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 extended practical experiment in refining a website with coding agents, especially hero-section motion, a dotted grid, connector lines, and a milestone progress bar. Luqman uses live previews, screenshots, and browser inspection to verify rendered results rather than trusting code alone, while investigating missing CSS keyframes, visual regressions, server or port confusion, and effects that disappear after refresh. Some motion became visible after switching to the project’s existing motion library, but other details remained under refinement; he emphasized limiting change scope and preserving good states in Git. He also explored a command center for projects with multiple agent roles and shared JSON status, while recognizing the risks of over-automation and of the management tool becoming a distraction. The choice between automated pipelines and steerable interactive sessions remained unresolved. Based on personal experience rather than a comprehensive benchmark, he compared Cursor, Warp, Antigravity, and Codex across embedded previews, terminal sandboxing, runtime control, context handoff, and usage limits. Audience questions covered WordPress versus newer platforms, Supabase and Firebase, turning a website idea into an Android app, and whether WordPress suits IPTV or OTT platforms, with an emphasis on content rights and legal risks. Unresolved issues included several interface animations, recovery of a deleted effect, YouTube API consumption, and YouTube’s inconsistent display of the live-stream duration.
108 moments from the stream
- 1:03:18
Tuning a Cursor Effect and Background Details
The interface review calls for removing the extra glow behind the cursor, reducing the circle, and checking its shape and gradient, while retaining the dotted grid and making its spacing more compact. A temporary cursor-tool problem and restart attempts interrupt the work.
- 1:07:50
Why Interactive Control Matters During Agent Tasks
In a partly unclear passage, the discussion contrasts running multi-stage agent tasks without intervention against a workflow that allows the task to be watched and steered while it runs. The speaker says the lack of control in an earlier pipeline was frustrating and concerning, and considers the more interactive approach much better.
- 1:13:23
Diagnosing Paste Failures in Cursor Terminal Sessions
The host described a recurring issue while using Whisper Flow to communicate with agents in Cursor terminals: after a few interactions, terminal sessions stopped accepting pasted text, even manually, while the agent manager continued to work. Restarting Cursor restored the behavior only temporarily, prompting discussion of moving to another terminal workflow and noting that terminal sandboxing had motivated the Cursor trial.
- 1:20:24
Refining UI Elements Through Visual Comparison
The host continued refining a website interface by reducing the dots and their spacing, then placing the site beside the development environment on another display for comparison. He also tried enlarging a circular element and adjusting a visual value, noting that the agent understood the requested change better because it already had the working context.
- 1:23:03
IDE Agents Versus Terminal-Based Workflows
The host explored IDE-integrated agents and compared them with direct terminal workflows across Cursor, Codex, and Antigravity. He discussed agent selection, an in-IDE browser, automation, and his view that terminal sandboxing may steer users toward the built-in agent manager. He also relayed an unverified article’s allegation that Composer 2 began as a wrapper around another model before further development; the model name is unclear in the captions.
- 1:31:30
Handing Off Project Context Through a Markdown File
Explains creating a handoff prompt and a Markdown file inside the project to carry context from a long Cursor session into Warp, allowing the new session to continue with website optimization without re-explaining the work.
- 1:32:17
Automatically Displaying Live Streams on a Website
Describes an automation using YouTube API calls to show a live stream and a stream section on the website while streaming.
- 1:34:55
Comparing Cursor and Warp for UI Preview Workflows
A practical comparison of Cursor’s embedded website preview with using Warp alongside a separate browser. It covers tab switching, mobile-view limitations, terminal sandboxing in Cursor and Antigravity, a preference for direct terminal work, and assigning Codex to continue polishing UI effects and subtle gradient differences.
- 1:42:44
Refining Animated Effects in a Website Hero
The work focused on a cursor-responsive circle and network-like signal traces in the website hero. Review showed that the lines were visible but static, so the remaining goal was to add clear motion and a flowing border effect while keeping the design simple.
- 1:55:48
Comparing AI Coding Tools
A quick comparison was considered between Cursor and another AI coding tool, focusing on plans, usage limits, and the ability to open multiple agents or chats. The impressions were preliminary, with no final comparative conclusion established.
- 2:03:32
Comparing AI Coding Tool Plans
The speaker considers trying a paid plan and compares options by public monthly price, model availability, and advertised context capacity. The captions do not identify the product or connect the plans to these details clearly enough.
- 2:05:51
Checking the Server When UI Motion Disappears
While troubleshooting, the speaker observed that UI motion had disappeared entirely and considered a server problem, suggesting a health check and restarting the terminal session. The captions do not establish a confirmed cause or final fix.
- 2:08:31
Giving a Coding Agent Precise Context with Browser DevTools
The speaker explains opening browser developer tools from a preview, using the element picker to locate an interface element and its source, then sharing code or screenshots so the agent can understand the exact UI area being discussed. A participant says he tried moving a CTA element identified with the inspector to his own site and that the effect worked.
- 2:09:51
Terminal Sandboxing Limits in AI Coding Tools
The speaker discusses how terminal sandboxing can constrain an agent’s access to the terminal environment and preferred models, especially when supplying screenshots, UI elements, or voice context. He says some capabilities may require a paid upgrade and expresses a personal preference for Warp for terminal work.
- 2:12:55
Diagnosing Missing Interface Animations
They checked whether the server was at fault, confirmed it was healthy on the active port, and tried a hard refresh and reopening the page. A hover highlight remained visible while other motion was absent, and the speaker considered macOS Reduce Motion an unlikely cause.
- 2:17:22
Critiquing a Visual Effect and Returning to Brainstorming
The replacement effect became visible in the hero section, but the speaker found it more distracting and less purposeful than the earlier code-rain effect. Rising circles and static lines were criticized as visually unclear and inconsistent with the desired futuristic, Apple-like direction, prompting a pause in implementation and further research and brainstorming.
- 2:20:59
Ruling Out Browser and Operating-System Causes
A general browser or macOS problem was ruled out because other animations in the hero section were working normally. Multiple browsers had also been tried, directing the investigation toward the missing animation in the specific section rather than system-wide settings.
- 2:23:07
Choosing Between Two Terminal Background Motions
The research suggested retaining a visual connection to the code-rain effect through subtle terminal drift, or using a breathing-screen treatment as a more premium alternative. Luqman wanted to try both and started with terminal drift.
- 2:24:27
Handling an Accidentally Deleted Code Effect
Luqman noticed that the agent had deleted a code-rain effect he had previously built in Cursor. He considered restoring it while remaining unsure whether the project history or a backup was available.
- 2:26:34
Designing a Command Center with Multiple Agent Roles
Luqman considered whether to automate the task board through a pipeline inside his command-center app or continue working interactively in terminals and Cursor. He demonstrated a design that selects an agent type by task, gives explorer, planner, and builder roles distinct system prompts, then generates a task-specific prompt that can be previewed and copied into another session. The more automated pipeline mode remained the more complex option, and no final architectural choice was established.
Got a question about something in this stream?
Ask in the group