Building AEO Scan Funnel System
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
A practical build session centered on preparing Burnish and its AEO scan funnel for Shopify merchants ahead of launch. Luqman reviewed the landing page, product demo, and scan-results interface; reviewed UI issues and requested fixes and alignment changes; and prepared a clearer video-editor handoff, while design polish, scan testing and delivery, email, and ad creatives remained unfinished. He explored offering an instant preview followed by a fuller report, balancing low signup friction against lead qualification because scans have an operating cost, and judging marketing through conversion and acquisition economics rather than assuming success. Technical work included examining crawlers, sitemaps, llms.txt, and robots.txt, alongside inconclusive discussions of privacy, application security, and reliance on Shopify; no comprehensive security test was documented. He also considered Graph Engineering and RAG as possible approaches to managing AI systems, while arguing that adoption should wait for a clear operational need. He examined role-sharing between Codex and Claude, context and usage-limit management, and task decomposition; his changing comparisons of speed and quality remained early impressions, not controlled tests. Audience discussions covered learning by building, freelancing, marketing and sales, e-commerce platform choice, and market validation, with emphasis on fundamentals and honest product claims and reviews. The supported conclusion is that the system was approaching launch and ready for further review—not complete or proven successful—while AEO remained an emerging, uncertain market requiring real-world testing and learning during the first weeks.
147 moments from the stream
- 1:37:11
Exploring Freelance Niches for AI-Assisted Coding
The speaker suggests researching Upwork profiles and services built with AI-assisted coding, such as web design and software work, to understand possible niches and market demand while distinguishing programming fundamentals from professional specialization.
- 1:40:05
Codex Session Communication in a Multi-Agent Workflow
According to the speaker, Codex had recently added communication between sessions, allowing one session in a multi-agent workflow to message another for the user to read. He says he was not aware of a separate integration he was asked about.
- 1:43:20
AEO Audit Workflow Summary
The proposed workflow asks five AI assistants store-related questions, then uses AI to rewrite one product page and develops a list of fixes with explanations of what they mean.
- 1:47:57
Coding-Assistant Effort Levels and Usage Consumption
The speaker discusses multiple effort settings in coding assistants and broadly compares his familiarity with Claude’s controls with his newer experience using Codex. He was initially satisfied with the current mode while usage consumption remained modest, though the differences between some levels were still unclear to him.
- 1:49:07
Why Follower Count Alone Does Not Reveal Engagement Quality
The speaker explains that weak engagement alone does not prove followers are fake: many may have arrived through a single viral post, without necessarily becoming consistently engaged with later content.
- 1:52:54
Choosing an E-Commerce Platform for a T-Shirt Store
Answering a question about building an AI-assisted T-shirt store, the speaker recommended using an e-commerce platform such as Shopify, BigCommerce, or WooCommerce instead of building the entire site from scratch, because established platforms simplify much of the store setup and operation. He contrasted this with software or service businesses, where a custom AI-assisted build may be more suitable.
- 1:54:33
Auditing the Audit Page UI and Change Safety
While working on the scan page, the speaker requested a visual widget and design direction, then asked for a quick audit of whether the recent changes could affect other backend behavior. The desktop call-to-action was also identified as broken, with an explicit request to confirm the issue before fixing it.
- 1:59:34
Showing an Instant Preview Before the Full Report Arrives
The speaker considered whether short session duration should be treated as a warning sign even after capturing the visitor’s email. He proposed improving delivery by showing a useful preliminary result within seconds, then telling the user that the full report would arrive by email a few minutes later, rather than relying only on email follow-up. This was a design proposal, not a confirmed implementation or result.
- 2:03:13
A Two-Stage AEO Report Funnel
The speaker outlines a funnel that provides a quick initial AEO report with immediate value and a cross-sell, then emails a fuller report after probing five AI engines.
- 2:07:31
Relying on Shopify for Application Security
The speaker says he relies heavily on Shopify for security while building the app. This is his view of the platform’s security responsibility, not a security test or proof that breaches are impossible.
- 2:10:20
An Early Positive Impression of Codex
The speaker shares an early positive impression of Codex, especially its perceived low usage consumption compared with alternatives that are not fully identified in the captions. This is an initial opinion, not a systematic comparison.
- 2:11:46
Pre-launch Readiness Review for a Shopify App
The project is described as usable now, but with acknowledged bugs and issues. A broader review and some changes are planned before publishing, advertising, and wider merchant adoption.
- 2:12:59
Turning Preparation into a Self-contained Codex Implementation Prompt
After accepting earlier recommendations, he asks for a self-contained prompt with an implementation plan, possibly as a Markdown file, to build the project with Codex. Following substantial preparation, he selects a high-capability setting and begins trying the mini-project implementation; no implementation result is shown yet.
- 2:23:08
Checking a Live-Site and Server Discrepancy
Luqman notices that the server state appears fixed while something still seems wrong on the live site, then later says the scan is working. The captions do not establish a confirmed root cause or completed test.
- 2:24:12
Natural Anxiety Before a Product Launch
Luqman reflects on feeling eager yet anxious before launch, describing uncertainty about the market’s reaction as natural after spending a long time building a product.
- 2:26:07
A Proposed Burnish Workflow for AI Answers
A draft product narration presents Burnish as analyzing AI assistants’ answers to buyer questions, combining signals about brand discovery and portrayal, then proposing product-content rewrites and additional answers. It says changes may await user approval or run automatically and that results are supported with evidence; these are marketing descriptions, not documented test results on this page.
- 2:29:10
Tracking Progress by Sessions Instead of Days
Luqman explains that the stream numbering represents work sessions rather than consecutive days because day-based labeling felt repetitive and could confuse viewers. The broader lesson is to use a progress unit that accurately describes the work cadence.
- 2:30:11
Refining Audio Transitions and Music Level
The experience review calls for removing an extra effect, using suitable fade transitions, and keeping the music at a consistent level throughout. Some effect terminology in the captions is unclear.
- 2:32:56
Reviewing Burnish’s Demo and Scan-Funnel Readiness
Luqman reviewed the product demo and desktop and mobile website versions, treating the essential website work as complete even though the desktop/mobile experience was not yet 100% finished, while the scan funnel continued in parallel. He also emphasized that cleaner input material tends to produce better output.
- 2:34:54
An Early Impression of Codex Speed Versus Claude
After starting to use Codex the previous day, Luqman said his early impression was that it executed faster than Claude, while stressing that this was about speed rather than overall superiority. He also mentioned the practice of having Claude review output.
Got a question about something in this stream?
Ask in the group