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
- 1:57:34
Refining a Report Card and Its Action Button
A practical report-card design iteration: testing a larger button in the brand color, then moving it to the right while preserving the card size and realigning the text and related metadata.
- 2:02:39
An Action-Capable AI Assistant Inside Shopify
An explanation of a Shopify app concept that acts as an AI assistant for merchants, answering questions about store activity and products while aiming to execute actions as well. The discussion also briefly covers validating market demand before expanding an app portfolio.
- 2:04:06
UI Iteration Can Reveal More Backend Work
Further review rejects some color and alignment choices, then explains why UX-flow refinement and verification can be substantial work: a problem uncovered during a frontend task may create an additional backend task instead of immediately shrinking the backlog.
- 2:08:13
Improving the AI-Readiness Report’s Placement and Copy
A UI review proposes better alignment for the readiness badge and report button, plus stronger, more concise copy that is not limited to products because the scoring may cover several domains.
- 2:09:30
Choosing Different AI Models for Coding Tasks
A discussion of assigning frontend, backend, debugging, and review work to different AI models. The speaker argues that frontend output is easier to inspect visually, while backend work and debugging demand greater scrutiny, and notes that model performance can be inconsistent or frustrating.
- 2:12:15
Strengthening the Primary Call to Action
After deferring a status-copy change, the work moves to the main card: keep it, add a subtle branded treatment and lightly tinted background, and enlarge the button so the card reads more clearly as the primary call to action.
- 2:15:42
Deferring UI Polish to Focus on Conversation-Flow Fixes
The host chose to postpone UI polish because it required more focused attention and instead address UX issues identified after conversation-flow verification. He prepared a new session with the needed context to work through the version-two fixes step by step.
- 2:18:26
Clarifying MCP Task-Management Capabilities
When MCP appeared not to provide the requested session details, the host clarified that the system could access tasks and create tasks and milestones, while referring to a database MCP server backed by local files.
- 2:43:23
Identifying an Agent Tooling Gap for Reading Subtasks
A review of a Command Center section showed that the agent could identify sessions, count items, and create tasks, but could not read existing card subtasks through the available MCP tools. The gap was confirmed and work began on adding a new tool.
- 2:46:28
Reviewing Interface Colors and Visual Contrast
Luqman criticized the blue background for not matching the brand’s teal color and suggested a black button background and black text to improve contrast, while noting that the overall design still needed improvement.
- 2:47:51
Testing Agent Access to the Conversation Flow Section
After MCP adjustments, the agent session was restarted and asked to read the “Conversation Flow v2” section in the Command Center. This was framed as a small debugging step in the user experience and conversation flow, since the agent had previously seen task counts without their contents.
- 2:48:38
AI Apps for Algeria and the Unit Economics Challenge
Responding to an idea for an ecosystem of apps serving daily needs in Algeria, Luqman said AI could positively affect people across Algeria, Egypt, and the wider Arab world. He distinguished strong demand from the ability to build a sustainable model and raised unit economics as a central challenge; the excerpt ends before the explanation is completed.
- 2:49:51
Currency Mismatch in Unit Economics
Explains that paying fixed or variable costs in dollars while earning in a volatile local currency can undermine unit economics even when demand exists, while a dollar-pegged currency may reduce that risk. This was a general assessment, not a completed financial calculation.
- 2:51:40
Classifying Agent Tasks and Fixing Bottom Spacing
The agent was asked to log tasks and identify which require operator input versus backend work. A UI review found the homepage broadly acceptable but recommended lifting the bottom box slightly so it does not touch the page edge.
- 2:53:21
Start With a Profitability Structure
Discusses why ads or apparent demand alone do not establish an app's viability. The target demographic, expenses, paid-service costs, profitability structure, and monetization strategy should be assessed before development begins.
- 2:56:19
Test Conversation-Flow Issues Before Closing Them
A routing issue was not accepted as resolved merely because no code change was proposed; earlier tests had partially passed or failed. The flow was to be tested again, especially because actions such as rewriting descriptions and showing affected products had previously appeared out of context.
- 2:58:23
Tie Customer Acquisition Cost to Profit Margin
Explains that an ad-supported model needs a repeatable way to acquire users for less than the expected profit. The acceptable acquisition-cost ceiling should follow from the profitability structure and margins, with market and cost testing performed before committing to a build.
- 3:00:51
Test Five Ideas With Landing Pages Before Building
Presents an app-validation method: choose several ideas with suitable margins and monetization, create lightweight variants of one landing page, then run ad campaigns and compare attention, waitlist sign-ups, and acquisition cost. Weaker ideas are dropped and the strongest is built quickly. This is a proposed method, not a completed test result.
- 3:03:58
Guided Testing for Conversation Quick Actions
Future conversation-trigger tasks were requested as exact, executable steps and prompts so the operator can run a real test and report the outcome. In the debugging example, a quick action did not trigger the expected onboarding scan or batch score, indicating a classifier or routing problem that still required a fix.
- 3:06:30
Pre-Validation Is Not the Only Path
Adds an important qualification: validating the market before building is not the only way to start a product, and a founder may build directly when conviction is strong. Talk Store was cited as having been built without prior market validation.
Got a question about something in this stream?
Ask in the group