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 developing a website and an AI-assisted coding workflow with tools including Codex, Claude, and Antigravity. Practical work covers interface hierarchy, CTA wording, project cards, live preview, and selecting page elements inside the development environment. Luqman also discusses a command center under construction for project maps, tasks, agent teams, and execution prompts generated from project context; inspection suggests the dynamic prompts remain too long and generic and need smarter, role-specific context profiles. Another major thread explores connecting the website to YouTube live status and separate live/offline thumbnails through options such as the YouTube Data API, Cloudflare Workers, and KV, but the integration was deferred and not shown as complete. The stream also covers automating content repurposing, organizing YouTube playlists, and separating recommended or affiliate-linked tools from the technical stack. In audience discussions, Luqman argues for learning by building while continuing college and stresses starting with a real problem, validating demand and unit economics, and planning distribution before implementation. A Telegram community and an initially manual paid-membership verification process are considered, while questions about automated GitHub pushes and permissions remain unanswered. Much of the work is exploratory or unfinished, and the session includes recovery from a power outage and a website crash whose cause was not established.
40 moments from the stream
- 0:01:47
Automating Stream Content Repurposing
The speaker explains that turning long streams into clips, Shorts, and Reels and distributing them across platforms is a separate workload, so he tries to automate as much of the workflow as possible.
- 0:04:31
Organizing YouTube Content into Three Playlists
The speaker outlines three YouTube tracks: a recurring main stream, a different playlist to be revealed later, and shorter videos collecting shortcuts and lessons learned. He frames the shorter format as his view of what YouTube's algorithm favors.
- 0:06:09
Polishing a Website with AI Coding Tools
The speaker switches from another project to polishing a website UI in Antigravity and demonstrates opening the local site inside the IDE. He says an earlier version was built quickly with Claude and then polished with a GPT model, while expressing a preference for Claude for core work because of what he considers its larger context window. His current plan is to use Codex for polish and Claude for overall direction. He starts by asking to remove a duplicate logo while keeping the top one, then notes that the landing page still contains placeholders; this page does not show a final result.
- 0:12:25
Clarifying CTA Destinations
A review of the page’s hero section distinguished a “Watch the stream” button leading to YouTube from a “Join the community” button leading to an email form. The suggested improvement is to name the actual destination so the CTA matches user expectations.
- 0:13:43
Selecting Page Elements Inside the IDE
An experiment used an in-IDE website preview to send a selected page element into chat, giving the assistant direct context without a screenshot or lengthy description. Changes appeared live, but copying or attaching the element did not work clearly during the attempt, and availability beyond the tool in use remained uncertain.
- 0:18:17
Options for Automating Stream Status and Thumbnails
The stream-section design discussion considered adding a thumbnail and connecting the broadcast window to live status. Options included an easily swapped image, deriving a thumbnail from a YouTube video ID, using the YouTube Data API, or updating live state through an OBS script with Cloudflare Worker and KV, with tradeoffs such as API keys, rate limits, or extra setup.
- 0:23:13
Deferring Setup-Heavy Integrations
Thumbnail work and live-status integration were deferred because implementation required external tools and YouTube API setup. The stated workflow choice was to preserve focus during the build session and complete the dynamic integration later.
- 0:27:20
Cross-Referencing Another Project’s Documentation
The session explored whether the assistant could reference another project to retrieve a planning document. The response indicated that this was possible when given a path, while directly bringing the document into the current context did not succeed in the shown attempt, so copying it was considered as an alternative.
- 0:31:20
Explaining AI Projects Within the Website
The site’s project cards are planned as clickable, concise introductions. Command Center is described as a project-management space with workflow lanes, tasks, and an agent hub offering substantial autonomy, while Talk Store is presented as a conversational store assistant for changes and product enrichment. The first UI result was treated as an early draft to refine later with Codex.
- 0:36:37
Planning Affiliate Monetization for a Tools Section
An exploratory discussion considered how to begin affiliate monetization for listed technology tools, understand enrollment and commission flows, compare programs, and eventually make the cards clickable. This was planning and research rather than a verified implementation.
- 0:38:55
Connecting a Website Section to YouTube Live Status
The stream section is planned to link directly to the current YouTube broadcast while live and fall back to a static thumbnail when offline, with further investigation into the mechanics and possible use of the YouTube API.
- 0:41:37
Separating Referral Links from the Rest of the Tool Stack
The discussion proposes a dedicated section for tools with referral programs while keeping the remaining tools in a separate section that presents the working stack. The aim was to keep recommendations clear instead of presenting every tool in the same way.
- 0:44:12
Detecting a YouTube Livestream for Dynamic Website Display
The discussion explores replacing a static website placeholder with a dynamic YouTube live-status link. The proposed design uses lightweight periodic checks against YouTube Data API v3 from an edge function and caches the result in Cloudflare KV so the site can read it without a fresh API request on every visit; the page does not establish that this was successfully implemented.
- 0:47:32
Previewing and Selecting Website Elements in Antigravity
The speaker explains that Antigravity can display a website inside the development environment, allowing a page element to be selected and sent to chat for modification instead of being described or captured in a screenshot. During the demonstration the site crashed, and it remained unclear which chat received the selected element. The speaker also offered the opinion that this kind of work does not necessarily require formal courses.
- 0:50:55
AI Agent Teams and Avoiding Work Overlap
After the website crashed, the speaker suggested—without confirming—that several agents running together might have caused it, then explained agent teams that communicate to coordinate work and reduce overlap.
- 0:52:41
Start With a Goal and Distribution Plan Before Building
AI can make websites and apps easier to build, but it does not answer what happens afterward: who will use the product, what will be sold, and how it will reach the market. The advice is to define the purpose, profit model, marketing, and distribution before starting.
- 0:54:39
Screenshots and Live Preview in Agent Development Environments
A practical discussion of being able to upload screenshots to the Antigravity agent but not to terminals, using multiple terminals and tools such as Codex, and the benefit of previewing a website live as it changes.
- 0:58:35
Validate Profitability, Market Demand, and Unit Economics First
The discussion rejects quick-riches claims and recommends using an AI assistant to examine the profit structure, sales path, and marketing before requesting ideas or building. Market research and unit economics may show that an appealing idea is unsustainable, especially when service costs do not match what the target market can pay.
- 1:05:46
Recovering a Coding Workflow After a Power Outage
A power outage stopped the local development session. After restarting, the developer checked that the environment and updates were working, then tried to identify the previous task and resume the coding assistant’s chat context before continuing an in-progress interface change. He noted that a UPS can protect the computer temporarily but does not prevent workflow disruption.
- 1:11:49
Project Map, Task Board, and Agent Hub
A walkthrough of a command center under development that turns complex work into a visual map with timelines and weekly tasks, alongside a task board and a hub where AI agents can coordinate and understand their responsibilities. The displayed week estimate was presented as the tool’s output, not a guaranteed schedule.
Got a question about something in this stream?
Ask in the group