BREAKING MY APP TO FIX IT
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 an AI assistant for Shopify as development shifts from building to review, debugging, and UI/UX polish. Luqman evolves a command center that separates UI, UX, and backend verification from fix tasks, records outcomes, and supports step-by-step retesting, while routing work among Codex, Claude, and other agents according to the task. Practical work includes migrating the interface to Polaris, reviewing the homepage and report card, and diagnosing classifier, routing, conversation-button, image, live-counter, and relative-date issues. Some cases were verified as fixed, while others remained unresolved, including app slowness, conversation-flow problems, and a vision-analysis failure. Audience discussions cover model and coding-tool selection, voice-chat applications, and local versus cloud model access, with the comparisons treated as informal impressions rather than definitive benchmarks. On product strategy, Luqman argues for examining demand, unit economics, pricing, and customer-acquisition cost before scaling, while acknowledging that pre-validation is not the only path and that currency mismatch and purchasing power can complicate sustainability in some Arab markets.
105 moments from the stream
- 0:03:09
Adapting a Command Center for Review and Debugging
The speaker explains how the workflow changed when the project moved from building into debugging and UI/UX polishing. He added a calendar for completed tasks, an application overview, and a review section where he and an agent work through tests task by task, record passes or failures, and reopen items that need fixes.
- 0:08:00
Why TalkStore Is Being Redesigned with Polaris
The speaker reviews TalkStore’s interface inside Shopify and explains his decision to follow Polaris. In his view, it limits design freedom but aligns the app more closely with Shopify’s application patterns and review expectations, which he prioritizes over visual effects and customization before submission.
- 0:12:02
Shopify’s AI Toolkit as Competition and a Signal of Demand
The speaker discusses Shopify’s announced toolkit for using different AI coding assistants and models. He sees TalkStore’s potential distinction as operating inside Shopify rather than being limited to a command-line interface, and views the announcement as an encouraging signal of demand and healthy competition. He also raises a possible, unconfirmed future Sidekick integration for exploration.
- 0:18:37
Improving a V1 App UI with Shopify Polaris
The speaker explains that a V1 application does not need an extraordinary custom interface, but should meet an appropriate standard efficiently. Shopify Polaris can reduce design work by providing components aligned with Shopify’s design language.
- 0:20:11
Verifying and Debugging Conversation Flows with an Agent
The speaker describes starting a session with Claude or another agent to test conversation flows step by step, including expected replies, classifier output, and logs. He notes that parallel agents can work well for building, while debugging is harder to parallelize because each fix may reveal further issues.
- 0:23:18
Verifying That Task Dependencies Are Actually Unblocked
After the system reported that completing a task would unblock downstream work, the speaker checked the board and found those tasks still blocked. The example highlights the need to verify actual workflow state rather than relying only on an automated status report.
- 0:25:21
Routing Different Workloads Between Codex and Claude
The speaker assigns component-migration tasks to Codex while keeping Claude available for interaction-heavy debugging. The example shows routing work by task type and accounting for usage limits instead of expecting one tool to handle everything.
- 0:28:00
A Customizable Project Command Center
The speaker describes preparing a specific Markdown file that an agent can use to build a project command center. It can be tailored to the owner’s needs, such as adding a bird’s-eye view and a calendar of completed work to restore context and make progress visible.
- 0:30:52
Validate the Market Before Building a Large Project
The speaker advises confirming market need and positive willingness to buy before investing extensive time in a large build, since initial assumptions about demand may be wrong. He also offers the personal opinion that B2B SaaS companies can provide attractive margins.
- 0:34:42
Building a Custom Command Center with an AI Agent
The speaker explains that a command center is not a fixed template and can be tailored to each person’s needs. He proposes supplying a Markdown file that can be placed in a project so an agent can build a starting point that users can develop further.
- 0:37:01
Why Debugging and UI/UX Review Resist Full Automation
The speaker describes the current workflow as simple and fast, while noting that debugging, UI/UX work, and review require repeated back-and-forth and therefore are unlikely to be fully automated in the same way.
- 0:39:00
Agent Orchestration and Clarification on Ambiguity
The speaker describes a workflow where an orchestrator delegates research and exploration. Agents return questions when a task is ambiguous or needs an operator decision, while straightforward work becomes an instruction file for a builder agent.
- 0:41:40
A Shared State File for Tracking Agent Work
Project events are recorded in a shared JSON state file so agents do not lose track of the work. As the file grows, sub-agents can navigate it and research options, while the best choice for a specific case remains a human decision.
- 0:43:10
Delegating Research and Reducing Agent Context Usage
The speaker recommends configuring agents for the project and delegating research, while retaining highly precise work for a human when needed. To reduce token usage, he suggests giving an agent only the relevant section of a large file instead of loading the entire file into context.
- 0:47:45
A Dynamic Debugging Workflow
The speaker expects debugging to be slower and more interactive than parallel task execution, using a session with repeated back-and-forth. He also describes a dynamic command-center section whose task list grows as verification uncovers additional issues before launch.
- 0:50:32
Validating Shopify App Ideas Before Building
At the time of recording, the stated plan was to use the Shopify app-review period to test several other ideas with landing pages and marketing before building, then develop the ideas that showed real buying interest. The speaker notes that marketing a novel concept with no clear direct competitor may be slow because the market must first learn the idea.
- 0:54:18
Using a whiteboard for goals and high-level planning
The speaker wants to reactivate the whiteboard for high-level notes and daily or weekly goals instead of leaving it as background decoration.
- 0:55:27
Turning a migration task into a full code audit
After completing a Polaris task without personally reviewing it, the speaker asks for the QA task to cover every migration subtask and investigate the codebase for clean files and changes.
- 0:57:37
Explaining a project through streams instead of a dedicated introduction
The speaker says the project can be understood progressively through streams, while separate videos are more likely to address specific practical topics than serve as a personal project introduction.
- 0:58:41
Making task enrichment visible and addressing dirty files
After unexpected modified files appear, the speaker asks whether the enriched task will clean them up and why the added guidance is not visible in the task notes.
Got a question about something in this stream?
Ask in the group