POLISHING UI/UX
Want to back the stream? The «داعم» membership on YouTube is EGP 60 a month: badges, emoji, and I answer your questions first on stream. Become a supporter
The gist
The stream follows a Shopify app moving from its core build into the harder pre-launch phase of polish and verification: UI/UX review, turning findings and screenshots into tasks, and manually exercising the app alongside AI agents. Luqman presents a command center that divides work among the human operator and agents, balancing parallel execution against file conflicts, and proposes teaching a lighter version from scratch. Practical investigation of slow loading and navigation occupies a major part of the work; routing, repeated database calls, Prisma, and network distance are considered, and some improvement is observed, but the complete cause and final fix remain unresolved. App-listing requirements, privacy policy placement, GDPR, Railway deployment, and the path toward Shopify submission are also discussed as unfinished work or plans rather than confirmed outcomes. Other useful discussions cover competitor research and positioning in a low-competition market, subjective tradeoffs among Codex, Claude, and other models, usage limits and subscription ROI, plus audience questions about Arabic dictation, email outreach, and channel growth through short-form video. The stream also includes attributed views on AI and cybersecurity and on how operator expertise shapes the value of AI tools.
73 moments from the stream
- 0:50:25
Codex and Claude Tradeoffs in Coding Sessions
The speaker compares his experience with both tools, describing Codex as stronger overall at the time while finding Claude easier to work with and better at handling conversational context; he emphasizes that each has pros and cons.
- 0:53:48
Moving from App Construction to Polish
After the app progressed faster than expected, the speaker explains that UI/UX polish and debugging are becoming the harder phase, so he created a dedicated system and milestone for that work.
- 0:57:33
Turning UI/UX Findings into Review-and-Polish Tasks
The speaker outlines a workflow for reviewing agent output before discrepancies from his standards accumulate: he points out problems, has them identified as UI or UX issues, and turns them into specific review-and-polish tasks for a subsequent process.
- 1:00:28
Arabic Voice Dictation with Wispr Flow
The speaker recommends voice dictation instead of manually typing prompts, noting that Wispr Flow supports Arabic, can be used across several messaging apps, and cleans up verbal hesitations during transcription. Another tool is mentioned, but its name is not clear enough to spell reliably.
- 1:03:58
How Shorts Support YouTube Channel Growth
The speaker discusses using Shorts, Reels, and TikTok to reach new viewers and subscribers. In his view, livestreams draw a smaller but more engaged audience interested in building and AI-assisted coding, while regularly publishing short-form videos can help grow a channel.
- 1:04:36
A Lighter Agent Workflow for UI Review and Debugging
The speaker designs a purpose-built workflow for UI polishing and debugging: a diagnostic agent reviews screenshots and relevant code, classifies the issue as UI, UX, or backend, and prepares the appropriate prompt file before a builder performs the task. For lightweight polish work, he proposes skipping research, security, and compliance agents and retaining only a build check and visual verification. Aesthetic judgment remains with the user, while the validator checks that the correct files changed and presents result screenshots for faster feedback. He also wants this workflow available when using Codex, with one or two review rounds expected.
- 1:11:58
Competitor Research and Positioning for a Shopify App
Researches potential competitors in the Shopify App Store and evaluates functional fit rather than superficial similarity. The discussion describes Shopify Sidekick as answering questions within a store, while the proposed concept aims to combine questions with taking actions. It also considers the risk of finding few direct competitors, the need for clear market positioning, and learning from the settings and plan sections of a successful app without claiming that a full analysis was completed.
- 1:13:52
Choosing a Tool for Bulk Email Outreach
Answers an audience question about sending emails that direct companies to a specific website. The advice begins with confirming that the email list exists and understanding its size, then choosing tools accordingly instead of defaulting to larger, costlier platforms. Mailchimp and Klaviyo are mentioned as established options, while simpler or free alternatives may suit small lists; LinkedIn is suggested as an additional outreach channel. A tool being tried is presented only as an initial option, not a definitive recommendation.
- 1:21:54
Finding a Clean App UI Reference
Exploring apps for a clean interface that could inspire a first app UI, including opening one option to inspect its internal design.
- 1:26:41
Turning UI Screenshots into a Structured Task
A workflow sends screenshots and instructions to an orchestrator agent, which analyzes the reference and creates a reviewable task for execution in a separate session. The aim is to use explicit preparation and start steps for more reliable work and fewer revision loops, even if that requires more model usage.
- 1:28:14
The Presenter’s Note on Claude Usage and a Codex Limitation
At the time of the recording, the presenter felt that Claude consumed tokens quickly and said Codex lacked MCP write access in the workflow setup shown.
- 1:30:08
Reviewing the Chat UI Refactor Task Scope
The generated task scoped a chat UI refactor around the reference, covering layout, the empty state, quick-action pills, and the input area; the presenter then tried to locate the task on the board.
- 1:30:48
Testing a UI Review and Polish Workflow
A trial of a review-and-polish workflow, assessing whether the tool can research the requested UI direction and prepare a stronger implementation prompt from a reference interface instead of requiring every change to be specified manually.
- 1:34:45
Diagnosing Slow Application Navigation
A discussion of delays when moving among chats, plans, and settings. The diagnostic report pointed to two possible causes: custom navigation elements instead of React Router links, and multiple database calls, while suggesting authentication admin-session storage latency as the next area to inspect. The app remained slow after a refresh, so no fix was confirmed.
- 1:40:31
Chat Home Being Skipped
An initial observation that the app does not show the chat home and instead opens directly inside a conversation; the caption ends before the behavior or its cause is fully described.
- 1:40:53
Diagnosing automatic navigation and slow initial loading in a chat screen
A practical debugging discussion about a chat screen that briefly shows its home view and then automatically enters a conversation. The investigation traced the behavior to the navigation mechanism while also examining slow first-load performance, caching, and lazy loading. Route-level lazy splitting was reported as already present, so adding more lazy loading was not considered the main fix; the chat screen was also noted to lack a render window that limits visible items. Some performance interpretation remained under investigation.
- 1:51:03
Copyright Caution During Livestreams
He avoids playing certain content during the livestream because it could create a copyright issue, despite finding it interesting. The remark highlights the need to separate reviewing content from publicly showing it when usage rights are unclear.
- 1:52:29
From Generated UI Prototype to Agent-Implemented Code
Luqman explores a tool that generates a UI with multiple conversations and understands the workflow as shaping the concept there, then handing its code to a coding agent for integration. He discusses copying, deployment, and distributing tool usage, but remains unsure whether the entire application code can actually be copied or exported.
- 1:53:45
Scheduling Streams Around Tool Load and Audience Availability
Luqman considers moving streams to later hours, citing a notice attributed to Claude about heavier demand during a certain period and his general observation that Arabic-speaking viewers are more active at night. He presents this as a possible experiment to improve interaction and avoid peak token usage, not a final decision.
- 1:55:17
Identifying Application Load and Navigation Latency
He briefly pauses the UI work to flag a fundamental performance issue: opening the application and moving between navigation sections have noticeable latency.
Got a question about something in this stream?
Ask in the group