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 combines practical work and early-stage exploration around AI-assisted building and agent coordination: taking a site live after moving DNS to Cloudflare and discussing the expected GitHub-to-Cloudflare deployment flow, then expanding an Electron swimlane timeline into a unified command center for roadmaps, tasks, dependencies, and agent state. Luqman discusses a phased, partly parallel implementation using Markdown plans, shared JSON state, and a human-review queue, while deferring OpenClaw integration until the core tool is ready. Separate-project troubleshooting established that the Electron tool had been created apart from the main project, with a tracker described as the bridge between them; reference documents were intentionally left at their canonical paths to avoid diverging copies. The stated development stack includes Supabase, Railway, Cloudflare, and Codex. Other discussions cover dividing work between Claude and Codex, context-window consumption, Warp, local versus VPS operation, usage limits, and the API alternative. The stream also describes an automated clipping workflow, with possible later redistribution on platforms such as TikTok. Audience questions touch on Egyptian dialect support and coding tools, while broader topics include an English-first interface with selective Arabic cues and a vision for growing an Arabic vibe-coding community through transparent live building. Luqman argues, as an opinion, that easier building does not ensure shipping without marketing, distribution, and profitability knowledge. Several comparisons remain unresolved, and the command center remained an unfinished concept and prototype.
37 moments from the stream
- 0:03:58
Taking a Site Live After Moving DNS to Cloudflare
The speaker reports that the DNS move from GoDaddy to Cloudflare completed and the site became live on its domain. The project name is not reliably clear in the captions.
- 0:04:46
An Execution Timeline from a Markdown Roadmap
The proposed Electron app converts a Markdown roadmap into a unified swimlane view of progress, remaining work, schedule, and dependencies. The speaker also wants the dashboard connected to the project so AI agents can follow the same plan, with Claude and Codex assigned according to the kind of work.
- 0:10:51
Choosing Arabic and English Priority for a Technical Site
The speaker discusses a site aimed mainly at Egyptian and regional audiences and weighs whether Arabic or English should come first. Although the stream is in Arabic, interaction with AI agents and much of the working terminology are in English, leading him to consider an English-first interface that also supports Arabic.
- 0:12:22
Splitting Website UI Work Across AI Tools
The speaker describes a workflow in which Claude produced an initial website skeleton and Codex was then used for visual UI refinement. He presents this as a practical experiment in matching tools to stages of the work; the quality assessment remains his own opinion.
- 0:15:29
Briefing an AI Agent to Reverse a Website’s Language Priority
Explains a prompt asking an AI tool to review a website UI in depth and change it from Arabic-first to English-first, while retaining an Arabic presence, possibly in the logo as well. Execution details were left to the tool, and the task was considered relatively limited in context scope.
- 0:18:56
The Need for More AI and Vibe-Coding Content in the Region
The speaker expressed the opinion that AI content—especially vibe-coding content—needs a stronger presence in Egypt and the wider MENA region, and encouraged a creator to continue working in that area.
- 0:20:01
How Project Files and Task Size Affect the Context Window
A practical observation that files such as manifests and roadmaps consume part of the context window in each session, while the language-priority change was viewed as a relatively small task rather than a full refactor.
- 0:21:32
Deploying Website Updates Through GitHub and Cloudflare
Explains that local changes will not appear on the published site immediately; pushing or committing them to GitHub is expected to trigger automatic deployment through Cloudflare.
- 0:22:09
Reducing Documents Loaded Into Claude Session Context
Investigates why project documents consume session context and concludes that CLAUDE.md directly imports some of them. The discussion considers removing unnecessary material while retaining roadmap visibility when agents need to understand project status.
- 0:26:10
An Agentic Swim Lane With Flexible Dates and Dependencies
Plans to make the roadmap swim lane adjustable when work falls behind or moves faster, while giving each milestone more detail about the subtasks and dependencies required for delivery. The intended view should support both the project owner and AI agents.
- 0:28:06
Variable Context Consumption in AI Coding Tasks
Observes that a small language-related website edit consumed more context than expected compared with an earlier, larger build task, without reaching a definitive explanation for the difference.
- 0:30:14
Egyptian Dialect Support and Coding Tool Choice
A brief exchange asks whether Egyptian dialect is understood, then turns to which coding tools are used, such as Claude or GPT; these pages do not capture a complete, detailed answer.
- 0:32:26
Coding-Assistant Usage Limits and the API Alternative
Discusses generally how some subscription capacities can make usage limits harder to reach and how a status command can show consumption and reset timing. It also identifies a paid API as an alternative, without exposing actual account data.
- 0:34:29
An English-First Interface With Selective Arabic Cues
Defines an English-first website direction to reduce the noise caused by translating every element, while retaining selected Arabic words that signal the platform is also built by an Arabic-speaking creator.
- 0:35:36
Codex IDE and Warp for Parallel Coding Workflows
The speaker says he believed Codex IDE was temporarily offering doubled usage limits to Pro users through April 2. He then explains his current preference for Warp as an agentic terminal that supports multiple tools or models and parallel sessions for different parts of a project.
- 0:39:33
Considering OpenClaw Locally Instead of on a VPS
According to the speaker, OpenClaw can be connected through a ChatGPT subscription without directly using the OpenAI API. He says a previous VPS setup involved too many steps, leading him to prefer running it on a local computer. These are attributed claims and experience, not documented tests in the captions.
- 0:41:09
Comparing AI Models with the Same Prompt
The speaker suggests testing two AI models with the same prompt and comparing their results. He considers one model mentioned in the discussion suitable for most tasks and less expensive, but its identity is not clear enough in the captions to assign a project name.
- 0:41:38
VPS Trade-Offs for Hosting Agent Tools
The speaker explains that a VPS offers an always-on connection and remote access, but he found the setup cumbersome and describes some hosting security systems as unreliable. This is his personal assessment rather than a systematic test result.
- 0:46:48
Building with Vibe Coding Without Deep Web Expertise
The speaker says the channel aims to show that people can participate in live coding and build useful things without being professional web developers, presenting himself as a vibe coder rather than a traditional web-development expert.
- 0:48:29
Designing an Agent-Aware Timeline for Drift and Milestones
The work explores a timeline swimlane that visualizes the gap between planned and actual milestone positions with a drift bar, allowing phases to stretch or compress as reality changes. Each node acts as a summary that opens a subtask panel. The design also proposes connecting the timeline to project agents so they can understand schedule state, with sections for specialized checklists such as Shopify and submission requirements.
Got a question about something in this stream?
Ask in the group