Inside the Core Engine
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 centers on developing the Burnish Shopify app, especially refining its multi-step onboarding so it feels native and consistent with Shopify, communicates merchant value clearly, and improves copy, cards, navigation, and pricing using visual references. The work remained unfinished: content and design revisions, backend wiring, deep-link behavior, and onboarding’s dependency on database state still required attention. In parallel, Luqman explores a safer Git workflow for multiple coding agents—cleaning the repository, isolating tasks with branches and worktrees, detecting conflicts before merge, validating development servers and URLs, and maintaining a living diagnostic log of difficult problems and fixes. Audience questions and broader discussions cover SEO and GEO, AI-driven product discovery, using funnel signals to assess demand, and practical advice for learning while building a SaaS product. The stream also contains personal, qualified impressions of Codex, Claude, Gemini, GPT, and coding benchmarks, stressing that outcomes depend partly on how tools are used and that informal comparisons do not support firm conclusions.
109 moments from the stream
- 3:48:35
Keeping Users Inside the Onboarding Flow
The discussion notes that a data-handling link can pull users out of onboarding and suggests removing or replacing it with an approach that communicates the information without interrupting the flow.
- 3:49:41
Turning Scan Explanations into a Strong Value Proposition
Rather than only listing what the scan measures, the review calls for more compelling language that explains Burnish’s value to merchants and gives them a clear reason to install, informed by a competitor-copy reference.
- 3:52:01
Aligning the Interface with Shopify-Native Patterns
The review emphasizes that the app should feel native to Shopify and questions using a separate interface-design skill without first confirming that it fits the project’s Shopify-specific tooling and conventions.
- 3:55:17
Clear Communication Around Branches and Worktrees
After a warning about another branch and an untracked file caused confusion, the speaker asks the agent to ignore work outside its task and proceed without unnecessary warnings.
- 3:56:54
Direct Access to App Sections in Development
The discussion covers starting the development server, checking currently running environments, and providing a URL that opens the app internally with access to settings and other main sections.
- 4:00:58
Deep-link routing issue in onboarding
A direct link to settings reached its destination, while links to other sections redirected to the welcome sequence; the issue was raised for diagnosis, with no confirmed resolution shown here.
- 4:03:40
Expressing product value from the merchant’s perspective
The interface review called for a more appealing presentation for merchants, a positive cue about what customers receive, and messaging focused on outcomes rather than back-end mechanics.
- 4:05:19
Sequencing onboarding work before the rest of the app
Because onboarding was tied to the database and repeated resets and per-worktree servers were cumbersome, the proposed workflow was to finish the onboarding sequence incrementally before resuming the rest of the app.
- 4:09:15
Highlighting visibility in AI product recommendations
The onboarding copy presents a benefit of seeing which competitors appear when people ask AI for product recommendations, including the relevant queries and placements.
- 4:10:04
Balancing heading semantics and visual size
The review noted the single page-level H1 and proposed visually enlarging an H2 to resemble a primary heading while preserving semantic structure, with compatibility with Shopify left as a question.
- 4:13:53
From Onboarding to the Core Engine: Staging App Development
A practical review points to the attached screenshot as a direct reference, then outlines a plan to finish and polish onboarding before moving to full application design and the core engine. The speaker says the engine architecture may take longer depending on standards and feature scope, while expecting the later Shopify audit to be quicker.
- 4:16:00
Comparing AI Models and the Limits of Coding Benchmarks
Answering an audience question, the speaker distinguishes general intelligence from coding performance when comparing Gemini, GPT, and Claude, but repeatedly notes uncertainty about interpreting the displayed agentic and coding benchmark sections. The claims are personal predictions, not results of a test conducted in the session.
- 4:20:14
Evaluating a Model Inside a CLI Tool
The discussion says that evaluating a model inside a command-line tool should include how it handles tools and context, then relays criticism of Gemini CLI reported by users and online communities. These are attributed reports, not a documented test performed during the session.
- 4:20:58
Consistency and Simplicity in Onboarding Step Two
An onboarding review argues that step two should be simple, straightforward, and visually consistent with step one. A competitor interface is used as an example, while the current version is described as inconsistent and not native-looking within Shopify.
- 4:22:03
How Should Agentic Coding Tool Comparisons Be Read?
The speaker distinguishes code-writing quality from an agent’s ability to work through a CLI and use tools, then examines what SWE-bench results mean and how strongly they shape perceptions of coding assistants. He notes that comparisons may depend on the model variant or operating setting, and says his own experience does not necessarily match the benchmark ranking. Some version names and figures are unclear in the captions, so they are not presented as confirmed details.
- 4:29:46
The Value of a Proactive Assistant Experience
The speaker says that, in his experience, the main appeal of Anthropic’s assistant is not code generation alone but its tendency to infer the intended direction and guide the next step instead of stopping after a short response. This is presented as a personal preference, not an objective model test.
- 4:31:46
Natural vs. Prompted Proactivity in AI Assistants
The speaker says Claude seems to infer the user’s direction and move the work forward naturally, while GPT may overdo it when configured to be proactive or to end every conversation with a question.
- 4:32:40
Matching Visual Grouping in a Reference Interface
During a UI review, the speaker explains that related elements should sit inside one white card because the current implementation makes them look like separate sections. The point highlights matching visual structure, not merely reproducing content.
- 4:37:20
An Opinion on Google and Gemini’s Position in the AI Race
The speaker argues that Google’s resources, infrastructure, and network give it a strong AI position, and that its relative lag is mainly in coding rather than the field as a whole. He presents possible future leadership by Gemini as personal speculation and suggests watching Google’s quieter releases.
- 4:38:57
Stopping Execution and Asking for Screenshot Interpretation Before UI Changes
After repeated attempts failed to match a screenshot, the speaker changes the debugging approach: the assistant is told not to execute anything, but first to describe what it sees and how the elements are arranged, then explain why earlier attempts failed. This separates visual interpretation problems from implementation problems.
Got a question about something in this stream?
Ask in the group