5 TASKS IN 7 HOURS
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 follows work on a Shopify app and a multi-agent “Command Center” workflow that divides exploration, implementation, and auditing while keeping the human responsible for setup and sensitive decisions. Practical work centered on connecting AI APIs, aligning environment variables across Railway and Supabase, configuring Sentry for production monitoring, and troubleshooting deployment, authentication, trial, approval, and cancellation flows. Some deployments and checks were reported complete, but issues including an HTTP 500 error and parts of the user flow remained unresolved, illustrating the limits of automation and the value of clear logs, staged audits, and CI/CD. Luqman also discussed choosing coding tools by task, retained context, and usage limits; his comparisons of Codex, Claude, and Kimi K2 were framed as limited workflow impressions rather than controlled benchmarks. Broader topics included designing defensible scores for stores’ visibility in AI answers, the unsettled GEO-versus-AEO terminology relative to SEO, and validating markets, pricing, distribution, and product economics before accelerating development. Research-oriented discussion extended to LLM interpretability and the lottery ticket hypothesis, as well as Luqman’s views on Chinese and US model ecosystems, reliance on NVIDIA chips, and the need for accessible Arabic models and local pricing. Audience discussions covered AI’s effect on developer roles, code comprehension and debugging, parallel milestones and dependencies, agent-rule context costs, productivity habits, and peer learning. This portion also included additional practical discussions about workstation and operating-system choices, stream latency, development practices before AI, and programmatic video creation with Remotion. Preparing Shopify submission media and polishing the Command Center for a possible GitHub release remained future plans, not completed outcomes.
127 moments from the stream
- 3:15:28
Learning AI Without a Programming Background
The speaker explains that his core background is in commerce and digital marketing rather than programming, and that his practical capability grew from learning how to work with AI tools. He describes the stream as a public record of his ongoing work sessions and journey, not a sudden starting point.
- 3:16:29
Beyond Prompt Engineering: Documentation and Workflow Design
The speaker argues that prompt engineering was a prominent term in early 2024, but model advances have shifted more attention toward documentation, back-end logic, and workflows that tell an AI how to interpret and deliver a task. He notes that his own interaction may begin with a short voice command because the implementation details are prepared elsewhere in the system.
- 3:18:15
Choosing an AI Tool by Task and Usage Limits
The speaker recommends choosing a model according to the task, especially the nature of a coding project, while comparing usage limits and the value available under a subscription. He expresses a personal preference for tools that provide more working capacity at constrained budgets, while noting that suitability depends on the user’s needs and access.
- 3:21:19
How Branding Supports Marketing
The speaker encourages keeping project costs low and argues that a solid understanding of branding covers an important part of marketing work, potentially reducing the early need for separate brand management.
- 3:22:20
A Peer Community for Learning and Market Readiness
The speaker encourages researching first and then asking focused questions, describing the community as a space where members learn from one another and exchange experience. He says its purpose extends beyond AI-assisted coding and model discussion to helping participants prepare for the market, while preferring to keep divisive political discussion out of the group.
- 3:24:35
Why Linear Project Estimates Fail
The speaker explains that estimating progress as a fixed percentage per day failed to account for blockers and troublesome implementation details, making the expected schedule overly optimistic.
- 3:28:18
Verification and Rework in Deployment Tasks
The speaker describes a workflow in which the tool performs its own verification, he reviews the result, and the cycle repeats if needed. He notes prior difficulty with this kind of deployment task and later reports that the deployment succeeded after taking additional time.
- 3:35:59
When Automation Stops Running on Its Own
The speaker contrasts a day when the command center worked largely on its own with a day that required substantial manual intervention because of unexpected blockers. Environment variables and APIs are cited as examples of details that can prevent a task from proceeding as automatically as planned.
- 3:45:20
Evaluating Codex in a Coding Workflow
The speaker plans to evaluate Codex on the real workflow to see how much work it can handle, with the pace of usage-limit consumption treated as one decision factor; no conclusion is reached yet.
- 3:48:28
Diagnosing a Failed Test and Verifying a Backend Deployment
A test failed while prior backend work was being verified, and the assistant appeared to be attempting a repair instead of only reporting the error. The speaker then monitored the worker and task-store components after pushing them together, waited for deployment results, and described the worker’s production deployment as successful.
- 3:52:52
Requesting a Handoff Prompt to Continue Two Tasks in Codex
After the deployment stage, the speaker requested a handoff prompt for two specific tasks and decided to move the remaining work to Codex to evaluate its performance after the previous workflow encountered friction and usage limits.
- 3:56:55
A Model Recommendation for Debugging
The speaker relays others’ recommendation of a model whose name is unclear in the captions, said to be good for debugging, and says he plans to try it. The page contains no actual test or confirmed result.
- 3:59:49
Why Tasks Requiring Human Effort Slow the Workflow
The speaker contrasts rapid progress on automatable tasks with slower progress when a task requires direct human effort, and says he will monitor how much work Codex can handle. This is a practical impression, not a controlled benchmark.
- 4:01:36
Programmatic Video Creation with Remotion and an AI Assistant
The speaker outlines a workflow that connects Remotion to a coding assistant and supplies a Markdown file describing a project’s mission and vision to help create a video, such as a marketing video. The description does not establish that a result was produced or validated on this page.
- 4:02:46
GEO or AEO? Comparing the Terms with SEO
The speaker reviews a Google Trends comparison of search engine optimization (SEO), generative engine optimization (GEO), and answer engine optimization (AEO). He considers GEO and AEO close in meaning, says GEO interest rose notably during 2025, and discusses which term better describes the field. These are his reading of the trends and personal expectations, not a definitive market conclusion.
- 4:05:04
Using Trend Signals to Identify GEO Opportunities Early
A general explanation of watching market signals and expected demand within a domain, using Google Trends to see what is gaining attention, while presenting forecasts as expectations rather than proven outcomes.
- 4:08:49
A Multi-Agent Workflow for Planning, Coding, and Review
An answer describing how explorer and researcher agents inspect the codebase and plan together, hand an instruction file to a builder agent, and then have an auditor review the implementation and accept or reject it when an issue is found.
- 4:10:12
Sentry’s Role in Detecting Production User Issues
A discussion explaining that Sentry helps track issues users encounter in production. It is contrasted with Railway deployment checks that may block a deploy when code has a problem, while noting that production monitoring extends beyond simply passing deployment validation.
- 4:14:58
Understanding and Configuring Sentry for Production Error Monitoring
While configuring Sentry, Luqman discusses his preliminary understanding of how an error-monitoring service differs from deployment platforms such as Railway and Vercel. Sentry can provide error context, a user journey, and more detailed logs, reducing guesswork when diagnosing production problems. The discussion also covers adding the DSN and configuring production, with clarification that local setup is unnecessary unless local monitoring is wanted.
- 4:23:50
Why Unclear Deployment Logs Slow Down Diagnosis
Luqman describes work being blocked by API setup steps and a Railway deployment issue whose logs did not clearly identify the source of the failure, leading to substantial guesswork. The experience illustrates the practical value of monitoring tools that provide better error context before redeployment.
Got a question about something in this stream?
Ask in the group