ALMOST 70% BUILT
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 documents the practical development of Burnish, a product beginning on Shopify and exploring how stores can improve their visibility in AI answers and platforms through GEO. Much of the work centers on the website and product message: clarifying the problem, solution, and customer outcome; structuring the homepage and pricing page; routing “Scan My Store” calls to action consistently; and refining cards, contrast, mobile layouts, scroll interactions, and a floating 3D logo. Visual reviews repeatedly led to revisions or removal of immersive effects when they distracted from the product, alongside recognition that excessive polish could delay shipping. The pricing experience advanced, but the homepage and especially the scan flow still required substantial frontend and backend work. The proposed scan collects a native Shopify store address and business email, then leads toward a deeper AI-visibility report. Important questions remained unresolved, including audit cost, the instant result or teaser, brand-name detection, usage limits, protection of shared public infrastructure, report-email delivery, and the terms of a refundable reservation. The discussion emphasized honest scarcity indicators and responsible marketing claims. Lighthouse and browser inspection provided preliminary performance signals, but measurements from an older site were not treated as final evidence for the heavier new version; screenshot and animation problems also remained under investigation. For implementation, Luqman compares Claude and Codex and assigns design, backend, migration, and audit work according to their perceived strengths. The workflow uses separated specifications, milestones, dependencies, a Command Center, coding agents, audits, and human visual review. His practical conclusion is that specific instructions, clean context handoffs, and deliberate sequential work are often more manageable than too many simultaneous sessions. The stream does not establish that every migration, audit, or automated workflow was completed successfully. The broader discussion turns to go-to-market work: validating demand before launch, building a list and organic distribution, warming an email domain, and testing advertising or direct sales in phases. Luqman argues that market response—not personal enthusiasm—is the real test of an idea, and that a strong product still needs marketing and distribution. He presents GEO and AEO as emerging relatives of SEO; based on his reading of trends, he observed growing interest during 2025 and predicted further growth by September 2026, explicitly as an opinion rather than confirmed market evidence. Audience questions also cover whether non-developers can build with AI, learning by doing, Micro-SaaS and B2B opportunities, model and hosting choices, and choosing Telegram or Discord for an Arabic-speaking community. Overall, this is an unfinished build-and-validation journey rather than a completed launch, with product viability and growth channels still awaiting real market evidence.
163 moments from the stream
- 6:14:20
Correcting the Exit Animation Toward the Logo
He reviews an exit animation that is not moving toward the logo as intended and asks for it to mirror the entrance zoom, with the green element centered inside the circle.
- 6:15:08
Sharing a Public Repository With Readiness Caveats
He says the Command Center repository is public and reusable, while cautioning that it was not prepared as a polished public product and lacks some installation guidance and documentation due to limited time.
- 6:16:18
Building a Product Around Real User Value
He argues that an application should deliver a genuinely useful outcome for people and merchants, and that the idea should be reworked if it does not create that value.
- 6:17:12
A Shopify Agent for Monitoring New Products
He proposes a Shopify agent that scans the product catalog, detects newly added products, and configures them automatically. It is presented as an early direction rather than a completed feature.
- 6:19:30
Delegating Exploration of the Remaining Homepage
He asks an agent to continue below a completed homepage section, inspect gaps and remaining content through the footer, then brief him and propose alternative design directions. He also prefers starting the task with fresh context.
- 6:21:55
Orchestrating Agent Execution and Review
He explains a workflow where task prompts are prepared and then executed in an automated mode, pausing for human confirmation or dependency conflicts. Agents enter Command Center to complete tasks, while reviewer agents check task and milestone results. He characterizes it as tailored to his workflow rather than a direct replacement for general project-management software, though it could be adapted for shared team use.
- 6:28:57
A Phased Plan to Complete the Website Pages
He requests a complete phased implementation plan covering unfinished areas after building full context across not only the homepage, but also the product-explanation and pricing pages.
- 6:32:18
Preparing to Move From Marketing Into Direct Sales
He discusses preparing interview scripts and a question framework before taking on direct-sales work, noting that his background is closer to digital marketing such as advertising, email, and SMS, so the transition requires new learning.
- 6:34:36
Replacing an Immersive Experience with a Simpler Design
A homepage review proposes preserving the important “why now” message and core content while removing or replacing an immersive experience that duplicates other sections and feels distracting. The discussion frames stepping back from a complex implementation as reasonable when the result is not clean enough.
- 6:43:58
Extending the Design Session Across the Full Homepage
After choosing an initial direction for one homepage section, the request expands into continued brainstorming across the full homepage so the design can be addressed cohesively in one session.
- 6:46:10
Automating Task Progression After Review and Audit
An explanation of a workflow that automatically moves tasks from review to done. Medium, large, and architectural tasks first pass through a post-build or post-deployment auditor, and their completion then unblocks tasks in the blocked section and moves them into the to-do queue.
- 6:46:56
Coding Assistants in Xcode and Shopify Tooling
A discussion of using Xcode for macOS development, with the speaker referring to assistants such as Claude and Codex inside the development environment, followed by a comparison with Shopify command-line tooling. These are the speaker’s remarks rather than documented tests of the capabilities.
- 6:48:45
The Idea of Shopify-Specific Skills
The speaker considers creating Shopify-specific skills, inspired by existing specialized skill packages, without claiming an implementation or test in this segment.
- 6:49:35
Delegating Design with a Clear Quality Bar
The speaker chooses to skip another detailed session and delegate the design direction while requiring it to meet or exceed the existing website’s standard. The segment highlights the value of shared context and an explicit quality bar when handing off a design decision.
- 6:59:17
Automating Review Approvals While Respecting Dependencies
A proposed workflow would approve review tasks, unlock the next wave, and then run an automated command across a milestone. The discussion emphasized ordering work according to dependencies and suggested reusing the workflow for product engineering after the website is complete; this page does not establish the execution outcome.
- 7:04:52
Question About Using DeepSeek with Claude Code
An audience member asked how DeepSeek could be used within Claude Code. No setup procedure was provided here; the speaker only explained in general terms that the current workflow focuses on Claude and Codex rather than additional models.
- 7:08:07
Configuring DeepSeek Through an API
A brief explanation of setting a DeepSeek endpoint, adding an API key, and selecting the model at runtime. The speaker says he will research the mentioned coding-tool integration further rather than claiming it has already been validated.
- 7:10:04
Delegating Work to Agents and Verifying Results
The speaker discusses executing work in phases with multiple agents, reviewing the result afterward, and providing feedback. He also requests checks on commands and prompts, with visual verification identified as the operator’s main verification duty at this stage.
- 7:12:17
Comparing Models by Testing Them on the Actual Task
The discussion presents differing views on trying several lower-cost models. One participant favors a Kimi model, while the speaker says it did not work well in his own use and that DeepSeek produced better results than the alternatives he tried, though it was somewhat slower. The exchange frames model choice as dependent on hands-on results and workload rather than a fixed preference.
- 7:18:48
Improving Project Presentation on a Portfolio Site
A suggestion to make project cards clickable, with each opening a dedicated page that provides more information about the project.
Got a question about something in this stream?
Ask in the group