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
- 5:32:46
Tracing a Blank Page After Approval
While testing an approval flow, no explicit error appears, but the page goes blank; after refreshing and returning, the change is visible. The speaker decides to reproduce it from the beginning and collect the log for debugging.
- 5:34:51
Building Solo or with Partners
The speaker weighs combining programming, marketing, and branding expertise in a team against starting solo with available tools and hiring help later. He also distinguishes later investment after proving the product from sharing ownership at the outset.
- 5:36:56
An Observation About Codex Reset Windows
The speaker observes that Codex’s displayed reset time appeared to move and cautiously hypothesizes that it may depend on session usage, contrasting it with a window that appeared fixed in Claude. This is a personal observation, not a confirmed account of system behavior.
- 5:39:19
Restarting the Environment While Debugging a Flow
While investigating the bug, the speaker decides to restart Shopify, close the server, and clean up browser windows; cancellation, approval, and a hard refresh are then mentioned as the debugging continues.
- 5:41:38
Clearer Terminology for Describing UI Errors
The speaker discusses using “user” rather than “customer” or “client” when describing what happens after a button click, considering it clearer and more professional in a development context, with plans to make the wording more natural later.
- 5:46:15
A qualified observation on Codex and Claude usage resets
The speaker observed that Codex session resets may shift and cautiously guessed that token usage could influence them, while describing Claude’s resets as more predictable. He explicitly treated this as an unconfirmed interpretation requiring further observation.
- 5:50:13
What is an enterprise resource planning system?
The discussion expands ERP as enterprise resource planning and broadly describes it as business software related to resource and project management, while noting that professional events can support networking with startups.
- 5:57:17
Testing a coding agent and escalating a recurring bug
While inspecting logs, the speaker gave a coding agent several attempts to fix a bug, then considered another agent when the error cycle persisted. Rather than continuing an unproductive loop, he requested a handoff prompt that documented the sequence of failures and directed the next model to the right task. The captions do not establish the exact bug clearly.
- 6:08:00
Looking for Coding and SaaS Conferences in Egypt
Luqman reviewed conference results in Egypt and asked whether any focused on coding or SaaS, noting that one visible result appeared to concern ERP rather than the topic he wanted.
- 6:11:49
Why Codex Took So Long on a Seemingly Simple Task
Luqman tried Codex on a billing-section change and questioned why its reasoning phase was so long for a task that appeared straightforward. He allowed that the backend might be more complex, then observed that the tool was testing that area; this page does not establish a final outcome.
- 6:15:52
The Trade-off Between Stream Latency and Quality
Luqman explained that the viewer-visible delay is tied to the stream-latency setting, and that choosing a lower-latency mode may reduce the delay at the cost of stream quality.
- 6:18:43
Different Coding Approaches in Kimi K2, Codex, and Claude
As an in-the-moment impression, Luqman says Kimi K2 took a long time and exposed an extended reconsideration of possible causes while Luqman was waiting for a result. He contrasts this with Codex, which may respond faster but sometimes before the work is complete, and expresses a personal preference for Claude. His assessment of Kimi K2 remains provisional and depends on whether it ultimately fixes the issue.
- 6:20:22
Diagnosing Authentication Errors and the Trial-Start Flow
The coding agent explored several possible causes for a problem that appeared related to a login page, including a server error, an unauthorized response, and middleware or server configuration. Luqman then explains the expected flow: starting a trial from billing should lead to a Shopify-hosted approval step, but that redirect was not occurring. He says an earlier implementation of the flow had been rolled back by one step during debugging, then contrasts the expected and actual behavior.
- 6:28:06
Waiting on Kimi K2 During a Debugging Attempt
The speaker discusses an unusually long Kimi K2 debugging attempt, saying the delay would be acceptable if the model ultimately produced a reliable fix; the outcome remains unknown at this point.
- 6:28:49
A Brief Claude-Assisted Website-Building Experience
An audience member briefly says they built a website with Claude in about a week instead of hiring a developer, without providing a detailed technical assessment of the result.
- 6:29:21
Why Agent Command Centers May Become Personalized Open-Source Tools
The speaker describes a command center organized around an agent workflow, an Agent Hub, and an MCP server, including an idea for switching workflow lanes across multiple projects. In his view, such tools will become common and highly personalized, so he leans toward publishing his work on GitHub instead of prioritizing it as a standalone product; he also expects application standards to rise and stronger offerings to be needed. This is his market opinion, not a documented test result.
- 6:40:21
Evaluating Coding Agents on a Real Workflow Problem
Luqman compares coding agents while troubleshooting a Shopify billing and trial flow, explaining that turnaround time and result quality matter more to him than a lower price alone. Based on this particular experience, he favors Codex and Claude, while acknowledging that his difficulty with other models may partly reflect his own workflow.
- 6:51:36
When to Commit in Git: Per Task or Per Milestone?
Luqman discusses an audience suggestion to make frequent commits so problematic agent changes can be reverted easily. He contrasts this with his current milestone-based approach, since he does not consider each intermediate part fully validated, while leaving open the possibility of changing the workflow once the application is in production.
- 7:00:42
Comparing Codex and Kimi K2 on a Practical Fix
The speaker said an attempt with Kimi K2 disrupted the work in this case, while Codex fixed the problem much faster by his comparison. He qualified the judgment as based on limited experience and a practical example, while questioning how well benchmark results reflect real-world performance.
- 7:04:52
Reviewing Approval-Flow Completion and Its Verification Checklist
The speaker reviewed completion of an approval-flow milestone and went through task checklist items covering notifications, a growth flow, trial or subscription cancellation verification, confirming an event appears in Sentry, and an SQL query for a shop row. The brief captions do not establish a separate result for every item.
Got a question about something in this stream?
Ask in the group