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 an evolving project command-center concept that connects a swimlane-and-milestone map with a task board executable by AI agents. Luqman discusses giving each task sufficient context, files, dependencies, and verification criteria, potentially assembling a dynamic prompt after a fresh codebase scan, while experimentally visualizing agent roles, collaboration, and activity. He stresses that the interface, the relationship among the Agent Hub, task board, and swimlanes, task and timeline accuracy, and the mechanics of shared memory and communication still require verification and iteration; this is not a completed result. Other useful discussions cover a research-based checklist for Shopify app submission, balancing Arabic and English, the unit economics of AI services in Egypt as the speaker’s opinion, and audience questions about Claude Code agents and cybersecurity. CTFs, deep training, and judicious AI use are suggested as practical starting points for security learning. The speaker also advocates learning AI app development through direct building and mentions deferred weekly and daily planning that was not completed in the recording.
25 moments from the stream
- 0:08:09
Planned Weekly and Daily Planning
Luqman says he plans to review what he wants to accomplish during the week and then write the weekly and daily tasks; this is planned later work, not a review completed in the recording.
- 0:08:23
From Swimlanes to a Project Command Center
Luqman traces how the idea evolved from swimlanes supporting an initial concept, to an agentic task board, and then to a full command or mission-control center. He explains that the version shown at the time of recording was not yet simple or easy to understand, and that the goal was to simplify the process and provide a very clear project map connecting domains, nodes, and tasks. He emphasizes that the whole concept still needed review.
- 0:10:01
A Unified Work View in the Command Center
The speaker presented an interface that brings together the foundation, weekly work, tasks, and an on-track indicator, with the task board connected to workflow stages and lanes. He noted that the interface still needed refinement.
- 0:10:40
A Pre-Submission Checklist for Shopify Apps
The speaker explained the value of tracking Shopify requirements before app submission. He said reviews can take time and, in his account, a single compliance issue may cause rejection and another wait. He therefore compiled a research-based checklist to review during development and reduce delay risk; the timelines and process are his claims, not independently verified here.
- 0:12:35
A Task Board Executable by AI Agents
The speaker contrasted a conventional task board with an agent-oriented one, where each task can carry system instructions, verification criteria, memory, dependencies, and requirement documents. The idea is to plan work periodically and then ask an agent to execute a defined task instead of rewriting its instructions each time.
- 0:14:30
Open Questions Before Adopting the Agent Command Center
The speaker regarded the command center as technically almost complete, but said he was not yet comfortable or familiar with it. He wanted to understand how the agent hub, task board, and workflow lane relate, and to verify task coverage, dependencies, and possible omissions. He planned to request a step-by-step AI walkthrough and refine the interface after understanding the system, leaving comprehension, verification, and improvement open despite its near-complete technical state.
- 0:20:04
Designing a Command Center with a Project Map and Swimlanes
A walkthrough and critique of a command-center design in which the swimlane view acts as the project’s high-level map, showing work areas, milestones, dates, dependencies, and the effect of schedule changes. Selecting a milestone reveals its subtasks and details, which can be presented as a Kanban-style task board. The discussion stresses accurate tasks and dates and identifies a need to switch between milestones from the task board, while noting that parts of the design remain unfinished.
- 0:29:34
Balancing Arabic and English in Technical Content
A brief discussion of alternating between Arabic and English to serve an Arabic community while retaining English for technical communication, with the qualification that following two languages can be difficult for some viewers.
- 0:32:25
Envisioning an Agent Hub and Requesting a Visual Example
The speaker outlines a tentative view of an Agent Hub as a place where models, agents, and possibly APIs come together, share memory, and show who is doing what. He says the communication mechanics, shared memory, and future interface remain unclear, so he asks for a practical visual example on the platform to validate the concept.
- 0:36:26
Unit Economics of AI Services in the Egyptian Market
An audience suggestion proposes subscription-based AI analysis or consulting. The speaker argues that a local Egyptian model may be difficult when variable AI API costs are denominated in US dollars while revenue is collected in Egyptian pounds. In his view, a global market may be more suitable, or the local model may be more viable if costs can be fixed rather than varying with usage.
- 0:40:40
Visualizing Agent Roles and Activity in Claude Code
The discussion proposes separating Claude Code activity into visible roles such as Explorer, Builder, and Planner, showing which agents participate in each task and how they interact through the interface and activity feed. An initial option was implemented and visually reviewed, with further refinement still considered.
- 0:47:31
The Difficulty of Cybersecurity and Bug Bounties
Cybersecurity and bug-bounty work are described as difficult fields that are not easy to master and require a suitable problem-solving mindset; a later comment adds that patience and deep training are important.
- 0:48:08
Question About Claude Code’s Internal and Custom Agents
A follow-up question asks what kinds of agents Claude Code invokes internally and whether custom agents can be built. This page does not contain a clear answer.
- 0:52:45
A Practical Start in Cybersecurity Through CTFs and Bug Bounties
The discussion notes that cybersecurity-company roles may require experience and certifications, and suggests CTF challenges as simulated training environments that approximate real situations. It also says AI may help with bug-bounty work, while stressing that it should be used judiciously.
- 0:57:28
How Agent Teams Coordinate Specialist Work
The captions describe an experimental setup where the user talks to a team-lead agent, which then spawns specialists such as frontend and backend agents. Spawned agents coordinate directly, while the terminal exposes subtasks and agent-spawning permissions.
- 0:59:11
Connecting an Electron Command Center to Swimlanes and a Task Board
The speaker gives a positive early impression of an Electron-based command-center app but emphasizes that he is still learning it. His focus is understanding how swimlanes relate to the task board and ensuring tasks are valid because the design remains experimental.
- 1:02:01
How Coding-Agent Teams Collaborate on One Project
The speaker explains an experimental feature for running several agents in parallel on different parts of a project, such as refactoring while another agent builds a component. In his account, the agents can communicate to coordinate overlapping work and reduce conflicts; this is a description of the feature, not a completed test result on this page.
- 1:05:07
Learning to Build AI Apps Through Practice
In the speaker’s personal view, courses can provide an introduction to working with AI, but direct experimentation and building are the best ways to learn. He describes the stream as showing the building process as it happens—including successes, failures, and bugs—rather than offering short, definitive tool comparisons.
- 1:06:27
Preparing an Agent-Team Prompt Trial
The speaker starts a new session and looks for a simpler task in his app for trying agent teams. He then reviews a prompt assigning roles to multiple teammates and watches the feature to understand its behavior; the page does not show a completed outcome.
- 1:12:32
The Agent Hub’s Role in Coordinating Work
The speaker explores the purpose of an Agent Hub after seeing its activity feed and connected agents, then explains that it exposes the agents working on a project and gives them shared visibility into goals and state to reduce overlapping work. He also discusses remote activity monitoring and using hooks and loops for trigger-based tasks.
Got a question about something in this stream?
Ask in the group