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
- 2:30:06
Cleaning the Repository Before Introducing a New Worktree System
After removing obsolete branches, worktrees, and unwanted files, the work moves toward designing a stricter workflow-management system. The stated goal is to begin from a clean state and work freely without branches or changes becoming entangled.
- 2:32:29
A Flexible Demo Mode for a Shopify App
A demo mode was proposed for switching the app’s view across different plan tiers, with a longer-term idea to simulate varied merchant and catalog sizes and reset the app and scans easily. The work was later deferred for the time being.
- 2:36:22
Automatic Agent Task Routing and Git Management
The proposed workflow lets the user describe a task while the agent decides whether it is new work and selects the appropriate branch or worktree using repository state and frontend, backend, or UX context. It should also surface untracked files or any unclean state immediately, either handling them or asking the operator what to do; continuous commits are mentioned only as a possible safeguard, not a settled decision.
- 2:42:13
When Should an Agent Use a Worktree or Pull Request?
A proposed workflow where not every task requires a worktree or pull request. Instead, the agent would decide when that isolation is warranted and when closing the task or opening a pull request is appropriate.
- 2:43:44
Organizing Work with Persistent Related Branches
An audience suggestion proposed a section that displays branches as related nodes and reveals each branch’s contents. The discussion interpreted these as persistent area-based branches whose completed work is merged through pull requests, while only tentatively comparing the idea to a polyrepo setup.
- 2:45:43
Repository State Checks and Uncommitted Work
The discussion proposes checking repository state at session start because accumulated uncommitted or untracked work later requires rereading, validation, and reconstruction of its branch context. It ends with a question about what “ahead of main” means and how that state arises, but this page does not contain the answer.
- 3:02:05
Building a Safer GitHub Workflow
The speaker explains the need to clean up the repository and organize branches and worktrees into a system that reduces manual intervention, merge problems, and repository file-state issues. He presents learning these tools through practice and establishing a clear workflow as necessary groundwork before continuing development.
- 3:03:34
From Design to Testing and Launch
The onboarding system remains unfinished, while the application design is nearing completion. The plan is to finish the design, test the application from several angles, and then launch.
- 3:04:46
Measuring Demand with a Marketing Funnel
The speaker discusses activating a marketing funnel and ads after releasing the application on Shopify to understand acquisition cost, cost per lead, other performance indicators, and the size and accessibility of demand. He also notes that install, usage, and LTV data become important once real users exist. Early results can indicate demand and funnel performance.
- 3:10:16
A System for Managing GitHub Branches and Agent Work
Explains a proposed system that reduces manual GitHub management by inspecting uncommitted files, dead code, branches, and worktrees, then choosing an appropriate branch or worktree for each task. The system was still awaiting practical evaluation; its aim was to reduce conflicts, lost work context, and distraction from the application itself.
- 3:16:26
Using Top-of-Funnel Signals to Test Product Direction
Discusses how Facebook top-of-funnel data may reveal demand and traction and provide early signals about whether a product direction is sound, while emphasizing that these signals do not by themselves prove the application will succeed.
- 3:23:27
Verifying a Merge and Cleaning the Git State
PR 54 is confirmed as merged, followed by a check that the final state is clean and a reminder to remove leftovers that could cause confusion before starting the next task.
- 3:29:22
Handing Off a Milestone as Independent Tasks
A concise handoff is requested for a new session to execute a technical milestone. The workflow treats each message as one task and lets the system handle workspace and branch setup without repeated manual instructions.
- 3:34:56
Starting Burnish’s Onboarding Sequence
A session begins work on the onboarding sequence of the Burnish Shopify app, using a server tied to the worktree for the section being edited to reduce interference and keep the development environment organized.
- 3:35:26
Impressions of AI Coding Tools and Models
The speaker discusses trying OpenCode with OpenRouter and several models, and gives the qualified opinion that GPT and Claude were stronger in his own experience at the time of recording.
- 3:38:26
Building a Handoff from Earlier Design Decisions
The session is asked to recover earlier design rules and page order—including the settings page and a self-scoring review system—and turn that context into a handoff prompt for another session.
- 3:39:31
A Rule for Validating Development-Server URLs
A hard rule requires the system to test the server and URL before returning them, and to keep self-diagnosing and retrying when they are unreachable instead of presenting a broken link.
- 3:43:30
Documenting Failures After Fixes
A self-diagnosis document in the project documentation is described as a place to record problems after they have been fixed.
- 3:44:00
Returning to Step One in Onboarding
A supplied URL opened the plan-selection step, while the Back action did not return to step one, so a direct step-one URL was requested.
- 3:47:21
Refining Step-One Onboarding Copy and Hierarchy
A review of onboarding step one proposes labeling the stated duration as approximate, making the AI-visibility scan prompt the main heading, and using shorter, clearer supporting copy around the continue action.
Got a question about something in this stream?
Ask in the group