BUILDING ≠ SELLING
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 balancing Shopify app development with early marketing. The speaker identifies the landing page and distribution as the current bottleneck, then works through a simpler site narrative, trust and waitlist conversion, section reviews, branding, logo variants, and social previews. A test sign-up was confirmed in Supabase and the OG image was reported as serving correctly, while Umami conversion-goal setup and analytics testing remained unfinished because of unreliable interface behavior. Because the guidance appeared to reference an older interface, research was requested for the UI available on April 2, 2026 before proceeding. The stream also presents a multi-agent workflow for task management, research, and building, alongside a theme-aware Liquid section-generation pipeline that validates syntax, JSON structure, and CSS scoping. A fragmented discussion proposes feeding validation errors back to the AI for another attempt, but does not establish a completed implementation or test. Shopify compliance rules, its distinct billing path, and post-deployment reviews are presented as launch requirements that belong in the implementation plan. An independent reviewer is discussed only as a possible later addition that could catch issues missed by the builder; the design of specialized review and security agents still requires research, and shared infrastructure for an app suite remains unresolved. In audience discussions, the speaker offers his personal experience of starting with a small toolset and a useful, enjoyable project despite lacking a strong programming background; he also explains short-form clip automation, context handoffs, and practical usage-limit management. Comparisons among tools and models are framed as preferences or uncertain observations, with benchmark rankings treated cautiously. Generated imagery, animation, multi-shot video, and an FFmpeg alternative are explored without claiming completed experiments, while extending Talk beyond Shopify is raised only as a future feasibility question.
59 moments from the stream
- 0:02:00
Prioritizing the Landing Page and Distribution Over More Building
The speaker says application development is moving quickly, but the landing page, distribution, and marketing have become the bottleneck. The plan is to finish the site first, then consider analytics, data from merchants, and an initial marketing strategy before returning to the main technical work, while potentially postponing a teaser video.
- 0:04:45
Considering a Standalone Open-Source Agent Command Center
The speaker considers separating an agent command center from the current project and making it available through GitHub or his website. He describes a setup with several preconfigured roles—such as security, building, exploration, and research—connected through MCP, but postpones the decision because preparing a standalone version would take additional time.
- 0:06:07
Using an Agent to Review Landing-Page Sections
The speaker prepares an agent to review the landing page section by section, with particular attention to the hero area, and asks it to identify missing sections and brainstorm additions before launch.
- 0:08:15
An Unverified Observation About AI Provider Usage Limits
The speaker believes the observed usage-limit or token issue appears to be provider-wide rather than a Cursor problem, and suggests increased demand as a possible cause. He presents this explicitly as personal, unverified speculation.
- 0:11:57
Designing a Landing Page to Build Trust and Grow a Waitlist
A brainstorm about landing-page sections intended to attract more merchants to an application's waitlist. The discussion emphasizes trust, lower friction, benefit communication, and objection handling, while deferring pricing because the page is currently explaining the concept. It also proposes offering some registrants a chance to try the application free as a conversion incentive.
- 0:15:21
Automating Short-Form Clips with StreamLadder
An answer to an audience question describing a StreamLadder workflow: submit a video to generate short clips automatically, lightly polish them, and schedule them. The aim is to reduce effort and preserve focus for building and AI-assisted coding.
- 0:18:30
Visualizing a Brainstorm During Planning
An exploration of a tool that creates a visual board alongside a planning or brainstorming session, making proposals easier to understand and compare through layouts such as icon grids and stacks, followed by consideration of a broader product-capabilities presentation.
- 0:21:03
Simplifying the Site Story and Checking LinkedIn Link Previews
The site narrative was simplified to focus on three domains instead of showcasing subdomains. A LinkedIn link-preview issue was then examined because the thumbnail had not updated, with a request to review the cause.
- 0:25:58
Handing Off a Website Improvement Plan Between AI Tools
An implementation plan was prepared for handoff to another tool, covering a new section, refinements to calls to action and wording, gradual polish, and possibly animation before shipping the site and moving on to marketing. Claude and Codex were mentioned in the workflow, without a confirmed final result.
- 0:33:35
Checking the Deployed Site and Its Social Preview Image
He asked to expose a file and launch a port for visual inspection, then noted that the OG image was being served correctly and moved on to check how the site banner looked.
- 0:34:38
What Does the Build-in-Public Channel Show?
Answering a question about the channel, he explained that it documents building and marketing software projects in public from the beginning, using AI-assisted coding, agents, and multiple tools while showing the work in detail.
- 0:37:56
Do You Need a Programming Background to Build with AI?
He said he began without a strong web-development or software background and argued that learning to communicate and collaborate with AI can enable non-specialists to build websites and apps. He also noted that he learned substantial programming knowledge along the way, presenting his experience as a personal example rather than a guarantee of results.
- 0:39:39
Improving a Landing-Page Footer from Reference Examples
He discussed the difficulty of meeting his high design standard for the landing page, then directed the agent to strengthen its bland footer by analyzing two reference screenshots and proposing ways to apply the same concept.
- 0:41:05
Shipping Before Every Detail Is Polished
Discusses shipping a product or landing page while acknowledging that more polish remains, in order to maintain momentum and move on to other work.
- 0:42:35
AI Session Context Compaction and Handoffs
Explains that long conversations may be automatically compacted, which can lose some older context, and suggests requesting a handoff prompt for a new session. The claim that performance worsens as context runs low is explicitly presented as unverified hearsay, while usage limits remain a practical concern.
- 0:45:48
Starting AI-Assisted Building with a Minimal Toolset
Argues that beginners do not need a large collection of subscriptions: they can start with a minimal toolset, discuss viable ideas with AI, assess them at a preliminary level, and create a roadmap or command center before building.
- 0:47:38
Reviewing a Blurred Image and Iterating the Interface
The review returns to an image that remains blurred in an inspection link, then the speaker describes what appears later as much better, with animation treated as a possible improvement for later.
- 0:49:29
A Custom Project-Management Command Center
Explains that the command center evolved across multiple sessions through a task board, workflow lanes, and ongoing additions and removals. AI can help build it, but the exact design need not be copied; the key is a system tailored to its owner that shows the schedule, roadmap, progress, and remaining time.
- 0:50:48
A Simple Workflow for Managing Agent Tasks
The speaker explains that the workflow does not necessarily need a complex command center or agents talking to one another. The essentials are a clear workflow map, a project-time estimate, and a task board for assigning work to the appropriate agent.
- 0:51:26
When Replit May Not Be Necessary
Answering a question about Replit, the speaker says he does not need it for his current case because he believes he can reproduce the required functionality with AI; this is presented as his own preference, not a universal recommendation.
Got a question about something in this stream?
Ask in the group