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 building and learning in public while planning a Shopify B2B SaaS product and improving a shared workspace for agents such as Codex and Claude. Luqman discusses preparing a project manifesto, roadmap, and Shopify review documentation before implementation; organizing dependent phases; sharing instructions and skills across agents; and reducing approval prompts while retaining safeguards for sensitive commands. The discussion develops into a visual project-map concept—favoring a swimlane view—that would turn a Markdown roadmap into an Electron app tracking milestones and dependencies while connecting to agents. This is compared with a custom task board and an always-on agent environment, but all remain ideas requiring testing. Practical topics also include separating a documentation app from the eventual Shopify app, possibly using simple HTML for a marketing page, and preliminarily favoring duplicate themes over direct edits to a live Shopify theme. The stream also considers developing a basic app while refining its concept, then presenting it through a polished landing page and clear explainer video, with only a general and uncertain reference to attracting early users. Luqman also revisits a domain transfer and a website built with Claude and polished with Codex, although DNS verification and the final review were not shown as complete. A broader content hub for streams, projects, explanations, and an Egyptian and MENA audience is proposed, with the sessions framed as a work-and-learning record rather than final tool verdicts. Open questions remain about coordinating instruction files, what happens to theme changes after app removal, the project map’s technical design, and whether an always-on agent environment could replace a separate task board.
33 moments from the stream
- 1:02:35
Is Simple HTML Enough for a Marketing Site?
A discussion of building a clean marketing page with email capture without an evident need for rendering or a complex stack, with the option to move to a framework later if the site expands.
- 1:04:56
Coordinating Agent Reviews and Building a Project Map
A proposed workflow for passing review findings between agents, escalating non-critical issues for human review, and performing a final audit. The discussion also stresses creating a project map to retain an overview before details become overwhelming.
- 1:12:55
A Custom Task Board for AI Agents
A concept for a task board connected to agents in a work environment. Each task can contain system instructions, context, a verification mechanism, and memory notes, then move through review, completion, and archival while preserving a record of completed work. The idea is still exploratory and might be offered to others if there is interest.
- 1:15:41
Could an Always-On Agent Environment Replace the Task Board?
An early comparison between building a standalone task board and using an always-on agent environment. The speaker speculates—not as a confirmed result—that agents could create the board themselves, operate within an organizational hierarchy, and be reached through messaging apps, so he wants to test the approach before deciding.
- 1:18:13
Presenting an App Concept Before the Full Build
A discussion of starting with a basic app version while refining its concept, then presenting it through a polished landing page and a clear explainer video showing what the application does. The passage also generally mentions attracting early users, but the specific targets are unclear and therefore omitted.
- 1:21:10
Reviewing and Optimizing a Project Environment for Codex
An in-depth review was requested, beginning with the project documentation and then covering folders, Markdown files, JSON files, and other configuration. The goal was to determine whether a workspace initially optimized for another assistant needed further changes to work well with Codex.
- 1:23:19
Designing a Visual Project Roadmap Tracker
The idea was to turn a Markdown project roadmap into a visual Electron application that tracks goals, important milestones, and progress from the starting point through the app launch, with the roadmap divided into understandable stages.
- 1:27:02
Reviewing Workflow Automation and Executable Tooling
The displayed review indicated that the workspace was optimized for another assistant while its Codex optimization was only partial, and it contrasted the existing workflow automation with executable tooling; one term in that assessment is unclear in the captions. Codex’s notes were then placed in a Markdown file for another assistant to analyze before implementation.
- 1:37:53
Envisioning a Live Project Map as a Desktop App
A concept for a project map showing lanes, nodes, and dependencies, with a possible Gantt-style timeline view. The desired result is a standalone Electron app rather than an HTML file opened in a browser, connected to the project so it can track progress and milestones and keep agents aware of the timeline. The discussion focused on visual concepts and requested an SVG prototype; implementation details and timing remained unsettled.
- 1:41:49
A Central Hub for Streams and Projects
The speaker outlines a website that would serve as a hub for streams and related content, showing his working approach, tools, and projects while hosting separate explanations and answers to selected questions. Social platforms would distribute shorter clips and attract viewers, with the hub evolving over time for an Egyptian and wider MENA audience.
- 1:43:38
Choosing a Swimlane View for a Roadmap
The speaker compares swimlanes, node graphs, and Gantt charts for presenting a multi-week roadmap. He considers the swimlane view clearer for this purpose and the node graph less detailed, while noting that several models estimated a longer project timeline than the launch-first, optimize-later approach he prefers.
- 1:48:31
Public Streaming as a Work and Learning Log
The speaker frames his streams as learning and building in public rather than delivering final verdicts on tools. He says recorded work sessions let him revisit his progress, and that thinking aloud while working with agents is personally useful regardless of audience turnout.
- 1:50:32
Connecting a Roadmap to AI Agents
The speaker wants more than a visual board for manual progress checks: he wants a more detailed swimlane connected to his agents so they can understand the current stage and completed work. He proposes turning the Markdown roadmap into this view and expects it to require a limited amount of additional work.
Got a question about something in this stream?
Ask in the group