Time to Face the Market
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
A broad work-and-discussion stream centered on preparing Burnish, a Shopify-focused AEO app, to move from technical availability toward marketing readiness. Practical review covered scan and report flows, loading states, clearer presentation of product signals and merchant value, interface and performance refinement, and confident copy that avoids unsupported guarantees. Luqman stressed separating genuine launch blockers from improvements that can wait, resisting feature creep, and seeking candid feedback from users outside the builder’s bubble. The stream also explored a funnel built around a landing page, free report, and email sequences; acquisition-channel measurement; and an experimental process for creating AI-assisted video ads from strong references. The strategy and creatives were still being tested, with no final performance result presented. Technical questions remained unresolved around attribution of commerce from AI assistants, access to pre-installation data, differing scores, and possible shared saved state between users. Audience discussion broadened into AI-built web-app security, agent requirements, B2B SaaS marketing, freelancing, learning programming in the AI era, and choosing among tools such as Codex, Claude, and n8n. The speaker framed tool comparisons and future-of-work predictions as time-specific opinions, while consistently emphasizing understanding AI-generated work, verifying outputs, planning adaptively, and shipping before everything is perfect.
127 moments from the stream
- 2:52:22
Why AI Estimates Make Roadmaps Difficult
The speakers discuss how AI tools, especially Claude, can produce unstable or inflated implementation estimates. They describe a swimlane-style planning interface whose weekly timeline was eventually removed because the estimates were not reliable enough.
- 2:54:10
Showing Signal Activity Without Ambiguity
A UI concept uses blocks with pulsing dots to convey that the system is continuously processing signals, alongside an emphasis on making every state and meaning clear rather than ambiguous.
- 3:04:15
Monitoring a Long-Running Codex Scan
An operational note that a Codex scan had continued for more than an hour despite a 1.5× speed setting, without establishing the cause or reporting a final result.
- 3:05:16
An Intermittent Journey into Deeper Vibe Coding
The speaker describes learning programming through vibe coding without a prior programming background. He began intermittently in September 2025, paused for a period, then returned to research and pursue it more intensively in March. The reasons for the interruption and personal circumstances are omitted.
- 3:06:50
Designing a Visual Widget for the Visibility Section
During an interface review, the speaker asks for a visual widget showing all four tabs in the visibility section and suggests making them clickable. The project identity is unclear from the captions.
- 3:09:01
Is Learning Programming Still Worth It in the AI Era?
Answering whether manual programming is still worth learning, the speaker advises a computer-science student to continue studying and use that technical foundation. He gives his view that many companies may soon need fewer developers to write code directly, while developers increasingly supervise and check AI-generated code, and encourages treating that predicted shift as an opportunity to adapt.
- 3:12:24
AI’s Potential Impact on Jobs and Industries
A brief discussion of a chart attributed, with some uncertainty, to Anthropic that mapped industries already affected by AI and those not yet affected, alongside a prediction that some jobs may become scarcer.
- 3:13:05
Using CLI Tools Inside IDEs
An answer about running a coding tool through the terminal inside IDEs such as Cursor and VS Code. The speaker says it is possible, but argues that terminal sandboxing and IDEs’ emphasis on their own agents can make the experience awkward, while noting that he had not personally tried the referenced extension.
- 3:20:06
Responsive Design Review and CTA Refinement
A practical review of the mobile interface, with a positive assessment of a frontend design that Codex produced largely autonomously. The requested refinements included centering the installation CTA text and having it fill the button appropriately on both mobile and desktop, while acknowledging that further polish remained.
- 3:22:18
Refining Burnish’s Message and Reviewing Its PDF Output
A practical review of hero copy that should explain Burnish’s benefit clearly and convey confidence without making an explicit guarantee, alongside feedback on product-image display, alignment, margins, and score presentation in the PDF.
- 3:26:38
Codex vs. Claude: Efficiency Versus Quality
Answering an audience question, the speaker says his current preference leans toward Codex for efficiency and getting more work done, while quality is more debatable and Claude may still produce strong results. This is a personal, time-specific assessment rather than an absolute verdict.
- 3:31:16
Managing Slow, Long-Running Coding-Agent Tasks
A practical discussion of mitigating a slow coding agent by running several tasks in parallel, having another model review selected output, and scheduling heavy work for times when the computer is not otherwise needed. The speaker notes that the best approach depends on the project.
- 3:32:20
Breaking Work into Manageable Tasks
A practical suggestion to divide work into parts, delegate a task lasting roughly an hour or more while away from the computer, and prepare other pieces when working at the machine.
- 3:34:19
Why Product Value Should Be Visible During a Trial
From a marketing perspective, Luqman argues that hiding important capabilities in the backend can prevent a new user from recognizing Burnish’s value and may lead them to uninstall after a short trial. In his view, making valuable work visible should inform product-design decisions.
- 3:36:25
Learn Marketing First, Then Amplify It with AI
Responding to an idea about building marketing agents, Luqman supports learning marketing principles and methods while using AI to study and apply them, with occasional reference to outside sources when needed.
- 3:37:19
Balancing Product Visibility with Competitor Access
Luqman discusses the risk of exposing product details to competitors and mentions restricting some developers’ access to the application. He relates this to examining competing products that allowed entry, then acknowledges that his earlier judgment was partly mistaken.
- 3:38:20
Why a Developer’s Tool Stack Shrinks Over Time
Luqman explains that his tool count has fallen over time even though core services remain for databases, access to multiple models, infrastructure, streaming, video, deployment, and error tracking. He expects expanding AI capabilities to reduce the number of separate tools many developers need.
- 3:40:31
Is n8n Still Necessary for Automation?
Luqman says n8n’s suitability depends on the project. He notes its prominence in 2025, particularly around agent automation. However, at the time of the recording, he argued that AI made it possible to build many automations directly without necessarily relying on it.
- 3:42:37
Using Codex and Claude Together in an IDE
An explanation of configuring an IDE to use Codex and Claude together, while noting that the choice between this workflow and graphical interfaces depends on the developer’s preferences and working style.
- 3:43:59
Agent Automation and the Rise of Graph Engineering
An opinion favoring hands-on learning in agent automation, alongside an observation that technical discussion has shifted from prompt engineering to agents and then toward what the speaker calls graph engineering. The latter is presented as an emerging topic he hears about and wants to explore, not as a settled market conclusion.
Got a question about something in this stream?
Ask in the group