FIXED 1 BUG. CREATED 3 MORE.
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 wide-ranging working stream focused on reviewing a Shopify app before launch, combining chat and merchant-UX refinement, fault diagnosis, and conversion of review findings into a debugging and testing queue inside a custom command center. Practical work covered crashes caused by products without images, disappearing text or attachments, slow replies, alignment problems, contextual actions, and confirmation states. A recurring conclusion was that a local fix is not proof of system readiness: human review and realistic scenario testing remain necessary. Luqman also explored agent-driven debugging through backend tests and log inspection, context management through documentation, state files, and fresh sessions, plus confirmation gates and Git checkpoints to constrain autonomous changes. Audience questions led to substantial discussion of Claude, Codex, OpenCode, OpenRouter, local models, and agentic coding tools. The practical view was that model choice depends on the task, quality, cost, and familiarity with the project, while benchmark tables must be interpreted through their methods rather than a single ranking. Many multi-model and cheaper-routing ideas remained proposed experiments rather than completed findings. The stream also covered validating product ideas with landing pages, taking products to market instead of repeatedly building, and moving from a useful first release toward scaling based on real usage. Visual brainstorming and interface design, SEO and GEO, precise product descriptions, and an Arab knowledge-sharing community were additional useful threads. Open questions remained around some latency and timeout causes, classifier-context cost trade-offs, automation reliability, and several fixes and tool trials that were not conclusively validated.
155 moments from the stream
- 0:52:53
Playwright Constraints When Debugging Inside Shopify
A discussion of using Playwright MCP for debugging, with repeated Shopify login requirements creating a practical browser-access obstacle. It also notes that moving among multiple agents and conversations can make the workflow difficult to follow.
- 0:56:18
Fixing One Bug Can Reveal More
The speaker explains that the challenge is not merely finding debugging tools: resolving one issue can expose several more. He emphasizes that the product is still early and that the priority is to deliver something usable and genuinely valuable while continuing to identify and fix defects.
- 0:58:34
Tracing an Unclear Classifier Timeout
While reviewing a classifier-related failure, the speaker notes that working across multiple terminals caused confusion, then investigates where a short timeout originates and whether it was intentionally configured. The page does not show a final resolution.
- 1:01:39
Review Each Stage or the Whole Project?
A discussion of two software-review approaches: divide the project into stages and review each completed task, or finish the project first and run one comprehensive review session. The speaker says the choice depends on working style and that this project used the intensive end-of-project approach.
- 1:04:21
Testing Multiline Paste in a Chat Input
Practical work on a chat-input issue: multiline text paste was checked and found to be working, after which an unwanted scrollbar inside the field was identified for removal.
- 1:06:07
Broad Review vs. Debugging and Parallel Work
An audience comment suggests structured specifications and separated parallel work to reduce conflicts between agents. The speaker clarifies that the current session is a broad project review covering UI, UX, and bugs—not only technical debugging—and says the suggestion may be considered later when application sections need independent tracking.
- 1:08:54
Removing the Scrollbar Revealed an Alignment Issue
After the chat-field scrollbar was removed, the text appeared slightly distorted and no longer centered, prompting a request to inspect and fix the alignment.
- 1:10:26
Chat UI Distortion and the Push-to-Dashboard Button
The chat input was described as distorted on initial page load or hard refresh, then returning to its expected appearance after typing. The review also called for confirming that the fix section retains a button for pushing an item into the dashboard.
- 1:12:19
Dividing AI Planning and IDE-Based Coding Work
The speaker described a personal preference for using a desktop AI tool for brainstorming, vision and mission documents, manifestos, roadmaps, and consultation, while keeping most implementation work in a terminal and IDE. The distinction was framed as broad planning versus detailed building.
- 1:14:43
Testing the Chat Input with Long Text
A more complete review of the chat-input experience was requested, covering focus, pasting a large block of text, scrolling through it in both directions, and keeping the input intact rather than delivering a partial implementation.
- 1:19:43
A Process Freezing Before Completion
A brief note describes a process freezing midway and never completing, including a case with zero images produced. The product identity and full technical context are unclear from the captions.
- 1:20:26
Choosing AI Models Based on Task Importance
The speaker discusses balancing quality and cost: using a stronger model for difficult or high-value tasks and lighter models for simpler prompts. He argues that economizing during development can create extra debugging work, while tools such as OpenCode and Codex can provide alternatives when another tool reaches its usage limit.
- 1:26:47
Why AI Is Difficult to Keep Up With
The speaker notes that AI is advancing too quickly for anyone to remain fully current at all times, making continual and shared learning an essential part of working in the field.
- 1:27:37
Building a Custom Multi-Agent Command Center
The speaker compares an open-source specify-plan-implement approach with a customizable in-house system. He describes agents researching and exploring a task, producing an instruction file, and handing it to an orchestrator, with optional review, coding, and security agents. Because a larger agent pool consumed many tokens, he reduced it; his broader point is that building the workflow yourself makes it easier to tailor to the project and add features such as a calendar and a debugging review interface.
- 1:30:28
A Customizable Multi-Project Command Center
The speaker explains that a custom command center can adapt workflow logic per project and bring task boards, workflow lanes, and tools into one interface. In his view, its main advantage is not absolute superiority over alternatives, but fully understanding a self-built system and letting agents control it through an MCP integration.
- 1:35:03
Managing Agent Context with State Files and Fresh Sessions
The speaker discusses balancing agent usefulness against token consumption. His approach reduces dependence on one long conversation by storing project state, tasks, and milestones in shared files connected to the command center, then executing work through fresh parallel sessions. He presents this as his own experience, not a guarantee against context failures in every system.
- 1:40:47
Using a Calendar to Track Completed Work
The speaker highlights the usefulness of a calendar view that shows work completed during the day, then marks a task as done.
- 1:43:30
Evaluating Agentic Coding Tools
The discussion considers Verdent AI, Warp.dev, and other agentic tools. The speaker relays a participant's preference for Verdent and says he will add it to his evaluation list. Mentioned criteria include distributing tasks across agents, project understanding, prompt simplicity, learning curve, and execution-environment suitability; no completed hands-on comparison is shown on this page.
- 1:45:41
Investigating Possible Duplicate Task IDs in a Fixes Queue
A small command-center backend issue appears during the work. The speaker suspects that some tasks in the review area’s fixes section that can be pushed to the task board may have repeated IDs, offering a concrete hypothesis for further debugging.
- 1:52:48
Trying New Coding Tools Without Abandoning Familiar Ones
The speaker says he has not personally tried the tool being discussed, but saw it demonstrated and found it interesting. He still uses his current tool because it is comfortable and sufficient, while remaining open to occasional experimentation.
Got a question about something in this stream?
Ask in the group