TOFU Acquisition Strategy
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 wide-ranging practical work on Burnish, a GEO application for Shopify merchants intended to measure store and product visibility in AI answers and recommend catalog improvements. The session focused on rebuilding and simplifying onboarding, polishing the scan page, report-wait experience, and mobile UX, and reviewing plan selection, Shopify billing, and data-processing consent. In parallel, it developed a Meta/Facebook acquisition funnel built around a free report and signals such as scan started, scan completed, and lead, with a recurring view that the mid-funnel and user journey should be improved before ads run. The work also covered clean-state testing and troubleshooting 500 errors, scan failures, deployments, Turnstile, tracking events, Stripe redirection, and Supabase, Cloudflare, Railway, and GitHub integrations; several causes and outcomes remained unresolved, and end-to-end verification was not shown as complete. Performance testing found that reducing sampling depth did not deliver the expected speedup in one attempt, leading to discussion of progressive results and a clearer waiting experience. Audience questions and broader discussions covered home-hosted databases, building Shopify apps through vibe coding, learning programming with AI, GEO versus SEO and AEO, dropshippers as a possible audience, and personal comparisons of Codex, Claude, and agent workflows. The indexed topics also cover separate discussions of Shopify-app market economics and saturation, D2C and Egyptian market strategy, cash flow and AI-tool spending, internet and OBS troubleshooting, digital-access constraints, digital-nomad visas, and asynchronous communication. The supported conclusion is that clear value, low-friction onboarding, reliable funnel behavior, and sound measurement should precede advertising scale; final product readiness, several bug causes, and the best landing-page and consent structure remained open.
183 moments from the stream
- 9:06:30
Balancing Work and Rest During Long Sessions
The discussion emphasizes balancing work and rest during long sessions and notes that visible progress can help even when work advances slowly. Private personal and health-related details are omitted.
- 9:07:51
Planning a Progressive Report and Testing Scan Time
The discussion considers revealing the report progressively even though the overall AI commerce index cannot appear at the outset, with strong visual execution seen as essential. The proposed next step is to reduce the sample count, start the worker, reset the referenced store, rerun the scan, and measure the actual duration before proceeding.
- 9:16:42
Using Claude and Codex for an Application Redesign
The speaker describes a workflow in which Claude inspected the website, debugged issues, and generated prompts for Codex, saying this helped refactor and substantially improve the application design. He also notes that productive progress can vary sharply from one day to another.
- 9:21:13
Tying AI Tool Spending to Return on Investment
A general spending principle is discussed for AI tools: watch the burn rate before recurring revenue exists, invest carefully at the beginning, and connect each expense to an expected return before expanding subscriptions.
- 9:24:42
How Revenue Exceeding Costs Changes a Project
The speaker argues that the beginning is the hardest stage and that revenue exceeding costs can clarify direction and strengthen confidence in a product’s market impact. He notes that working for months without revenue can be discouraging, while later growth brings a different kind of problem.
- 9:28:37
Evaluating a Dedicated Facebook Ads Landing Page
The proposed work is to audit the site’s main pages—including the homepage, product explanation, pricing, and scan pages—and then weigh the pros and cons of a single landing page optimized for Facebook Ads. The suggested page would center on a visibility report, while the final conversion path remains undecided.
- 9:31:54
Inspecting Background Agents in Codex
The speaker praised Codex for exposing background agents, each agent’s chat, and the prompt assigned by the orchestrator, describing this visual oversight as useful compared with their experience in Warp.
- 9:38:10
Choosing One Clear Landing-Page Headline
The speaker proposed dropping the multi-variant optimization task and showing a single hero message, arguing that optimization across variants would be premature without sufficient traffic. They preferred a short AI-visibility idea but found “Stop being invisible to AI” potentially accusatory, prompting a search for concise, less confrontational wording backed by adequate sources before making a strong claim.
- 9:48:51
Diagnosing Dropped Frames and Network Settings in OBS
A troubleshooting discussion about stream issues and dropped frames: whether device heat in a hot environment could contribute, or whether the internet connection is the likelier cause. It also covers OBS's network section being disabled during the live stream, whether settings can be changed afterward, and how to investigate more deeply. A response suggests OBS is not intentionally throttling the bitrate and that the local connection may be imposing a limit, but the cause remains unresolved.
- 9:55:59
Preventing changes from the wrong worktree
The discussion examines why a change was committed from the funnel-verify worktree despite unrelated code being present. The explanation was that the choice was not thought through and the commit was made from the directory already open, followed by a question about adding a CLAUDE.md rule to prevent the mistake with Claude and Codex. This page does not show a final agreed rule.
- 9:59:13
Testing how reduced sampling depth affects scan time
The scan was described as examining the niche, catalog, and metadata and generating prompts across several angles. Sampling depth was reduced from three to one with an expectation that runtime would fall from roughly four minutes to a little over two, but the observed measurement was about four minutes and five seconds, so this attempt did not show the expected improvement.
- 10:04:42
Isolating Unrelated Tasks with Git Worktrees
A proposed coding-workflow rule checks the current branch and worktree before acting, creating a separate worktree when a task is unrelated to the active work. The discussion explains how this avoids trapping unrelated changes on a branch that cannot currently be pushed, then follows a pull-request merge and confirmation that the deployment turned green.
- 10:17:47
Landing-Page Routing for Paid and Organic Traffic
The discussion tentatively settles on optimizing the existing scan page for ads, then raises an open question about whether paid and organic visitors should see different landing experiences that both lead to the scan. The excerpt ends before the clarification is completed.
- 10:20:04
Questioning a GitHub Workflow Change
A discussion questions why the GitHub workflow shifted after hours of work toward small hotfixes and separate branches, noting that the workflow took time to understand and that changes had previously been previewed without those steps.
- 10:24:22
Planning a Mobile-First Conversion Landing Page
The discussion favors a simple page-scan approach without depending on UTM, then defines a concise, informative landing page by starting with copy, content, and structure before visual design. It emphasizes understanding the conversion funnel, requesting a Shopify store URL and business email, designing mobile-first, and briefly educating visitors about the relatively new GEO domain.
- 10:31:59
Structural Questions for an AI Commerce Visibility Scan Page
A review of page decisions covering navigation, how much explanation the AI commerce visibility and GEO concept needs, whether a short primer or explanatory block is sufficient, and the pattern for the final mobile scroll-back call to action.
- 10:33:21
Positioning the Form Above the Fold on Mobile
The discussion revisits the hero hierarchy and proposes showing the form at or near the fold on mobile and desktop. It compares arrangements of the headline, animated demo, trust strip, and form—including putting the form before the demo or compressing the layout while retaining the demo first—so visitors encounter the action early and can scroll for context.
- 10:38:21
Critiquing a 3D Effect and the Current Logo
A visual critique notes that a graphic reportedly uses an unclearly named technique to create a 3D appearance. The speaker says it is not the best logo they have made and stops short of claiming it will remain forever.
- 10:39:11
A Structured Logo Need Not Explain the Product Literally
The discussion favors a distinctive geometric logo that makes the brand feel structured without literally depicting SEO or GEO. The initial logo was made in Canva, then an SVG was used to create a floating 3D glyph for the website, while acknowledging that the design could be improved or replaced later.
- 10:40:03
Positioning the Offer as a Deep Audit, Not a Quick Scan
The review notes that the process takes about two minutes rather than 60 seconds, so the inaccurate timing claim should be removed. It also recommends presenting the result as a genuine deep audit offered for free, rather than describing it as merely a scan.
Got a question about something in this stream?
Ask in the group