Building AI Apps Live from Cairo 🔴
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 centers on time-boxed polish for a terminal-themed website: reviewing visual hierarchy, reducing clutter, and iteratively refining a “code rain” effect—its motion, density, visibility, and content—through precise feedback, screenshots, and component references. Luqman argues that coding agents need specific direction and should understand a project before editing it, then explores an Explorer–Planner–Builder workflow with automatic context assembly, prompt previews, and verification before handoff. The value of a fully automated pipeline, exploration quality, CLI task resumption, and unexpected behavior in plan mode remain unresolved. Other discussions cover Nightbot and social-distribution automation, an unfinished YouTube API integration for live status and schedules, and pre-launch marketing through a clear landing page and waitlist while distinguishing a global product market from an Arabic knowledge audience. Audience topics include practical streaming advice, choosing web or mobile development according to the project, reading AI benchmarks cautiously, and brief questions on core AI concepts. Preferences involving Claude, Codex, Cursor, and Warp are presented as personal judgments, and the stream does not establish that every proposed improvement or test was completed.
63 moments from the stream
- 2:10:58
An Invisible Effect Is Functionally a Failed Effect
A practical UI-review observation: if an effect cannot be perceived, it is not serving its purpose and may effectively be disabled.
- 2:12:26
Lighter Streams for Discussion, Questions, and Peer Learning
A tentative idea for some future streams to use a lighter format centered on discussion, questions, and peer learning for an audience with varied levels of experience. It was presented as a possibility, not a confirmed plan.
- 2:27:02
Refining the Direction and Content of a Code-Rain Effect
Luqman clarified that the effect should show horizontal lines of code falling from top to bottom rather than moving sideways, using English programming-related text such as HTML elements. After reviewing the result, he found the motion acceptable but said the token list still needed tuning into fuller code lines with a strong mix of HTML and React, with density and speed available for adjustment.
- 2:32:18
Automating an Explorer–Planner–Builder Agent Pipeline
The discussion identifies weaknesses in an explorer agent that starts from a thin task prompt without enough context about directories, patterns, or scope, while the current workflow relies on manual dispatch and copying output between stages. Proposed improvements include automating the Explorer-to-Planner-to-Builder flow, passing structured summaries instead of truncated output, optimizing the explorer and selecting an appropriate model for it, and allowing it to search the internet for some tasks as well as the internal codebase.
- 2:37:18
Refining a Code Animation for an Egyptian Audience
Requested refinements to an animated code display include slightly increasing its speed, preventing lines from overlapping, and mixing HTML-, React-, and hook-related content with subtle, humorous Egyptian Franco-Arabic terms to make it more relatable to Egyptian viewers.
- 2:41:02
Automating a Multi-Agent Pipeline
A discussion of replacing manual handoffs among exploration, planning, and building agents with an automated sequential pipeline that feeds each stage into the next. The proposed design includes live status and elapsed-time reporting, progress events across processes, cancellation, and per-step timeouts. It also argues that strong exploration—including findings beyond the codebase—shapes downstream quality, while acknowledging that web search adds latency.
- 2:43:29
Improving a Web Page’s Initial Visual Impact
A review of falling-line animation on a web page: the lines should appear immediately on load, begin around the middle of the visible fold rather than the very top, move slightly faster, and use sharper, more polished Franco-Arabic phrasing.
- 2:52:02
Dividing Web Development Tasks Between Codex and Claude
The speaker describes his personal workflow: Claude handles the main build while Codex polishes the UI. He recommends first asking the assistant to read the project and explain its understanding before making changes.
- 2:54:28
Iteratively Refining an Animated Visual Effect
A practical iteration on an animated code effect: increasing speed and density, starting it immediately on page load across the full height, then keeping elements visible while adjusting the emphasis of parts of the effect.
- 2:56:22
How Should AI Model Benchmarks Be Read?
The speaker discusses the difficulty of interpreting model rankings and the need to understand measures such as intelligence and speed before choosing a model. He states a personal preference for Claude while using GPT for some tasks, without claiming certainty about the cited comparisons.
- 3:01:47
Refining a Code-Themed Interface
The interface is refined with varied code-line sizes and more visual space, while witty Franco-Arabic lines receive a stronger green highlight and most of the content retains an HTML-and-React feel.
- 3:04:45
Planning a Project with a Swimlane
An explanation of respecting timelines and building a project map that shows what should be achieved across successive weeks. The speaker considers a swimlane the essential tool for most people, while task-board automation and an agent hub are optional additions suited to some workflows.
- 3:09:19
How to Choose Between Web and Mobile Development
Responding to a beginner’s question, the speaker recommends choosing according to the project and researching the idea and its viability rather than following a universal recommendation. Web development is described as easier to start and publish, while mobile apps must meet app-store submission requirements and checklists.
- 3:12:06
Refining an Animated Interface with Precise AI Feedback
He chose to focus on desktop first because the mobile concept was not yet convincing, then requested varied animated line lengths, wider side coverage, and deliberate highlighting of Franco-Arabic phrases instead of random accents. To reduce misunderstandings, he asked the assistant to restate its interpretation before making changes.
- 3:15:57
Aiming to Automate the Social-Media Workflow
He felt the thumbnails looked too generic and needed improvement, and expressed a preference for automating the social-media workflow as fully as possible so it would require little ongoing attention.
- 3:17:47
Exploring a CLI-Based Task Workflow
He discussed a tentative understanding that a prompt-pipeline mode might require placing a CLI inside a command center. Questions remained about back-and-forth interaction and resuming a task, so this was an unresolved workflow exploration rather than a demonstrated success.
- 3:19:19
A Clear, Clean Pre-Launch Landing Page
He judged the current design too busy and inelegant, while treating the work as useful web-design practice for a later landing page. Because that page would support a pre-launch campaign, he wanted it to remain very simple, crystal clear, and polished.
- 3:20:22
Reducing Visual Noise in Line Animation
He identified two concrete animation problems: overlapping lines and too many lines changing at once. He asked to prevent overlaps and reduce simultaneous changes for a calmer visual result, while leaving speed and dimensions unchanged according to the clearest captioned instruction.
- 3:26:25
Directing Codex to Refine a Dynamic Design and Highlight Franco-Arabic Lines
Luqman reviewed an interface with sections moving at different speeds and asked for ordinary lines to use lower opacity while witty Franco-Arabic lines stood out. When the implementation failed to identify them and highlighted unrelated lines, he requested a return to the previous version.
- 3:28:40
Brief Discussion of an OpenRouter Wrapper
Luqman reacted positively to an audience project described as a wrapper using OpenRouter and said he would check it out. The captions do not provide enough technical detail about its implementation.
Got a question about something in this stream?
Ask in the group