<?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 Gideon Egbeyemi on Medium]]></title>
        <description><![CDATA[Stories by Gideon Egbeyemi on Medium]]></description>
        <link>https://medium.com/@gideonegbeyemi?source=rss-63229f7361a1------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*ItJitja35xwhDdJW2-K2-w.png</url>
            <title>Stories by Gideon Egbeyemi on Medium</title>
            <link>https://medium.com/@gideonegbeyemi?source=rss-63229f7361a1------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 16:22:07 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@gideonegbeyemi/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[Reverse Engineering Teardown — Todoist]]></title>
            <link>https://medium.com/@gideonegbeyemi/reverse-engineering-teardown-todoist-d53a6a7c18f0?source=rss-63229f7361a1------2</link>
            <guid isPermaLink="false">https://medium.com/p/d53a6a7c18f0</guid>
            <dc:creator><![CDATA[Gideon Egbeyemi]]></dc:creator>
            <pubDate>Fri, 16 Oct 2026 20:00:08 GMT</pubDate>
            <atom:updated>2026-10-16T20:00:08.538Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Product analysed:</strong> Todoist<br><strong>Category:</strong> Productivity and Task Management<br><strong>Teardown focus:</strong> Creating, organising and completing a task<br><strong>Purpose:</strong> To understand how a real product solves a user problem and identify lessons for building better software.</p><h4>1. Introduction</h4><p>Reverse engineering teardown involves studying an existing product to understand its user experience, core features, workflows and underlying technical requirements. The goal is not to copy the product but to understand the decisions that make it useful and apply those lessons to our own projects.</p><p>For this teardown, we selected Todoist, a task-management application that helps users organise tasks and keep track of work they need to complete.</p><h4>2. The Problem It Solves</h4><p>People often struggle to remember responsibilities, prioritise activities and keep track of deadlines. Relying entirely on memory, scattered notes or messages can lead to missed tasks and poor organisation.</p><p>Todoist addresses this problem by providing a central place to record tasks, organise them and mark them as completed.</p><h4>3. Target Users</h4><ul><li>Students managing assignments, tests and personal responsibilities.</li><li>Professionals organising work and deadlines.</li><li>Individuals managing personal goals and daily activities.</li><li>Teams coordinating shared responsibilities.</li></ul><h4>4. Core User Journey</h4><p>The main workflow selected for this teardown is creating and completing a task.</p><ol><li>The user opens Todoist.</li><li>The user navigates to a task list or project.</li><li>The user selects the option to add a task.</li><li>The user enters a task title.</li><li>The user optionally adds a due date, priority or other details.</li><li>The user saves the task.</li><li>The task appears in the relevant list.</li><li>When the task is finished, the user marks it as completed.</li></ol><p>This journey should be tested directly in the live product before the final submission.</p><h4>5. Reverse Engineering: How It May Work</h4><p>The following is a reasoned technical model, not a claim about Todoist’s proprietary implementation.</p><p><strong>Frontend:</strong> The interface displays task lists, forms, buttons and completion states.</p><p><strong>Backend:</strong> The server processes operations such as creating, updating and completing tasks.</p><p><strong>Database:</strong> Persistent storage holds tasks and their associated information, such as titles, dates, completion status and ownership.</p><p><strong>Authentication and authorisation:</strong> An authenticated user’s tasks must be associated with the correct account and protected from unauthorised access.</p><p><strong>State management:</strong> The interface must update when a task is created or completed so that the user sees the current state.</p><p><strong>Validation and error handling:</strong> The application should handle invalid input and network failures without losing the user’s work unnecessarily.</p><h4>6. What Makes the Product Useful?</h4><p><strong>A. Clear primary action</strong></p><p>Creating a task is the central action. A good interface should make it easy to discover and use.</p><p><strong>B. Progressive detail</strong></p><p>Users can begin with a simple task title and add more information when needed. This can reduce the effort required to record a task.</p><p><strong>C. Organisation</strong></p><p>Grouping tasks helps users manage different responsibilities without treating every task as part of one large list.</p><p><strong>D. Visible progress</strong></p><p>Marking a task complete gives users immediate feedback and helps them track their progress.</p><p><strong>E. Flexible planning</strong></p><p>Dates and priorities can help users decide what to do and when to do it.</p><h4>8. Potential Weaknesses to Investigate</h4><p>These are questions for testing, not confirmed defects.</p><ul><li>Is the add-task action easy to find for a first-time user?</li><li>How many interactions are needed to create a basic task?</li><li>Are errors clearly explained if a task cannot be saved?</li><li>Is the interface comfortable to use on a small mobile screen?</li><li>Can users easily correct an accidental completion?</li><li>Is keyboard navigation supported throughout the main workflow?</li><li>What happens when the user loses their internet connection?</li></ul><p>The answers should be based on direct observation and documented evidence.</p><h4>9. Lessons for a New Project</h4><p>This teardown suggests several useful principles for building software:</p><ol><li>Make the main action easy to find.</li><li>Keep the simplest workflow short.</li><li>Make optional details available without forcing users to complete every field.</li><li>Provide clear feedback after each important action.</li><li>Preserve user data and handle errors gracefully.</li><li>Design for accessibility and different screen sizes.</li><li>Test the complete journey rather than judging the interface by appearance alone.</li></ol><h4>10. Conclusion</h4><p>Todoist provides a useful case study for understanding task-management software. By examining the journey from creating a task to completing it, we can learn how interface design, data persistence, feedback and organisation work together to solve a real user problem.</p><p>The most important lesson is that a successful product does not need to make every action complicated. It should help users complete their main task efficiently, understand what has happened and recover when something goes wrong.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d53a6a7c18f0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Authentication Is More Than a Login Form: Lessons from Building Spotter]]></title>
            <link>https://medium.com/@gideonegbeyemi/authentication-is-more-than-a-login-form-lessons-from-building-spotter-f60bd3074bdf?source=rss-63229f7361a1------2</link>
            <guid isPermaLink="false">https://medium.com/p/f60bd3074bdf</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[rags]]></category>
            <category><![CDATA[agentic-rag]]></category>
            <dc:creator><![CDATA[Gideon Egbeyemi]]></dc:creator>
            <pubDate>Fri, 16 Oct 2026 19:26:54 GMT</pubDate>
            <atom:updated>2026-10-16T19:26:54.085Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Qs6-v3BWPLReuIJJgnR1LQ.png" /></figure><p>When building a web application, it’s easy to think authentication begins and ends with a login page.</p><p>A user enters an email address and password, clicks a button, and gains access to the application.</p><p>But what happens behind that button?</p><p>How do we know the account exists? How do we confirm that the person logging in owns the account? How do we keep other people from accessing their records? What happens when a user logs out or wants to delete their account?</p><p>These are some of the questions we’ve been working through while building Spotter, a web application designed to help gym members access information from their gym’s own records.</p><p>Working on Spotter’s authentication flow has helped us understand an important lesson: <strong>authentication is not just a form. It is a system of connected processes that must work together.</strong></p><p>Here are some of the key concepts we’ve been exploring.</p><h4>1. Form Validation: Help Users Get It Right</h4><p>Before an application can create an account, it needs to collect the right information.</p><p>Consider a sign-up form with fields for a member’s full name, email address, member number, and password.</p><p>What happens if the user submits the form without entering a name? What if the email address is incorrectly formatted? What if the member number contains letters instead of numbers?</p><p>Without proper validation, these mistakes can lead to confusing errors or bad data.</p><p>We worked on making Spotter’s sign-up form more helpful by exploring:</p><ul><li>Required-field validation.</li><li>Inline error messages when a field is empty or invalid.</li><li>Real-time email validation.</li><li>Full-name validation.</li><li>Restrictions on the member number.</li><li>Clear password hints.</li><li>Disabling the Create Account button until the form is valid.</li></ul><p>The goal is to help users identify and correct mistakes before submitting their information.</p><p>However, frontend validation alone is not enough. The server must validate incoming data too, because users can bypass browser-side checks.</p><p>Good validation improves the user experience while helping protect the integrity of the data entering the application.</p><h4>2. Database Persistence: Creating an Account Must Actually Save It</h4><p>A successful sign-up message does not necessarily mean an account has been created correctly.</p><p>During our work on Spotter, we investigated a problem where users could attempt to sign up, but their information was not being saved in the database.</p><p>This highlighted the importance of database persistence.</p><p>When a user creates an account, the application must validate the information, process the request on the server, and save the necessary account details in the database.</p><p>The database schema must also support the information the application needs to store, such as the user’s email address, password hash, member number, and email verification status.</p><p>Database migrations are important here. They allow us to apply changes to the database structure as the application develops.</p><p>We also explored configuring the database connection through environment variables rather than scattering connection details throughout the codebase.</p><p>One important security rule is that passwords should never be stored as plain text. Applications should store secure password hashes and verify submitted passwords against those hashes.</p><p><strong>Creating an account is not just about displaying a success message. The account must be correctly and securely persisted.</strong></p><h4>3. Email Verification: Confirming Ownership of an Email Address</h4><p>Suppose someone creates a Spotter account using an email address that belongs to another person.</p><p>If the application immediately treats that address as verified, it has no reliable confirmation that the user controls the inbox.</p><p>Email verification helps address this problem.</p><p>We explored a flow in which Spotter generates a six-digit verification code and sends it to the user’s email address.</p><p>The intended process is straightforward:</p><ol><li>The user creates an account.</li><li>The application sends a verification code to the supplied email address.</li><li>The user enters the code.</li><li>The server validates the code and confirms that it is still valid.</li><li>The account’s email verification status is updated.</li></ol><p>A robust implementation also needs an expiry time, limits on repeated attempts, and protection against reusing a code after successful verification.</p><p>The code should be generated using a cryptographically secure random source, stored safely, and compared on the server.</p><p>Email verification is not proof of a person’s full identity. It provides evidence that they can access the email address used during registration.</p><p>For Spotter, this is one part of building a more reliable account system.</p><h4>4. Password Verification: Never Trust the Login Form Alone</h4><p>When a user logs in, the application needs to determine whether the submitted credentials are correct.</p><p>We explored using Argon2id to verify passwords.</p><p>Argon2id is a password-hashing algorithm designed to make password guessing more expensive for attackers. Instead of storing the original password, the application stores a hash and verifies future login attempts against it.</p><p>We also explored normalizing email addresses by trimming unnecessary whitespace and converting them to lowercase before looking up accounts.</p><p>This helps prevent avoidable inconsistencies during login.</p><p>However, password verification is only one step. The application must also check relevant account conditions, including whether the account is permitted to sign in and whether email verification is required.</p><p>The important distinction is that authentication should not rely on the appearance of a successful form submission. The server must verify the credentials and enforce the application’s rules.</p><h4>5. Session Management: Remembering Who Is Logged In</h4><p>Once a user successfully logs in, the application needs a safe way to recognise that user across subsequent requests.</p><p>This is where sessions become important.</p><p>Without a session mechanism, the application would have no reliable way to maintain a user’s authenticated state as they navigate between pages.</p><p>We explored a session flow involving cookies, database-backed session records, and middleware.</p><p>In a typical implementation:</p><ol><li>The server verifies the user’s credentials.</li><li>It creates a new session.</li><li>It sends the browser a session cookie.</li><li>The browser includes the cookie in subsequent requests.</li><li>Middleware validates the session and identifies the associated user.</li><li>Protected routes use that verified identity to decide whether access is allowed.</li></ol><p>Session cookies should be configured securely, including appropriate HttpOnly, Secure, and SameSite settings.</p><p>If session tokens are stored in the database as hashes, the application can compare a hash of the presented token with the stored value without keeping the raw token in the database.</p><p>The server should also check whether the account remains active and whether the session is valid.</p><p>Logging out must invalidate the server-side session, not merely remove a visual indication that the user is logged in.</p><p>This is especially important for Spotter because members may access personal information such as their attendance history and outstanding balances.</p><p><strong>A login establishes access, but session management helps control that access over time.</strong></p><h4>6. Protecting Member Records: Authentication and Authorisation Work Together</h4><p>Authentication and authorisation are related, but they are not the same.</p><p>Authentication asks: Who is making this request?</p><p>Authorisation asks: Is this person allowed to perform this action or access this information?</p><p>This distinction matters for Spotter.</p><p>Imagine two gym members using the application. Both have valid accounts, but one member should not be able to view the other’s attendance history or outstanding balance.</p><p>A valid login must not automatically grant access to every member’s records.</p><p>Instead, the server must derive the authenticated user’s identity from a trusted session and use that identity when querying personal records.</p><p>For example, rather than trusting a member ID supplied by the browser, the server should obtain the member identity from the verified account and restrict the database query accordingly.</p><p>The same principle applies to staff and owner access. Different roles may have different permissions, and those permissions must be enforced on the server.</p><p>This is central to Spotter’s design: members should receive information from the correct gym records, and personal information should only be available to the appropriate user.</p><h4>7. Account Deletion: What Happens When a User Wants to Leave?</h4><p>Account management does not end when someone signs up or logs out.</p><p>Users may eventually want to delete their accounts, and applications need a deliberate process for handling those requests.</p><p>We explored a deletion flow with three main stages:</p><p><strong>Soft deletion:</strong> The account is marked for deletion rather than immediately removed.</p><p><strong>Grace period:</strong> The user has a defined waiting window during which the deletion may be cancelled, depending on the application’s rules.</p><p><strong>Hard deletion:</strong> Once the grace period expires, the system permanently removes the data covered by the deletion policy.</p><p>We also explored the idea of a scheduled job that checks for accounts whose grace periods have expired and processes them.</p><p>This approach requires careful decisions about which records should be deleted, how related records are handled, and whether any information must be retained for legitimate legal or operational reasons.</p><p>The scheduled job should be reliable, safe to rerun, and designed to avoid deleting the wrong records.</p><p>Account deletion is therefore not just a button on a settings page. It is a lifecycle process involving the application, database, and background jobs.</p><h4>8. What We Learned</h4><p>Working through Spotter’s authentication requirements has reinforced several lessons.</p><p>First, the visible interface is only one part of the application. A clean form must be supported by server-side validation, correct database operations, and appropriate security controls.</p><p>Second, authentication features depend on one another. Account creation, email verification, login, session management, and logout must form a consistent flow.</p><p>Third, security cannot depend entirely on the browser. The server must validate data, verify credentials, enforce permissions, and protect access to personal records.</p><p>Finally, a feature is not complete simply because the code has been written. It needs to be tested under normal conditions, invalid input, expired sessions, failed database operations, and other edge cases.</p><h4>Conclusion</h4><p>Building Spotter has shown us that authentication is much broader than a login page.</p><p>It involves validating information, saving accounts correctly, verifying email ownership, protecting passwords, managing sessions, controlling access to member records, and handling account deletion responsibly.</p><p>Each part solves a different problem, but all of them contribute to the same goal: making the application reliable and safe to use.</p><p>As we continue building Spotter, the focus is not just on making the interface work. It is on making sure the systems behind the interface behave as expected.</p><p>Because a good authentication experience is not simply one that lets users in.</p><p><strong>It is one that lets the right users in, protects their information, and handles the rest of the account lifecycle responsibly.</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f60bd3074bdf" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Reverse Engineering Drill: Bolt Ride Booking]]></title>
            <link>https://medium.com/@gideonegbeyemi/reverse-engineering-drill-bolt-ride-booking-6c539c17796c?source=rss-63229f7361a1------2</link>
            <guid isPermaLink="false">https://medium.com/p/6c539c17796c</guid>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[bolt]]></category>
            <dc:creator><![CDATA[Gideon Egbeyemi]]></dc:creator>
            <pubDate>Fri, 02 Oct 2026 18:55:33 GMT</pubDate>
            <atom:updated>2026-10-02T18:55:33.867Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*uO2-2asn7hZivgKd" /><figcaption>Photo by <a href="https://unsplash.com/@appshunter?utm_source=medium&amp;utm_medium=referral">appshunter.io</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><h4>What the product does</h4><p>Bolt connects passengers with nearby drivers. A user enters a destination, selects a ride option, sees the estimated cost, and requests a ride.</p><h4>User Flow</h4><p>Set location → Enter destination → Select ride → Review fare → Request ride → Driver matched → Trip → Payment</p><h4>What happens behind the scenes?</h4><p>When the user enters a destination, the system needs to determine the route and estimate the journey time and distance.</p><p>It then uses this information, alongside factors such as the selected ride category and current demand, to calculate or display an estimated fare.</p><p>Once the user requests the ride, the system needs to find an appropriate nearby driver. The driver’s location and availability are important here.</p><h4>The interesting part</h4><p>The <strong>“Request Ride”</strong> button looks simple, but several systems have to work together immediately:</p><ul><li>GPS/location services</li><li>Maps and routing</li><li>Fare calculation</li><li>Driver availability</li><li>Driver matching</li><li>Real-time trip tracking</li><li>Payment processing</li></ul><p>If one of these systems fails, the experience can break even though the button itself works.</p><h4>One Improvement I’d make</h4><p>Before the user confirms a ride, make the <strong>estimated fare, ride type, pickup location and estimated arrival time</strong> especially prominent.</p><p>This gives users the key information they need to make an informed decision before requesting the ride.</p><h4>Key takeaway</h4><p>Reverse engineering Bolt shows that a simple user action can depend on several interconnected services. The interface hides much of this complexity, which is what makes the product feel simple to use.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6c539c17796c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[From Design to Development: Building a Strong Foundation for an App]]></title>
            <link>https://medium.com/@gideonegbeyemi/from-design-to-development-building-a-strong-foundation-for-an-app-cef100b182b5?source=rss-63229f7361a1------2</link>
            <guid isPermaLink="false">https://medium.com/p/cef100b182b5</guid>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[scaffolding]]></category>
            <category><![CDATA[token]]></category>
            <category><![CDATA[figma]]></category>
            <category><![CDATA[google-antigravity]]></category>
            <dc:creator><![CDATA[Gideon Egbeyemi]]></dc:creator>
            <pubDate>Thu, 01 Oct 2026 18:46:25 GMT</pubDate>
            <atom:updated>2026-10-01T18:46:25.075Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*MZi1NkgHwUiCLOr6" /><figcaption>Photo by <a href="https://unsplash.com/@joshuaryanphoto?utm_source=medium&amp;utm_medium=referral">Joshua Reddekopp</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>When building an application, it is easy to focus on the screens users will see. We think about the dashboard, login page, buttons, forms, cards, and other features. But before all of these come together, there is another important part of the process: creating a foundation that makes the application consistent and easier to build.</p><p>This week, I spent time learning about three areas that helped me understand this better:</p><ul><li>Moving a design system from Figma into Antigravity</li><li>Creating the scaffold of an application</li><li>Using colour roles according to Google Material Design</li></ul><p>These may seem like separate activities, but they are closely connected.</p><p>They are all about taking an idea from design and creating a structured system that can be developed and maintained.</p><h4>Moving a Design System from Figma to Antigravity</h4><p>Figma is a great environment for designing an interface.</p><p>It allows designers to define colours, typography, spacing, components, layouts, and different states before development begins.</p><p>But a design system shouldn’t stop at the design file.</p><p>At some point, those design decisions need to become part of the actual application.</p><p>This was one of the things I explored when moving a design system from Figma into Antigravity.</p><h4>Exporting the Design Tokens</h4><p>One of the practical steps I took during this process was exporting the design tokens from Figma.</p><p>To do this, I made use of a Figma plugin called <strong>Design Tokens: Export Design Token Files</strong>.</p><p>The plugin helped me export the design tokens from my Figma design system into files that could be used as part of the development workflow.</p><p>This was useful because instead of manually copying individual values such as colours, typography, or other design tokens from Figma into the application, I could work with the exported token files.</p><p>This also reinforced an important idea for me:</p><p><strong>Design tokens can serve as a bridge between design and development.</strong></p><p>For example, rather than manually entering a colour value every time it is needed, a token can represent that decision and be reused throughout the application.</p><p>The same concept can be applied to other design decisions.</p><p>Instead of creating every button independently, the application can have a reusable button component.</p><p>Instead of manually choosing a colour every time a component is created, the application can use defined colour tokens.</p><p>Instead of deciding spacing differently on every page, spacing can follow a consistent system.</p><p>The goal isn’t simply to copy what is visible in Figma.</p><p>The bigger goal is to understand the rules behind the design and translate those rules into something that can be reused during development.</p><p>This changed the way I think about the relationship between design and development.</p><p>A design system is not just a collection of beautiful screens.</p><p>It is a set of rules that should make both design and development more consistent.</p><h4>Creating the Scaffold of an Application</h4><p>After thinking about the design system, another important question comes up:</p><p><strong>Where does everything go?</strong></p><p>This is where application scaffolding becomes important.</p><p>Before building individual features, I worked on creating the basic structure of an application.</p><p>This included thinking about:</p><ul><li>Project structure</li><li>Routes and navigation</li><li>Layouts</li><li>Reusable components</li><li>Pages</li><li>Design tokens</li><li>Data structure</li><li>Development conventions</li></ul><p>The purpose of the scaffold is not to build the entire application before development starts.</p><p>Instead, it provides a structured starting point.</p><p>Imagine starting a project without deciding where components, pages, layouts, or other parts of the application should live.</p><p>It may work when the project is small.</p><p>But as more features are added, things can quickly become difficult to organise.</p><p>A good scaffold gives the project some initial direction.</p><p>It also makes it easier to understand where new functionality should be added.</p><p>One lesson I took from this is that scaffolding is not about predicting everything the application will ever need.</p><p>It is about creating enough structure to support the current stage of development while leaving room for the application to grow.</p><h4>Understanding Colour Roles in Material Design</h4><p>Another part of the process that caught my attention was colour.</p><p>Previously, I could think about colour simply as a hex value.</p><p>For example:</p><p><strong>#533AFD</strong></p><p>But a design system needs more than a list of hex codes.</p><p>It needs to define what those colours are used for.</p><p>This is where Material Design’s approach to colour roles becomes useful.</p><p>Instead of assigning colours randomly to individual components, colours can be given semantic roles.</p><p>Examples include:</p><ul><li>Primary</li><li>On Primary</li><li>Primary Container</li><li>On Primary Container</li><li>Secondary</li><li>Tertiary</li><li>Surface</li><li>Background</li><li>Error</li></ul><p>The important part is the word <strong>“role.”</strong></p><p>For example, instead of saying:</p><blockquote><em>“Use #533AFD for this button.”</em></blockquote><p>A design system can say:</p><blockquote><em>“Use the Primary colour role for the main action.”</em></blockquote><p>That difference may seem small, but it makes the system much more flexible.</p><p>If the actual primary colour changes later, the components using the Primary role can continue to work without every component needing to be redesigned individually.</p><p>The role stays the same.</p><p>The value can change.</p><p>This is one of the ideas that made me appreciate semantic design tokens more.</p><h4>Connecting the Three Ideas</h4><p>At first, moving a design system, scaffolding an application, and defining colour roles may seem like three different tasks.</p><p>But they are connected.</p><p>The process can be viewed like this:</p><p><strong>Design → Tokens → System → Structure → Implementation</strong></p><p>Figma gives you a place to define the visual direction.</p><p>Design tokens capture important decisions from that design system in a reusable format.</p><p>The exported token files can then become part of the development workflow.</p><p>The application scaffold provides the structure for implementing those decisions.</p><p>Semantic colour roles and other design tokens make those rules easier to apply consistently.</p><p>The result is an application where design decisions aren’t recreated from scratch on every screen.</p><p>Instead, the application has a common visual and technical language.</p><h4>What I Learned</h4><p>The biggest lesson from this process is that building an application is not only about writing features.</p><p>The decisions made before and around feature development can have a significant impact on how easy the product is to build and maintain.</p><p>I also learned to think beyond individual screens.</p><p>A screen might look good on its own, but a product needs consistency across many screens.</p><p>That consistency comes from systems.</p><p>A good design system defines reusable decisions.</p><p>Design tokens make those decisions easier to transfer between design and development.</p><p>A good scaffold provides a place for those decisions to live.</p><p>And semantic tokens give those decisions meaning that can be reused throughout the application.</p><h4>Conclusion</h4><p>I’m beginning to see development as more than turning designs into code.</p><p>It is about understanding the system behind the design and building an environment where that system can be reused.</p><p>Moving from Figma to Antigravity helped me think about design-to-development workflows.</p><p>Using the <strong>Design Tokens: Export Design Token Files</strong> Figma plugin gave me a practical way to move design-token information out of Figma and into the development workflow.</p><p>Creating an application scaffold helped me think about structure.</p><p>Learning about Material Design colour roles helped me understand how visual decisions can become semantic and reusable.</p><p>These are small pieces of a much larger process, but they have changed how I approach building applications.</p><p>Instead of asking only:</p><p><strong>“How do I build this screen?”</strong></p><p>I’m starting to ask:</p><p><strong>“What system can I create that makes building this screen — and the next ten screens — easier?”</strong></p><p>That shift in thinking is one of the most useful things I took away from this week’s learning.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cef100b182b5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Reverse Engineering Jumia’s Product Ordering Flow: What Happens Behind the Scenes?]]></title>
            <link>https://medium.com/@gideonegbeyemi/reverse-engineering-jumias-product-ordering-flow-what-happens-behind-the-scenes-6a0b519b2357?source=rss-63229f7361a1------2</link>
            <guid isPermaLink="false">https://medium.com/p/6a0b519b2357</guid>
            <category><![CDATA[product-reviews]]></category>
            <category><![CDATA[jumia]]></category>
            <category><![CDATA[review]]></category>
            <dc:creator><![CDATA[Gideon Egbeyemi]]></dc:creator>
            <pubDate>Sun, 20 Sep 2026 19:21:21 GMT</pubDate>
            <atom:updated>2026-09-20T19:21:21.268Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*-SIiwVVxptxnZKPc" /><figcaption>Photo by <a href="https://unsplash.com/@rowanfreeman?utm_source=medium&amp;utm_medium=referral">Rowan Freeman</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>When I order a product online, the experience can feel deceptively simple. I search for a product, click <strong>Add to Cart</strong>, provide my delivery information, pay, and wait for it to arrive.</p><p>But behind that simple experience is a series of systems working together.</p><p>For this week’s reverse engineering drill, I decided to look at <strong>Jumia Nigeria</strong> and break down what I think happens behind the scenes when a customer orders a product.</p><h4>The Flow</h4><p>At a high level, the journey looks something like this:</p><p><strong>Search → Product Page → Cart → Checkout → Payment → Order Confirmation → Fulfillment → Delivery</strong></p><p>Let’s break it down.</p><h4>1. Finding the Product</h4><p>The process starts when a customer searches for something or browses through Jumia’s categories.</p><p>The platform needs to return relevant products along with information such as the product name, images, price, ratings, seller information, and availability.</p><p>From a technical perspective, I would expect Jumia to have a product catalogue and a search system that can retrieve and rank products based on the customer’s query.</p><p>The key point is that the customer isn’t simply searching a static webpage. The application requests product information from backend systems and presents the results dynamically.</p><h4>2. The Product Page</h4><p>Once a customer selects a product, Jumia displays more detailed information.</p><p>This can include the product description, images, price, seller information, reviews, delivery information, and available options.</p><p>At this stage, the system also needs to make sure that information such as price and availability is reasonably current.</p><p>This is important because a product that was available when the customer opened the page might not necessarily still be available when they attempt to purchase it.</p><h4>3. Adding the Product to the Cart</h4><p>When the customer clicks <strong>Add to Cart</strong>, the system needs to remember that selection.</p><p>The cart could be associated with a logged-in customer’s account or with a session for a visitor.</p><p>Behind the scenes, I would expect the application to send a request to a backend service that records the selected product, quantity, and other relevant information.</p><p>The interface then updates the cart so the customer can continue shopping or proceed to checkout.</p><h4>4. Checkout</h4><p>This is where the simple shopping experience becomes more complicated.</p><p>The system needs to bring together several pieces of information:</p><ul><li>The products being purchased</li><li>Quantity</li><li>Product price</li><li>Delivery location</li><li>Delivery method</li><li>Applicable delivery charges</li><li>Available payment options</li><li>The final order total</li></ul><p>Jumia’s current Nigerian shopping experience supports multiple payment options, including online payments and payment on delivery for eligible orders.</p><p>This means the checkout system also needs to determine which payment and delivery options are available for a particular order and location.</p><h4>5. Payment</h4><p>Jumia gives customers more than one way to complete an order.</p><p>For eligible orders, a customer can choose <strong>Payment on Delivery</strong>, meaning payment is made when the order is delivered rather than online during checkout.</p><p>For online payments, another important system becomes involved: the payment infrastructure.</p><p>Jumia provides JumiaPay as a payment solution, with options including cards, bank transfers and USSD.</p><p>From the outside, online payment can look like one simple step:</p><p><strong>Click Pay → Payment Successful</strong></p><p>But technically, several things have to happen.</p><p>The payment request needs to be processed, the transaction needs to be verified, and the shopping platform needs to know whether the payment succeeded or failed.</p><p>Payment on Delivery creates a different flow. Instead of requiring successful online payment before the order can proceed, the order can move into fulfillment with payment expected at the point of delivery, subject to Jumia’s eligibility conditions.</p><p>This means the order system needs to keep track of not only the order itself, but also the <strong>payment method and payment status</strong>.</p><p>For example:</p><p><strong>Order → Payment Method: Payment on Delivery → Payment Status: Pending</strong></p><p>After successful delivery and collection of payment, that status can be updated.</p><p>This is one of the interesting things I noticed from looking at the ordering process: <strong>the order and the payment don’t necessarily have to happen at exactly the same point in the customer’s journey.</strong></p><h4>6. Creating the Order</h4><p>Once the necessary checkout and payment conditions have been satisfied, the system can create the order.</p><p>At this point, I would expect the backend to generate an order record containing information such as:</p><p><strong>Order ID → Customer → Products → Quantity → Amount → Delivery Information → Payment Status → Order Status</strong></p><p>This order record becomes the source of truth for what the customer purchased.</p><p>The customer’s interface can then use the order status to show what is happening with the purchase.</p><h4>7. Fulfilment and Delivery</h4><p>Creating the order is not the end of the process.</p><p>The product still needs to move from the seller or fulfilment location to the customer.</p><p>Jumia’s platform therefore has to connect the digital order with physical fulfilment and delivery operations.</p><p>This is one of the interesting parts of reverse engineering an e-commerce platform.</p><p>The website might feel like a software product, but its success depends on a combination of:</p><p><strong>Software + Payments + Inventory + Sellers + Logistics + Customer Support</strong></p><p>The software coordinates these different parts.</p><h4>What I Think Is Happening Behind the Scenes</h4><p>Based on the user experience, I would describe the architecture at a high level like this:</p><p><strong>Customer</strong></p><p>↓</p><p><strong>Jumia Web/App</strong></p><p>↓</p><p><strong>Product &amp; Search Services</strong></p><p>↓</p><p><strong>Cart &amp; Checkout Services</strong></p><p>↓</p><p><strong>Order Service</strong></p><p>↓</p><p><strong>Payment Service</strong></p><p>↓</p><p><strong>Fulfillment &amp; Logistics</strong></p><p>↓</p><p><strong>Customer</strong></p><p>These are inferences based on the observable behaviour of the product, not claims about Jumia’s private internal architecture.</p><p>That’s an important distinction in reverse engineering.</p><p>The goal isn’t to pretend that I can see the source code.</p><p>The goal is to observe the product’s behaviour and reason about what systems would likely be required to make that behaviour possible.</p><h4>What Would I Change?</h4><p>One area I would explore further is <strong>making the total cost and delivery information even more prominent before the final order confirmation</strong>.</p><p>Jumia already provides delivery information and encourages customers to review item details, quantities, totals, shipping methods and available offers before completing checkout.</p><p>My proposed improvement would be to make a final <strong>“Order Summary”</strong> particularly prominent immediately before the customer commits to the purchase:</p><p><strong>Product price + Quantity + Delivery fee + Discounts + Final amount + Payment method + Delivery location</strong></p><p>The reason is simple: the customer should be able to understand exactly what they are committing to before completing the order.</p><p>This isn’t necessarily a claim that Jumia’s current checkout is broken. It’s a product improvement I’d test to reduce uncertainty at one of the most important points in the purchasing journey.</p><h4>What I Learned From This Teardown</h4><p>The biggest thing I learned is that a simple user experience can hide a complicated system.</p><p>From the customer’s perspective:</p><p>Search → Buy → Receive</p><p>Behind the scenes, the platform has to coordinate product data, cart state, pricing, checkout, payment verification, order management, inventory, sellers and delivery.</p><p>That’s what made this exercise interesting for me.</p><p>Reverse engineering a product isn’t just about asking:</p><p><strong>“What does this button do?”</strong></p><p>It’s about asking:</p><p><strong>“What must be happening behind this button for the experience to work?”</strong></p><p>That is the mindset I want to develop further as I continue building software.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6a0b519b2357" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Context Engineering: Giving AI the Information It Needs to Build Better]]></title>
            <link>https://medium.com/@gideonegbeyemi/context-engineering-giving-ai-the-information-it-needs-to-build-better-da8836967626?source=rss-63229f7361a1------2</link>
            <guid isPermaLink="false">https://medium.com/p/da8836967626</guid>
            <category><![CDATA[prompted-play]]></category>
            <category><![CDATA[context-engineering]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[vibe-coding]]></category>
            <dc:creator><![CDATA[Gideon Egbeyemi]]></dc:creator>
            <pubDate>Sat, 19 Sep 2026 18:23:05 GMT</pubDate>
            <atom:updated>2026-09-19T18:23:05.106Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*6DnGD9SN4vx7XvFH" /><figcaption>Photo by <a href="https://unsplash.com/@tec_razy?utm_source=medium&amp;utm_medium=referral">Om Kamath</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>AI has changed the way we build software.</p><p>Today, developers can use AI coding tools to generate code, debug problems, explain unfamiliar concepts, and even help build entire features. But as I have been learning more about AI-assisted development, I have realized that getting useful results from AI is not only about writing good prompts. It is also about <strong>providing the right context</strong>. This is where <strong>context engineering</strong> comes in.</p><h4>What is Context Engineering?</h4><p>In simple terms, context engineering is the process of giving an AI system the relevant information, instructions, rules, and background it needs to produce a useful result.</p><p>Think about working with a new developer on a project.</p><p>If you simply tell them:</p><blockquote>“Build a login page”</blockquote><p>They can build something, but they don’t know much about the project.</p><p>They don’t know the purpose of the application, the technologies being used, the design decisions that have already been made, the coding conventions to follow, or what limitations they need to consider.</p><p>You would have to explain all of those things before they could work effectively.</p><p>AI coding agents have a similar challenge.</p><p>A prompt tells the AI <strong>what you want it to do</strong>, but context can help explain <strong>what it is working on, how it should work, and what constraints it needs to respect</strong>.</p><p>That is what makes context engineering interesting to me.</p><h4>Prompt Engineering vs. Context Engineering</h4><p>Prompt engineering focuses heavily on how we communicate a particular request to an AI.</p><p>For example:</p><blockquote>“Create a registration page for this application.”</blockquote><p>Context engineering takes a broader approach.</p><p>Instead of only giving the AI the instruction, we can provide information about the project, its requirements, rules, technical decisions, and available capabilities.</p><p>The request might still be:</p><blockquote>“Create a registration page.”</blockquote><p>But the AI could also have access to information explaining:</p><ul><li>What the application does</li><li>Who the users are</li><li>What technology the project uses</li><li>How the existing code is structured</li><li>Coding conventions</li><li>Design requirements</li><li>Rules the agent must follow</li><li>Tools or capabilities available to it</li></ul><p>The important lesson for me is that <strong>a good prompt does not exist in isolation</strong>. The quality and relevance of the surrounding context can have a major effect on how useful the AI’s response is.</p><h4>Context Files</h4><p>One practical way of organizing this information is through context files.</p><p>During my learning this week, I looked at four types of files:</p><ol><li>PRD.md</li></ol><p><strong>PRD</strong> stands for <strong>Product Requirements Document</strong>.</p><p>The PRD explains what we are building and why we are building it.</p><p>It can contain information such as:</p><ul><li>The purpose of the product</li><li>Target users</li><li>Features</li><li>Requirements</li><li>User flows</li><li>Product goals</li><li>Functional expectations</li></ul><p>For an AI coding agent, this provides important product-level context.</p><p>Instead of simply saying:</p><blockquote>“Add a dashboard.”</blockquote><p>The agent can understand what the dashboard is supposed to accomplish and who is expected to use it.</p><p>The PRD therefore helps connect the code being written to the <strong>actual product requirements</strong>.</p><p>2. AGENTS.md</p><p>AGENTS.md can be used to provide instructions for AI agents working on a project.</p><p>For example, it could explain:</p><ul><li>How the project is organized</li><li>Which commands should be used</li><li>Coding conventions</li><li>Development workflows</li><li>Important directories</li><li>Testing expectations</li><li>Instructions the agent should follow when modifying the project</li></ul><p>This is useful because the AI doesn’t have to guess how you expect it to work.</p><p>Instead of repeatedly explaining these instructions in individual prompts, they can be documented as part of the project context.</p><p>3. RULES.md</p><p>RULES.md can contain specific rules and constraints that should guide the agent.</p><p>For example:</p><ul><li>Don’t modify certain files without permission.</li><li>Follow a particular naming convention.</li><li>Don’t introduce a new dependency unless necessary.</li><li>Always test changes before considering a task complete.</li><li>Follow the project’s existing architecture.</li></ul><p>The exact rules will depend on the project.</p><p>The important idea is that <strong>constraints are part of context</strong>.</p><p>When building software, knowing what you should not do can be just as important as knowing what you should do.</p><p>4. SKILLS.md</p><p>SKILLS.md can document the skills, capabilities, tools, or workflows that are available to the AI agent.</p><p>This gives the agent additional information about what it can use when working on a task.</p><p>For example, if an agent has access to particular tools or established workflows, documenting them can make it easier for the agent to understand how those capabilities should be used.</p><p>Again, the exact structure will depend on the project and the tools being used.</p><h4>Why Does Context Matter?</h4><p>Imagine asking someone to assemble a piece of furniture but giving them only one instruction:</p><blockquote>“Build this.”</blockquote><p>They might figure it out.</p><p>But their chances of getting it right improve if you also give them:</p><ul><li>The instruction manual</li><li>The list of parts</li><li>The finished design</li><li>The tools they should use</li><li>Important warnings</li><li>Any restrictions</li></ul><p>AI-assisted development works in a similar way.</p><p>The AI may be capable of generating code, but it still needs relevant information about the environment in which that code will be used.</p><p>Without enough context, an AI agent may make assumptions.</p><p>Those assumptions can lead to:</p><ul><li>Code that doesn’t fit the existing architecture</li><li>Features that don’t match the product requirements</li><li>Unnecessary dependencies</li><li>Changes to the wrong files</li><li>Inconsistent implementation</li><li>Solutions that technically work but don’t fit the project</li></ul><p>Context engineering helps reduce unnecessary guessing.</p><h4>Context Is Not About Giving the AI Everything</h4><p>One thing I find important about context engineering is that <strong>more context does not automatically mean better context</strong>.</p><p>If you give an AI agent thousands of lines of irrelevant information, you haven’t necessarily helped it.</p><p>The goal is to provide <strong>relevant context</strong>.</p><p>For example, if I am asking an agent to fix a problem with authentication, it may need information about the authentication system, relevant project rules, and the files involved.</p><p>It probably doesn’t need every piece of documentation ever written about the application.</p><p>So I see context engineering as a process of asking:</p><blockquote>What information does the AI need to make a good decision about this task?</blockquote><p>That is a much more useful question than simply asking:</p><blockquote>How much information can I give the AI?</blockquote><h4>A Simple Example</h4><p>Imagine I have an application and I ask an AI coding agent:</p><blockquote>“Add a student results page.”</blockquote><p>Without additional context, the agent has to make several assumptions.</p><p>What does a student result contain?</p><p>Who can access it?</p><p>Should teachers see it?</p><p>Should administrators see it?</p><p>What technology should be used?</p><p>Where should the page be located?</p><p>How should it fit into the existing application?</p><p>Now imagine that the project already contains relevant context.</p><p>The PRD explains the purpose of the results system.</p><p>AGENTS.md explains the project&#39;s development conventions.</p><p>RULES.md establishes constraints.</p><p>SKILLS.md explains available tools and workflows.</p><p>The agent now has much more information with which to reason about the task.</p><p>The prompt has not necessarily become more complicated.</p><p><strong>The environment around the prompt has become better.</strong></p><h4>Context Engineering Changes How I Think About AI</h4><p>Before learning about context engineering, I mostly thought about AI interaction in terms of:</p><p><strong>Prompt → Response</strong></p><p>Now I am starting to think about it more like:</p><p><strong>Project Context → Instructions → Task → AI → Result</strong></p><p>The task is still important, but it is only one part of the interaction.</p><p>This has changed how I approach AI-assisted development.</p><p>Instead of constantly trying to find the perfect prompt, I can also improve the information available to the AI.</p><p>That means documenting decisions, establishing rules, describing the product clearly, and organizing useful project information.</p><h4>The Bigger Lesson</h4><p>The biggest lesson I am taking away from context engineering is that <strong>AI is only as useful as the context we give it for the task at hand</strong>.</p><p>That doesn’t mean AI needs every piece of information about a project.</p><p>It means we should be intentional about what information it needs to make good decisions.</p><p>For me, the four context files we explored — PRD.md, AGENTS.md, RULES.md, and SKILLS.md — helped make this idea practical.</p><p>They represent different kinds of information:</p><p><strong>PRD.md → What are we building and why?</strong></p><p><strong>AGENTS.md → How should the agent work within the project?</strong></p><p><strong>RULES.md → What constraints should it follow?</strong></p><p><strong>SKILLS.md → What capabilities and workflows are available?</strong></p><p>When these pieces of information are organized properly, AI-assisted development becomes less about repeatedly explaining the project and more about creating an environment in which an AI agent can work with the right information.</p><p>I am still learning and experimenting with this approach, but it has changed one thing for me:</p><p><strong>I no longer see prompting as the entire conversation with AI. The context surrounding the prompt matters too.</strong></p><p>And as AI becomes a bigger part of how we build software, learning how to provide the right context may become just as important as learning how to write the right prompt.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=da8836967626" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Understanding Retry and Backoff in Software]]></title>
            <link>https://medium.com/@gideonegbeyemi/understanding-retry-and-backoff-in-software-51fee724aef8?source=rss-63229f7361a1------2</link>
            <guid isPermaLink="false">https://medium.com/p/51fee724aef8</guid>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[prompt-engineering]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[retry]]></category>
            <dc:creator><![CDATA[Gideon Egbeyemi]]></dc:creator>
            <pubDate>Fri, 11 Sep 2026 19:26:35 GMT</pubDate>
            <atom:updated>2026-09-11T19:26:35.863Z</atom:updated>
            <content:encoded><![CDATA[<p>One thing I started thinking about differently this week is what happens when an application makes a request and that request fails. At first, the solution seems simple: Just try again.</p><p>But in software development, “try again” isn’t always as simple as it sounds. If an application immediately retries every failed request, it can sometimes make the original problem even worse. That is where <strong>retry and backoff</strong> come in.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XpuchdwBVoMsdV8pVwzuKQ.png" /></figure><h4>What Does Retry Mean?</h4><p>A retry simply means attempting an operation again after it fails. For example, imagine your application needs to retrieve information from an external API. The request fails because of a temporary network problem. Instead of immediately showing an error to the user, the application might attempt the request again.</p><p>Retries are useful because not every failure is permanent. Networks can be unstable. Servers can become temporarily overloaded. External services can experience short interruptions. A request that fails once might succeed a few moments later.</p><p>But there is an important question:</p><p>How quickly should we try again?</p><h4>Why Immediate Retries Can Be a Problem</h4><p>Imagine an external service is already struggling because it is receiving too many requests. Now imagine thousands of applications experiencing failed requests and immediately sending those requests again. The service receives even more traffic while it is already under pressure. This can create a cycle where failures cause more retries, and more retries cause even more failures. Instead of helping the system recover, the retries can contribute to the problem. This is one reason simply adding a retry mechanism isn’t enough. We also need to think about <strong>when</strong> to retry.</p><h4>What Is Backoff?</h4><p>Backoff is the practice of waiting for some amount of time before making another attempt. Instead of immediately retrying a failed request, the application waits.</p><p>For example:</p><ul><li>First retry: wait 1 second</li><li>Second retry: wait 2 seconds</li><li>Third retry: wait 4 seconds</li><li>Fourth retry: wait 8 seconds</li></ul><p>This particular approach is called <strong>exponential backoff</strong>. The waiting period increases after each failed attempt. The idea is simple: If the service is temporarily unavailable, give it some time to recover instead of continuously hitting it with requests.</p><h4>But there is another problem</h4><p>Imagine 10,000 clients all make a request at the same time. The requests fail. If all 10,000 clients use exactly the same retry schedule, they might all wait 2 seconds and then retry at exactly the same time. The server suddenly receives another huge wave of requests. This is where <strong>jitter</strong> becomes useful. Jitter adds a small amount of randomness to the waiting period. So instead of every client retrying at exactly the same moment, the retries are spread out over a period of time. That small change can make a significant difference in distributed systems.</p><h4>When Should You Retry?</h4><p>Not every error deserves a retry. This is one of the things I found particularly interesting. A temporary network failure may be worth retrying. A timeout may be worth retrying. A server-side 500 error may be retryable depending on the situation. A 429 Too Many Requests response may also be retryable, especially when the API tells you when to try again using a Retry-After header. But consider a 401 Unauthorized response. Retrying the same request repeatedly without changing the authentication information probably won’t solve anything. The same applies to many 400 Bad Request errors. The application needs to understand <strong>why the request failed</strong>, not simply assume that every failure should be retried</p><h4>There Is Also a Limit</h4><p>A retry mechanism should not continue forever.</p><p>We normally define things such as:</p><ul><li>Maximum number of attempts</li><li>Maximum waiting time</li><li>Maximum backoff delay</li><li>Overall request deadline</li></ul><p>For example, an application might allow three attempts before finally returning an error. This prevents a temporary failure from turning into an endless process.</p><h4>What About Payments?</h4><p>This is where retries become even more interesting. Imagine a user clicks a <strong>Pay button</strong>. The payment request reaches the payment provider, but the response gets lost because of a network problem. Your application doesn’t know whether the payment succeeded or failed. If it blindly sends the payment request again, there is a possibility of charging the customer twice.</p><p>So retrying isn’t only about timing. It is also about understanding whether an operation is <strong>safe to repeat</strong>. This is why concepts such as <strong>idempotency</strong> are important when dealing with operations like payments. For example, an idempotency key can allow a payment provider to recognize that two requests are actually attempts to perform the same operation rather than two separate payments. That means the application can retry without accidentally creating duplicate side effects.</p><h4>Retry, Backoff and Timeout Work Together</h4><p>Another thing I learned is that these concepts don’t exist in isolation.</p><p>A well-designed request might have:</p><ul><li><strong>Timeouts</strong> to prevent waiting forever.</li><li><strong>Retries</strong> to recover from temporary failures.</li><li><strong>Backoff</strong> to avoid retrying too aggressively.</li><li><strong>Jitter</strong> to spread retries out.</li><li><strong>Circuit breakers</strong> to temporarily stop requests to a dependency that is consistently failing.</li></ul><p>Each one solves a slightly different problem.</p><h4>A Simple Mental Model</h4><p>The way I now think about it is:</p><p><strong>Retry answers:</strong><br>“Should I try this operation again?”</p><p><strong>Backoff answers:</strong><br>“How long should I wait before trying again?”</p><p><strong>Jitter answers:</strong><br>“How can I avoid everyone retrying at the same time?”</p><p><strong>Timeout answers:</strong><br>“How long am I willing to wait for this attempt?”</p><p><strong>Idempotency answers:</strong><br>“Is it safe to perform this operation again?”</p><p>That combination is much more useful than simply adding a try again loop.</p><h4>The Bigger Lesson</h4><p>Before this week, I might have thought of a failed API request as something that simply needed an error message or another attempt. Now I’m starting to see failures as something that needs to be <strong>designed for</strong>.</p><p>Real applications operate in environments where networks fail, services become unavailable, requests time out, and dependencies become overloaded.</p><p>The goal isn’t to pretend failures won’t happen. The goal is to build systems that can handle them without making the situation worse.</p><p>That’s probably the biggest takeaway for me:</p><p><strong>Good software isn’t just about what happens when everything works. It’s also about what the application does when things don’t.</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=51fee724aef8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Anatomy of a Good Prompt for Product Development]]></title>
            <link>https://medium.com/@gideonegbeyemi/the-anatomy-of-a-good-prompt-for-product-development-f110d043b39e?source=rss-63229f7361a1------2</link>
            <guid isPermaLink="false">https://medium.com/p/f110d043b39e</guid>
            <category><![CDATA[claude]]></category>
            <category><![CDATA[writing-prompts]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[prompt-engineering]]></category>
            <dc:creator><![CDATA[Gideon Egbeyemi]]></dc:creator>
            <pubDate>Sat, 29 Aug 2026 11:10:26 GMT</pubDate>
            <atom:updated>2026-08-29T11:11:28.983Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*H6Kcc-ZQfMLoOQhW" /><figcaption>Photo by <a href="https://unsplash.com/@berctk?utm_source=medium&amp;utm_medium=referral">Berke Citak</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>When I first started using AI to build products, I thought the process was pretty straightforward.</p><p>I would describe an idea, give the AI a command like:</p><blockquote>“Build me a school management system”</blockquote><p>…and expect something close to what I had in my head. Sometimes I got something impressive. Other times, I got something that technically worked but completely missed the point.</p><p>That experience made me realise something important:</p><p><strong>Building with AI isn’t just about knowing what you want to build. It’s also about knowing how to communicate it.</strong></p><p>As I continue learning about AI and product development. I have started to see a good prompt differently. A good prompt isn’t simply a question or instruction. <strong>It is a product brief. </strong>It gives the AI enough information to understand the problem, the users, the task, the boundaries, and the expected outcome.</p><p>In this article, I’ll break down the five things I am learning to include when writing prompts for product development.</p><h4>1. Give the AI a Personality</h4><p>One of the first things I learned was that I don’t always have to approach AI as if I’m simply asking it a question. I can tell it <strong>who I want it to act as</strong>.</p><p>For example:</p><blockquote>“Act as an experienced Product Manager”</blockquote><p>Or:</p><blockquote>“Act as a Senior Software Architect and Product Manager with experience building education technology products.”</blockquote><p>This gives the AI a role and perspective to work from. Why does this matter? It matters because different roles naturally focus on different things. A product Manager might think about: User needs, Product requirements, Features, User journeys, Business goals and Priorities while A Software Architect might focus more on: System architecture, Databases, APIs, Scalability, Security, Technology choices.</p><p>So instead of simply saying:</p><blockquote>“Help me build an app”</blockquote><p>I can say:</p><blockquote>“Act as a senior product manager. Help me turn this idea into a clearly defined MVP.”</blockquote><p>Giving AI a personality doesn’t mean pretending that it is actually a human with years of experience. For me, it’s simply a way of <strong>giving the model a useful perspective from which to approach the task</strong>.</p><h4>2. Give the Prompt Context</h4><p>AI needs context. Imaging telling someone:</p><blockquote>“Build a CBT platform”</blockquote><p>They might ask:</p><ul><li>Who is it for?</li><li>Is it for schools or universities?</li><li>Is it online or offline?</li><li>Who administers the exams?</li><li>How many students will use it?</li></ul><p>Without those questions, they have to make assumptions. AI does the same thing. So instead of:</p><blockquote>“Bui’d a CBT platform”</blockquote><p>I could write:</p><blockquote>“I am building a CBT platform for secondary schools. The platform will allow teachers to create examinations and students to take them. Schools should be able to conduct examinations both online and through a local network where internet access is unreliable.”</blockquote><p>Now the AI has something to work with. It understands</p><ul><li>What you are building: A CBT platform</li><li>Who you are building for: Secondary schools</li><li>Users: Teacher and Students</li><li>Purpose: Conduct examinations</li><li>Important requirement: Online and local network support</li></ul><p>A useful question to ask yourself before writing your prompt is:</p><blockquote>“If I gave this prompt to another person, would they understand what I am trying to build?”</blockquote><p>If the answer is no, your AI probably won’t either.</p><h4>3. Give the AI a Clear Task</h4><p>Another mistake I made early on was giving AI a general instruction and expecting it to figure out everything.</p><p>For example:</p><blockquote>“Help me build this platform.”</blockquote><p>That’s not really a task. It’s an invitation to guess. A better prompt would be:</p><blockquote>“Design the user roles, key features, user flow, and database structure for the CBT platform.”</blockquote><p>Now the AI knows exactly what I want it to work on. The difference might seem small, but it changes the quality of the output. You can even break larger tasks into smaller ones.</p><p>For example:</p><blockquote>“First, identify the different types of users the platform will have and define what each user should be able to do.”</blockquote><p>Then:</p><blockquote>“Now design the examination creation workflow for teachers.”</blockquote><p>Then:</p><blockquote>“Based on these requirements, propose the database entities and their relationships.”</blockquote><p>This approach has helped me understand something important. <strong>AI doesn’t always need one massive prompt containing everything. </strong>Sometimes, breaking a product into smaller, clearly defined tasks makes the process easier to manage.</p><h4>4. Include Constraints</h4><p>This is another part of prompting that I initially underestimated. A product doesn’t exist in a perfect world. There are always constraints. Maybe you have a specific technology stack. Maybe your users have limited internet access. Maybe the application needs to work on older computers. Maybe you have security requirements. Maybe you have a limited budget. These things matter.</p><p>For example, imagine I want to build the CBT platform using a specific technology stack.</p><p>I could tell the AI:</p><blockquote>“Build the application using React for the frontend, Node.js for the backend, and PostgreSQL for the database. The system must support both online and local-network examination modes.”</blockquote><p>Now the AI has boundaries. Instead of suggesting any technology it wants, it has to work within the requirements I provided.</p><p>Constraints can include:</p><ul><li>Tech stack: HTML, CSS, JS, React, Node.js, PostgreSQL, etc.</li><li>Platform: Web, Mobile, Desktop.</li><li>Connectivity: Online, offline, local network.</li><li>Security: Authentication, permissions, encryption.</li><li>Performance: Expected number of users.</li></ul><p>Constraints aren’t necessarily restrictions that make product development harder. <strong>They help define what the product actually needs to be.</strong></p><h4>5. Define the Expected Output</h4><p>This is probably one of the easiest things to forget. You can give AI a great task and still receive an output that isn’t useful because you never explained <strong>what you want the result to look like</strong>.</p><p>For example, I could say:</p><blockquote>“Analyse this product idea.”</blockquote><p>That’s extremely open-ended. Instead, I could say:</p><blockquote>“Analyse the product idea and provide the response as a Product Requirements Document containing the product overview, target users, user problems, core features, user stories, technical requirements, database structure, and MVP roadmap.”</blockquote><p>Now I know what I’m expecting to receive. The important thing is to <strong>tell the AI what form the outcome should take</strong>.</p><h4>Let’s combine the five elements.</h4><p>Instead of writing:</p><blockquote>“Build me a school management system.”</blockquote><p>I could write something more like:</p><blockquote><strong>Act as an experienced Product Manager and Software Architect.</strong></blockquote><blockquote>I am building a school management platform for secondary schools. The platform is intended to help schools manage students, teachers, classes, examinations, results, and administrative activities.</blockquote><blockquote><strong>Your task:</strong> Define the MVP for this product. Identify the primary user roles, core features, user journeys, and key workflows.</blockquote><blockquote><strong>Constraints:</strong> The application should be web-based and responsive. Use React for the frontend, Node.js for the backend, and PostgreSQL for the database. The system should use role-based access control and be designed to support multiple schools in the future.</blockquote><blockquote><strong>Expected output:</strong> Provide a structured Product Requirements Document containing the product overview, problem statement, target users, user stories, feature requirements, technical requirements, database entities, MVP priorities, and suggested development roadmap.</blockquote><p>Now the AI has much more information. It knows: Who to act as, What we’re building, Who it’s for, What to do, What limitations exist, and What the output should look like. I’m still very much a beginner when it comes to AI and product development but one thing I’ve started to understand is that <strong>AI is only as useful as the direction we give it</strong>. A good prompt doesn’t guarantee a perfect product. But a clear prompt gives you a much better starting point.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f110d043b39e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Designing Vantage: A tool for “Online Vendors”]]></title>
            <link>https://medium.com/@gideonegbeyemi/designing-vantage-a-tool-for-online-vendors-bb8b111ad7b1?source=rss-63229f7361a1------2</link>
            <guid isPermaLink="false">https://medium.com/p/bb8b111ad7b1</guid>
            <category><![CDATA[ecommerce]]></category>
            <category><![CDATA[figma]]></category>
            <category><![CDATA[product-design]]></category>
            <dc:creator><![CDATA[Gideon Egbeyemi]]></dc:creator>
            <pubDate>Wed, 03 Apr 2024 21:32:00 GMT</pubDate>
            <atom:updated>2024-04-03T21:32:00.059Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KCnszqTxnBGPC5SEJP-f-g.png" /></figure><p>Swiping through two or more mobile apps just to keep up with orders from your customers, frequent rifts with delivery agents, and difficulty measuring your business’s profitability are definitely a lot.</p><p>Vantage is an e-commerce tool designed to streamline the online selling experience for vendors operating primarily on Facebook, Instagram, and WhatsApp. With the surge in social media as a platform for commerce, there is a growing need for a comprehensive solution that integrates seamlessly with these platforms, empowering vendors to efficiently manage orders, deliveries, and transactions. Vantage aims to fill this gap by providing a user-friendly mobile application that caters specifically to the needs of online vendors.</p><h4>Introducing Vantage</h4><p>Vantage is a two-way mobile app for online vendors and delivery agents. The app aims to simplify the management of online businesses.</p><h4>The challenge with regards to running an online business</h4><p>The rise of social media as a marketplace has presented challenges for online vendors in effectively managing their sales operations. Without a dedicated tool, vendors often find themselves overwhelmed with manual processes for receiving orders, coordinating deliveries, managing finances, and sending invoices and receipts. This results in inefficiencies, errors, and a poor customer experience. Vantage seeks to address these pain points by offering a centralised platform that simplifies and automates these tasks.</p><p>Online vendors often have difficulties successfully managing their businesses, particularly when dealing with orders received from multiple social media channels. Working with delivery agents is always tough, and inventory management can also be time-consuming. We designed vantage, an integrated environment for internet enterprises.</p><h4><strong>Gathering Insights</strong></h4><p>In our research with 12 (twelve) “online vendors,&quot; we needed to look into a couple things right away:</p><ul><li>Why do SMEs move their businesses online?</li><li>What kind of struggles do they have?</li><li>What can we modify without altering their productivity?</li></ul><p>We identified the following pain points after synthesising the data from the interview:</p><ul><li>Order tracking and fulfilment.</li><li>Finance management.</li><li>Inventory management.</li><li>Managing orders from multiple social media accounts.</li><li>Customer relations.</li><li>Receipt and Invoice management</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OCqIaETfSHwFZ3qhkU0j9A.png" /></figure><p>During the research phase, we also conducted competitive analysis to build on existing solutions and identify gaps to address.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/1*rn5FnBHq9ayPtRA8MMifUw.png" /><figcaption>Competitive Analysis</figcaption></figure><p>Engaging in firsthand experiences with some of these applications enabled the team to document successful practices, aligning with our current understanding of the reasons behind their choices and how they had executed them.</p><h3><strong>Designing Vantage 🎨</strong></h3><h4>Orders, order fulfilment and tracking</h4><p>In the realm of e-commerce, order fulfilment stands as a critical bridge between the virtual storefront and the customer’s doorstep. It’s the pivotal moment where promises are either kept or broken, customer satisfaction solidified or lost.</p><p>After a customer places an order from a sales channel (store website, WhatsApp, Facebook, or Instagram), it appears on the orders tab in the vendor’s app, and a notification is also sent. Orders can also be created manually for orders made physically or outside the sales channel.</p><p>When the payment status has been confirmed, the next step involves preparing and delivering the item. This process is referred to as order fulfilment. After the order has been packaged and is ready for delivery, the vendor clicks on the “Mark as ready for delivery” button. Subsequently, the order appears on the courier app, allowing the delivery agent to proceed with delivering it to the customer.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*s1DV0l6-ggKVOoQTJHrdnQ.png" /></figure><p>To achieve a seamless order fulfilment process, we designed a courier mobile app where trusted delivery agents are onboarded.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nAx2nqrXi4K3UBg6jEd_QA.png" /></figure><p>To ensure customers receive the correct packages and to prevent false complaints of non-delivery, we designed a pin feature for precise delivery to the intended recipient. Upon delivery, the customer provides a 4-digit PIN, which is subsequently confirmed by the delivery agent.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lVGOYjphABt0d2Z14F6hQg.png" /></figure><h4><strong>Finance Management</strong></h4><p>Vantage Wallet provides entrepreneurs with a transparent, trustworthy, and easy-to-use money management tool tailored to the needs of businesses. Payments for orders are paid to this wallet, allowing vendors to separate personal and business finances.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xzylOhfe6DShBXPZRW3brA.png" /></figure><p>The Finances Overview page offers a detailed breakdown and analysis of your store’s financial performance within a chosen date range, presented through a series of cards and an exported PDF document.</p><h4>How customers pay on Vantage</h4><p>Payments are made through payment links and bank transfers. Customers’ payments are deposited into the vendor’s Vantage wallet.</p><h4>Inventory Management</h4><p>Inventory tracking serves as a crucial tool in preventing the sale of products that are out of stock and in notifying you when it’s time to reorder or produce more of a particular product.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*A8T7QiQM3sCs9pu-TzjBxg.png" /></figure><h4>Receipt and Invoice Management</h4><p>Receipts serve as your customers’ evidence of their purchases and the corresponding amounts paid.</p><p>Receipts and invoices are automatically sent to customers. However, if vendors need to send them manually, they must select the particular order for which they wish to send an invoice or receipt. At the bottom of the screen, a “Send invoice” button appears if the payment status reads &quot;unpaid,” and a “Send receipt” button appears if the payment status reads “paid.” The customer receives the receipt or invoice from the sales channel (social media app) through which they ordered the product.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HxNqUmWzYOlET2I_xm-l8A.png" /></figure><h4>Managing Multiple Social Media Apps</h4><p>To reduce the burden of managing chats across multiple social media apps and increase response time, we designed a chat feature in the Vantage mobile app. This feature enables vendors to consolidate messages from WhatsApp, Instagram, and Facebook into a single app. Vendors can seamlessly record sales, send automated messages and updates on orders, and streamline the process of sending invoices and receipts.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9LiEHqZRtht99kucNn86iQ.png" /></figure><h4><strong>Customer Relation</strong></h4><blockquote>The customer is king</blockquote><p>Establishing stronger customer relationships may seem daunting at first. Nevertheless, we designed Vantage to streamline customer relations for online-based SMEs in the following ways:</p><ul><li><strong>Effective communication</strong> through automated messages and notifications.</li><li><strong>Marketing and campaigns</strong>.</li><li><strong>Rewarding loyalty</strong> through discounts and coupons.</li><li><strong>Excellent customer service </strong>through chat bots.</li></ul><h4><strong>Effective Communication</strong></h4><p>According to our research, customers appreciate being kept in the loop. They value updates on their orders, including delivery status and notifications when their preferred products become available.</p><p>To ensure effective communication between vendors and customers, we designed a campaign feature and an automated messaging tool where vendors can communicate effectively with their customers.</p><h4><strong>Marketing and Campaigns</strong></h4><p>Marketing campaigns are structured sets of marketing activities aimed at promoting a product or sale within a specific timeframe, typically leveraging channels such as paid advertising and social media.</p><p>These channels are a great way to share new products, discounts, and deals, keep in touch with your customers, and reach potential new customers.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*x-8ZKZXdlbO7Srw9vHw7yg.png" /></figure><h4><strong>Rewarding Loyalty</strong></h4><p>From our research, customer loyalty programmes offer an effective means to enhance relationships with customers while simultaneously boosting revenue. A robust loyalty programme holds significance for both longstanding clients and newly acquired customers. By offering carefully curated discounts, rewards, and promotions, vendors can provide valuable benefits to those who consistently support their business.</p><p>We designed the Vantage discount &amp; coupon generator. Discounts are implemented through two methods: utilising a discount code and activating an automatic discount. With the discount code, customers must input a unique code manually, whereas automatic discounts are applied when customers fulfil specific conditions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dN76xDwngfDJ1sGk4S9Qlg.png" /></figure><h4>Outcome</h4><p>This project was incredibly thrilling and enjoyable for me to be a part of, as it delivered tangible value, required extensive research, and involved intricate interaction efforts. From this experience, I gleaned several crucial insights regarding product and business processes, including:</p><ul><li><strong>Communication and Collaboration:</strong> Effective communication and collaboration among cross-functional teams were critical to the success of the project.</li><li><strong>User-centric design is key.</strong> Throughout the entire process, it became evident that understanding the needs, behaviours, and pain points of the target users is paramount. By prioritising user-centric design principles, we were able to create a product that truly resonated with our audience and addressed their specific requirements.</li><li><strong>Iterative Design Leads to Improvement:</strong> Embracing an iterative design process allowed us to continuously refine and enhance the product based on user feedback and testing results. Each iteration brought us closer to our goal of creating a seamless and intuitive user experience.</li></ul><p>Through a comprehensive design process focused on understanding user needs and addressing pain points, Vantage has successfully delivered a solution that empowers online vendors to thrive in the competitive world of social media commerce. With its intuitive interface, seamless integration, and powerful automation features, Vantage is poised to become the go-to tool for online vendors.</p><p><em>Hello, I’m Gideon, dedicated to crafting exceptional experiences and designing solutions that align your business strategy with the needs of your customers. For further discussion, feel free to </em><a href="https://twitter.com/geattweets"><em>connect with me on Twitter</em></a><em>. Additionally, you can explore more of my work in </em><a href="https://linktr.ee/GideonAyomide"><em>my portfolio.</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bb8b111ad7b1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[FindDAO — The Case Study]]></title>
            <link>https://medium.com/@gideonegbeyemi/finddao-the-case-study-f0496e56790e?source=rss-63229f7361a1------2</link>
            <guid isPermaLink="false">https://medium.com/p/f0496e56790e</guid>
            <category><![CDATA[defi]]></category>
            <category><![CDATA[blockchain]]></category>
            <category><![CDATA[dao]]></category>
            <category><![CDATA[product-design]]></category>
            <category><![CDATA[web3]]></category>
            <dc:creator><![CDATA[Gideon Egbeyemi]]></dc:creator>
            <pubDate>Sat, 15 Oct 2022 16:16:47 GMT</pubDate>
            <atom:updated>2022-10-15T16:19:33.852Z</atom:updated>
            <content:encoded><![CDATA[<h3>Designing FindDAO — The Case Study</h3><p>A solution for newbies getting started with DAOs.</p><figure><img alt="Mockup for FindDAO Hero Section." src="https://cdn-images-1.medium.com/max/1024/1*u5V-NImbJiwM-GhM5KQXFw.png" /></figure><p>Role: UX Designer UI Designer | Project time frame: 2 weeks</p><h4><strong>Overview</strong></h4><p>The web 3 space is evolving at a rapid pace with a lot of concepts to keep up with. One of these concepts is the <strong>Decentralized Autonomous Organization (DAO).</strong></p><p>A DAO is an entity owned and managed collectively by its members with no central authority governing the organization. These organizations are built on a <a href="https://www.ibm.com/topics/what-is-blockchain">blockchain</a> using <a href="https://www.ibm.com/topics/smart-contracts">smart contracts</a>. Membership is mostly based on owning a governance token that allows members to create proposals and vote on decisions.</p><p>FindDAO is a design solution that helps newbies get started with DAOs.</p><p><strong>The Problem</strong></p><p>The future of traditional organizations is DAO. However, this future won’t be a reality if users struggle to take part in it.</p><p>As a newbie, Finding DAOs to join and knowing what these DAOs represent or seek to achieve can be overwhelming.</p><blockquote>You can’t give what you don’t have</blockquote><p>Newbies tend to join and contribute to DAOs that does not align with their interest or skill sets which place them at crossroads on how best they can serve.</p><h4>What do we seek to achieve?</h4><p>The goal of this project was to create a distinctive solution, focusing on three key features:</p><ul><li>Help our users find and join DAOs that aligns with their interest and skill sets.</li><li>Help our users with a clear path on how to contribute to DAOs.</li><li>Our users should be able to gather basic knowledge about some DAO’s terminologies.</li></ul><h4>Research O’clock</h4><p><strong>Understanding how to get started with a DAO.</strong></p><p>I needed to first understand what DAOs are, what they entail, and how they work. To kick things off, I started doing some secondary research to get a hold of the necessary steps to take to join a DAO and how to contribute to the DAO. From this exploration, I was able to identify how users contribute to a DAO.</p><ul><li>Investing in DAOs by buying their tokens.</li><li>Vote and help decide the future of the DAO.</li><li>Creating proposals to share ideas on how to accomplish a DAO’s project or goals.</li><li>Offering to work for these DAOs as a developer, designer, content creator, writer, etc.</li></ul><blockquote>Reading reports and articles is more work and less fun than talking to live humans and hearing about their specific situations.</blockquote><p>To gain more insights and validate my secondary research, I also conducted primary research (interview) with a newbie in the DAO space and also a user with vast experience with DAOs.</p><p>The major aim of the primary research was to identify pain points that will inform my design and find answers to these questions.</p><ul><li>How do users go about finding DAOs to join?</li><li>What are the current measures in place to help users find and contribute to DAOs?</li><li>What are the steps to take to join a DAO?</li><li>How do users contribute to DAOs?</li><li>How do users know the goals of DAOs they are interested in?</li><li>Where might newbies get lost in their quest of finding DAOs to join?</li></ul><h4>Findings</h4><p><strong>Some key notes from the user interview conducted.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hQrupZOfAtdzBdqEjqmWkA.png" /></figure><h4>Setting things straight</h4><p>By combining ideas from primary and secondary research, I was able to reach the following conclusions about my users.</p><ul><li><strong>Easy Search: </strong>Newbies want an easy way to look for DAOs to join and also get adequate information about what is being offered by the DAO.</li><li><strong>Learn: </strong>Newbies need a platform to learn about terminologies that seem bogus, learn about how best they can contribute to DAOs and where to ask questions.</li><li><strong>Interest: </strong>Newbies want to be able to join DAOs that match their interests or skill set.</li><li><strong>Noise: </strong>Newbies have difficulty in distinguishing important information from background noise on Discord.</li></ul><h4>Exploring existing solutions</h4><p>To investigate current solutions and build upon them by examining their strengths and weaknesses, I conducted a competitive analysis.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Jr637747pOyTj75wQrmKMg.png" /></figure><h4><strong>Building Personas</strong></h4><p>I decided to create a user persona to keep my designs user-centric.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jUJEgKwBUl96WgsN1tjk2A.png" /></figure><p>Since most users using this product are newbies, I mapped out a user journey map to highlight the steps users will take to attain their goals and create obstacle-free paths for users, highlighting new pain points and identifying improvement opportunities.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QPNQrpA_2dg9WjLuMTh2-w.png" /></figure><h4>Ideating and Proffering Solutions</h4><p>To start the ideation phase of my design process, I used How Might We statements to turn problems into opportunities.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LbkBEuHxW-xUxLElq-jPrA.png" /></figure><p>At this point, I also developed a broad set of solutions to address the key insights identified in the research data.</p><p><strong>Easy Search</strong></p><p>To solve the problem of looking for DAOs to join;</p><ul><li>Design a platform where users can search for DAOs to join.</li><li>Allow users to narrow their search based on the type of DAOs they want, the type of membership they want, and the network a DAO is built on.</li><li>Quick access to DAO socials, community and whitepaper.</li><li>A clear guide on how to get started.</li></ul><p><strong>Learn</strong></p><ul><li>Design an online dictionary where users can look up the definition of unfamiliar words.</li><li>A blog where users can read articles.</li></ul><p><strong>Interest</strong></p><ul><li>A brief survey of users’ skill sets and interests, as well as the types of DAOs they are interested in.</li><li>The survey’s findings should be used to recommend DAOs to users.</li></ul><p><strong>Noise</strong></p><ul><li>Design a notification panel in which users will be notified of important information, community calls, and in-demand skills, as well as updates on new and completed proposals.</li></ul><h4>Introducing FindDAO</h4><p>One-stop shop to explore the world of DAOs.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hWzagPtDuWwO9n1XxdEVDQ.png" /></figure><h4><strong>Onboarding our users</strong></h4><p>To recommend the best DAOs for our users and narrow their search, I created some modals that serve as onboarding screens. These modals introduce users to DAOs, the major ways they can contribute to a DAO, and collect some information so we can recommend DAOs that match their interests and skill sets.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8q0HEAqpdm0ExpXHTPJzFA.png" /></figure><h4>Explore DAOs</h4><p>The results of the onboarding process are used to create the explore page. Users can view the recommended DAOs on this page and decide which DAOs are of interest to them.</p><p>I also had to consider the possibility that users are dissatisfied with the recommended DAOs. To address this, users should still be able to narrow their search by DAO categories, type of membership, and network on which the DAOs are built.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vsJqnBSV1y2hVfFF96ALUQ.png" /></figure><h4><strong>Inside the DAO section</strong></h4><p>The DAO section allows users to</p><ul><li>Access necessary information about a DAO, what they seek to achieve and a clear path on how to get started.</li><li>Perform actions like creating proposals, viewing proposals, and voting on proposals.</li><li>Access DAO bounties.</li><li>Access DAO socials, website, community and the whitepaper.</li><li>View the performance of a DAO token and also be able to claim the token.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qQgkO_1s6GQvy_dzNoW-6Q.png" /></figure><h4>Filtering the noise</h4><p>I designed a notification panel where the necessary information from a DAO discord channel will be accessible from this panel to suggest content that is essential to users and ensure they are up to date with the latest happenings in their preferred DAO.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CQj5bEnRJ76Z3XM9AF1Big.png" /></figure><h4><strong>Learn about DAOs</strong></h4><p>The web 3 space is cluttered with a lot of words that are not newbie friendly. To help users find meanings to terms that might seem bogus I designed a dictionary section for FindDAO for users to explore meanings. There are also blog sections where users can explore articles to read.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*baWIqJ9t2kdWPSlIHHessw.png" /></figure><p><strong>View the FindDAO prototype </strong><a href="https://www.figma.com/proto/VobxnbMOJsclLpNryPgC83/Druids-Design-Challenge?page-id=226%3A3713&amp;node-id=430%3A4007&amp;viewport=-6284%2C-1450%2C0.22&amp;scaling=scale-down-width&amp;starting-point-node-id=430%3A4007"><strong>here</strong></a></p><h4>Final Thoughts</h4><p>This is one of the best projects I have worked on. It helped me understand how complex the DAO space can be for newbies. Designing a solution for this complexity newbies face with DAOs gladdens my heart.</p><p>Working on this project also confirmed my perception that thorough research forms the basis of excellent designs.</p><p><strong>Thank you for reading.</strong></p><p><strong>If you’d like to reach out, you can connect with me via </strong><a href="https://twitter.com/geattweets">Twitter</a> <a href="https://www.linkedin.com/in/gideon-ayomide-614b47231/">LinkedIn</a> <a href="https://t.co/smiJfUWuc8">dribble</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f0496e56790e" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>