigate to the GPT Workspace: Find the option for creating a custom GPT in your dashboard.</li><li>Start a New GPT: Click on the “Create” button to begin the process.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_x47GDidne8UUBygWO3ZQQ.png" /></figure><h3>Selecting a Base Model</h3><p>Start by selecting a base GPT-4 model. The base model will serve as the foundation for your customization. Depending on your needs, you may choose a model that’s optimized for specific tasks, such as UX Design, Psychiatrist, Diet Doctor etc.</p><p>Let’s build Psychiatry GPT…</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/958/1*__bK0cZFo0Tw4DQNpfz3ww.png" /></figure><blockquote><em>After instructing GPT to create a psychiatry assistant, you’ll need to name your GPT. Interestingly, GPT offers suggestions to help you select a name.</em></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/947/1*jXMqd-CmDkBUDxhXzLU36g.png" /></figure><blockquote><em>Interestingly, you can see how the AI gives you suggestions to create a unique and personal GPT for your purpose. After selecting a name, ChatGPT provides you with a profile picture, which you can change by instructing ChatGPT to do so.</em></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/900/1*gm9ClvKQaQLVDDBFIrUgyw.png" /></figure><h3>Defining Your GPT’s Personality and Behavior</h3><p>This step is where you can make your GPT your own. You can customize:</p><h3>Response Style: Choose whether it should be concise or elaborate.</h3><p>Example: The response style should be empathetic, validating the user’s experiences and emotions. It should also be reflective, often paraphrasing the user’s concerns to demonstrate understanding while guiding them toward self-awareness. Responses should be thoughtful, encouraging users to explore their feelings and thoughts more deeply.</p><h3>Tone: Decide if your GPT should be formal, casual, humorous, etc.</h3><p>Example: The tone should be warm, understanding, and non-judgmental. This helps create a safe space where users feel comfortable sharing their thoughts and emotions. The language should be gentle and reassuring, avoiding any phrases that could be perceived as dismissive or critical.</p><h3>Behavior: Set instructions for how it should handle different types of queries or tasks.</h3><p>Example: The behaviour should be consistently supportive, offering encouragement and positive reinforcement. The GPT should prompt the user to explore their emotions and thoughts while suggesting coping strategies or therapeutic techniques, such as mindfulness or journaling. It should prioritize active listening and gentle guidance, rather than offering direct advice or solutions immediately.</p><p>You’ll be prompted by ChatGPT, which will ask you about your GPT’s purpose and behaviour.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/907/1*0lnf6gUzgSEtUZWxvNKuDQ.png" /></figure><blockquote><em>After creation, if there are no more questions, you can access the “Configure” tab. This allows you to manually manage and modify settings, including changing the name.</em></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CZwLMoawwiZZdf0iGDAO-Q.png" /></figure><blockquote><em>After that, you can click on “Create.” It will then show 3 options. The choice depends on your intended use: you can build it for personal use, or if you have a team, select “Anyone with the link” so you can share it with your colleagues. The last option is “GPT Store,” which means your GPT will be available for everyone to use.</em></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/591/1*6-nKFVYkHlk0WS4RAkDfQw.png" /></figure><h3>Testing and Refining Your Custom GPT</h3><h3>Testing Scenarios and Fine-Tuning</h3><p>After setting up your GPT, it’s time to test it out. Run through common scenarios to see how it performs. You can ask it various questions or give it tasks to ensure it behaves as expected.</p><h3>Gathering Feedback and Iterating</h3><p>If you plan to use your GPT with a team or for clients, gather feedback from others to refine its behaviour. ChatGPT-4’s interface allows you to easily adjust settings and retrain the GPT based on this feedback.</p><h3>Deploying and Using Your GPT</h3><h3>Integrating Your GPT into Daily Tasks</h3><p>Once your GPT is fine-tuned, you can integrate it into your daily workflow. Whether you’re using it to draft emails, manage your calendar, or brainstorm ideas, your custom GPT is now ready to assist you seamlessly.</p><h3>Sharing and Collaboration Options</h3><p>ChatGPT-4 enables you to share your custom GPT with others. This feature is especially valuable for team projects or collaborative environments where consistent communication is crucial.</p><blockquote><em>By following these straightforward steps, you’ll have your personalized GPT assistant ready in no time. Whether you’re using it for business, education, or personal tasks, this tool can revolutionize your workflow — helping you stay organized, boost creativity, and enhance efficiency.</em></blockquote><p>Please review this GPT and provide feedback for improvement.</p><p>URL: <a href="https://chatgpt.com/g/g-vzTFRzUHr-psychiatrist">https://chatgpt.com/g/g-vzTFRzUHr-psychiatrist</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5e7441448ecc" width="1" height="1" alt=""><hr><p><a href="https://medium.com/etechviral/how-to-build-your-own-gpt-assistant-in-chatgpt-4-without-coding-a-step-by-step-guide-5e7441448ecc">How to Build Your Own GPT Assistant in ChatGPT-4 Without Coding: A Step-by-Step Guide</a> was originally published in <a href="https://medium.com/etechviral">ETechViral</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>