AI COMMERCE INDEX SCORING 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 documents a build-in-public effort around an engine for measuring how visibly Shopify merchants’ products appear in AI answers, alongside a Command Center for classifying agent tasks by size and dependency, executing them, and reviewing them through task- and milestone-level audits. Practical work covered code-informed prompt preparation, correcting illogical dependencies, monitoring tasks and logs, and improving the README and installation options. Although some workflows were described as becoming more autonomous, defect review, the frontend, and several implementation decisions remained unfinished. Luqman also compared Codex, Claude, and other tools from personal experience, consistently qualifying observations about speed, quality, resource use, and session variability as non-benchmark impressions rather than settled conclusions. Audience questions and wider discussions covered market validation and marketing before Shopify approval, organic outreach versus paid acquisition and waitlists, landing-page structure, animated backgrounds and calls to action, and GEO/AEO compared with traditional SEO. Other useful threads included choosing Telegram or Discord for an Arabic community, learning programming, AI, and DevOps through practical projects, data residency, hosting and DNS, plus ideas for automating marketing and short-form clip extraction. A recurring conclusion was that building a product does not automatically solve distribution; multi-platform expansion, the pre-launch strategy, Hermes adoption, and parts of the website and domain work remained open or under investigation.
123 moments from the stream
- 1:09:32
Organizing Project Prompts by Milestone
A practical task proposes giving each milestone its own subfolder inside the prompts directory and placing that milestone’s prompts there to improve organization.
- 1:11:55
Comparing Subscription Tiers and Usage Capacity
A brief general comparison of monthly subscription prices and the amount of work they were believed to support; the product and precise tier details are unclear in the captions.
- 1:14:30
Proposed Automated Task-and-Agent Workflow
The speaker asks whether an automation command will launch agents for medium tasks, unblock dependent work, and continue approvals unless an auditor agent raises a red flag. He also asks whether the flow stops at an unprepared large-task dependency; these are proposed behaviors, not confirmed results.
- 1:19:06
Configuring Safe Stops and Starting the Workflow
The speaker accepts having automation stop at an unprepared task and requests a clean setup that defaults to subagents. He then starts the automation command and leaves the command center open to see whether two tasks are completed; this page does not establish the outcome.
- 1:21:23
A command center that unblocks dependencies and continues automatically
The speaker tests a workflow in which tasks move through execution, review, and completion, automatically unblocking dependent work and starting the next task. The stated aim is to prepare a large task set and run it with minimal intervention while observing the system before expanding its use.
- 1:22:06
Test the market before becoming attached to a product idea
Responding to an audience discussion, the speaker cautions that a similar product’s success is not sufficient evidence. In his view, the idea should be backed by market research and demand testing, a real problem solved for a defined audience, and a plan for profitability, distribution, and monetization.
- 1:32:46
Preparing an automated workflow with a CLI and two-layer auditing
The speaker describes an evolved command center with a CLI and commands that prepare task prompts and run an entire milestone automatically until a problem is encountered. It handles task dependencies, runs agents in parallel, and applies one auditor after each task plus another after each milestone; he says the milestone audit found issues missed by the individual reviews.
- 1:32:52
Planning a GitHub release and incorporating workflow feedback
The speaker says he plans to publish the command center on GitHub and describes incorporating useful workflow suggestions, such as milestone-based commits and shortcuts for copying task identifiers. This is presented as planned and ongoing work, not a completed release.
- 1:35:27
Milestone auditing and a qualified practical comparison of debugging agents
The speaker discusses the difficulty of debugging agent-built projects and presents per-task and per-milestone audits as useful review layers. He also recounts an uncontrolled prior experience with a model whose name is unclear in the captions; according to his account, it failed to fix a problem, while Codex reviewed the result and corrected it faster. He frames this as reviewable practical experience, not a general benchmark.
- 1:42:17
Telegram or Discord for an Arabic Community?
The speaker considers Discord stronger for organizing communities and notes that it has expanded beyond its gaming roots. He initially chose Telegram because the target audience in Egypt and the Arab world is more familiar with it, reducing download and onboarding friction. A gradual move to Discord is presented as a possible future option if the community grows, not a firm decision.
- 1:46:54
Cleaning Up Command Center and Auditing Agent Tasks
The work includes removing documents unrelated to Command Center before preparing a strong README. The speaker also explains that tasks subsequently pass through an auditor agent that reviews the code and other outputs, with an audit indicator showing that the review occurred.
- 1:51:23
Data-Residency Requirements for App Deployment
A participant reports encountering an obstacle while trying to deploy an app in Algeria: he was told that its data center had to be located inside the country. The example presents a reported data-residency constraint without independently establishing its legal scope or current status.
- 1:51:54
Data Residency and Local Hosting Options
The speaker discussed whether a data center could be hosted in Algeria, explaining generally that some countries require user data to remain within their borders. He cited Saudi Arabia as an example but said he did not know Algeria’s requirements or infrastructure. Asked whether an old laptop could serve as a server, he acknowledged limited hardware expertise and suggested deeper research or asking specialists, mentioning a VPS only as a possible option rather than a confirmed recommendation.
- 1:55:54
Coding-Agent Usage Limits and Workflow Impact
The speaker compared his practical experience with AI coding tools such as Codex and Claude, focusing on the difficulty of judging session and token consumption while several tasks are running. In his view, anxiety about exhausting an allowance disrupts development flow, and he described keeping side GitHub work separate from the main task. These were personal usage impressions, not a controlled performance test.
- 2:00:53
Model-Release Rumors: Speed Versus Workflow
The speaker relayed unconfirmed talk of a possible model release the following day, with no clearly stated official date, then said that workflow quality mattered more to him than speed alone. Some model names in the captions are unclear and are therefore not presented as facts.
- 2:04:38
Why direct AI model comparisons are difficult
The speaker said two models may accomplish similar work despite having very different philosophies, so he was unsure whether one model’s apparent overall strength would make it better in practice. The captioned version names are unclear.
- 2:06:13
Scoring product visibility in AI answers
The speaker described an engine that scores how visibly ecommerce products appear in answers from services such as GPT, Claude, Perplexity, and Gemini, while also considering competitor visibility. He characterized it as a substantial project that also tests autonomous task execution.
- 2:10:07
Autonomous task execution through dependencies and blockers
The speaker reviewed a workflow in which an agent continues milestone tasks according to dependencies rather than a fixed order, unlocks work as it becomes ready, and stops when it reaches a blocker requiring intervention. He also wanted to clean up the repository and clearly document the tool’s purpose and behavior for users.
- 2:11:48
An Opinion on Google’s Releases and Antigravity
The speaker views Google as known for silent releases but still needing to prove its presence. In his impression, Antigravity was the most prominent product it pushed strongly before its market presence later faded somewhat.
- 2:13:21
Should We Stop Learning Programming Because of AI?
Responding to an engineering student, the speaker gives his personal view that learning to use AI is most effective through building real things and interacting with models during the work, rather than relying only on courses and videos. He distinguishes this from studying AI and model internals in depth, and strongly favors hands-on learning without claiming that programming should be abandoned.
Got a question about something in this stream?
Ask in the group