System Prompt Downstream Funnel
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 wide-ranging working stream combining a Burnish review and submission readiness with discussion of AI-assisted product building. Practical work covered installation, scopes, billing, the storefront, tiered human-approved Autopilot optimization, the brand manifesto, traffic and revenue dashboards, pixel/GA4 integrations, and meaningful empty states. Some narrow checks worked—such as a level transition and selected product-data updates—but the stuck optimization/publishing flow and missing analytics remained unresolved, prompting discussion of tunnel vision, error monitoring, debugging orchestration, and subagent workflows. The stream also explored product-scope discipline, capped trials, market and competitor validation, unit economics, and learning software development while building with AI. Luqman argued that deep software-lifecycle expertise is not an absolute prerequisite for starting, while making clear that AI tools do not guarantee bug-free production software. Audience questions additionally touched on APIs versus webhooks, local and future AI models, and context-window limits, alongside lighter conversations about stream music, World Cup predictions, and a virtual-office idea for tracking project progress.
69 moments from the stream
- 0:52:34
A QA Copilot That Orchestrates Bug-Fixing Sessions
A workflow in which a quality-assurance orchestrator spots, studies, and collects bugs before automatically spawning focused repair sessions. The speaker says this approach changed his debugging workflow and reduced the manual effort needed to start each session.
- 0:58:18
Diagnosing an Analytics Page With No Recorded Traffic
A post-deployment issue is described in which an analytics page remains empty, with no traffic or sessions recorded. Possible causes involving Google Analytics 4, a web pixel, and a potential authorization step are raised, but the cause is not resolved in this excerpt.
- 0:59:26
Do AI-Built Projects Still Need a Programmer?
Answering an audience question, the speaker explains that AI-built projects can reach production with bugs, whose number and difficulty vary with project size, construction approach, and the amount of auditing during development. He offers the opinion—not a guarantee—that someone skilled with AI tools may be able to begin without a programmer.
- 1:01:11
Understanding app embeds and the Evidence Pixel’s role in measurement
A request was made to explain the roles of three app embeds and whether a disabled Evidence Pixel could account for missing session and revenue data. The captions present questions to investigate, not a confirmed cause.
- 1:07:31
Choosing suitable background music for a stream
A light discussion about avoiding copyright issues and using ambient or oud music, with the observation that a meditation track had sounded like horror music to some listeners.
- 1:09:11
Reviewing feature redundancy and the revenue dashboard’s empty state
The discussion questions whether the Evidence Pixel or any of three features are redundant, then describes a revenue-page issue: cards appear disabled rather than showing zero, and no traffic-and-sales chart or table renders when data is absent. The suggested behavior is to keep meaningful empty-state components visible.
- 1:13:31
Reviewing the Revenue Page and Removing Redundant Elements
The discussion proposes removing an AI evidence pixel if it proves redundant, then reviewing the entire revenue page so its displayed metrics match the app’s scope and capabilities. The review is framed as something to brainstorm before making immediate changes.
- 1:14:27
Retesting Installation, Permissions, and Billing
After deleting the app, the installation flow was retested from the beginning. The review covered merchant-visible permissions, noting that requesting extra scopes can discourage installs, followed by checking the billing transition, approval behavior, and the next storefront step. The speaker noted that an anticipated error did not appear and that a product score looked correct, without presenting this as comprehensive testing.
- 1:20:46
Enterprise Mode for Using the Funnel Beyond Shopify
Luqman explains that Enterprise mode exposes the Burnish API, allowing its funnel to be used beyond Shopify and across multiple stores.
- 1:21:30
Testing Tiered Autopilot Product Optimization
Luqman tests an Autopilot flow that selects product batches using factors such as tier, product count, product score, and plan, then optimizes them. At level one, changes are not published without merchant approval; during the run he checks score updates and confirms that image alt text reaches the product, while identifying a UX issue he no longer considers a software bug.
- 1:23:57
A Brand Manifesto as the Source for Internal Prompts
Luqman explains a feature that builds a brand manifesto from business information such as mission, products, audience, and voice, then uses it to construct an internal system prompt that tailors optimization quality to the store. Merchants can edit the manifesto directly or redraft it by answering questions.
- 1:29:40
Proposed Metrics for AI-Driven Traffic and Revenue
A draft dashboard is reviewed with metrics for AI sessions, conversion rate, revenue, revenue share, performance by AI platform, and top AI-driven pages. Two labels are requested for removal, followed by a question about whether the remaining set is sufficient for a first version.
- 1:32:39
Testing Level Progression in an Automated Publishing Flow
The test state was adjusted through SQL and the flow was rerun to check whether adding products and advancing levels happened automatically after an earlier issue that required a page refresh. In the demonstrated attempt, the state advanced as expected, while other product areas remained in progress.
- 1:35:44
Measuring Product Contribution to Value
The discussion proposes a metric estimating the value attributable to Burnish actions compared with what might have happened without the product. It is described as double-edged: a positive result could demonstrate strong value, while a zero result could create a negative impression, so its presentation and interpretation require care.
- 1:38:44
Choosing a Free-Trial Length and Usage Limits
The answer compares starting with a seven-day free trial with the fourteen-day trial that, according to the speaker, most competitors offered. The shorter period may involve higher drop-off, while the longer one increases the cost of serving trial users. Usage gating is proposed as a way to test a longer trial without granting the plan’s full capacity, acknowledging that some users may complete their task and then cancel.
- 1:40:49
Designing a Capped Trial While Preventing Trial Resets
A discussion of offering users a meaningful trial while capping how many products can be processed, with the duration and limits to be revisited after market cost analysis. The speaker also flags a reportedly observed reinstall flow that may reset a trial and says it should be fixed.
- 1:43:31
Differentiating a Product Through Feature Depth and UX
A general comparison argues that basic optimization alone may not be enough to differentiate a product; clearer value can come from additional sections, options, automation, and useful scoring signals. Some feature names are unclear in the captions.
- 1:46:59
Checking a Batch Change and Representing Missing Data Safely
A request to verify what happened after a batch-related change in the backend or database, alongside a requirement that expected data elements still exist when values are absent and must not be populated with fabricated numbers. The type of those elements is unclear in the captions.
- 1:49:51
Reviewing the Level-Two Optimization and Auto-Publish Flow
The speaker outlines the expected level-two flow: select a product batch, optimize descriptions, titles, and metadata using GEO practices, then auto-publish some results according to risk. He also notes that an earlier part had not yet been tested and that the flow's success was still being checked.
- 1:52:14
A Multi-Agent Workflow Fails to Push Product Updates
The speaker shows a coding workflow that failed to push any products despite repeated attempts and many subagents, contrasting it with an earlier stage that had performed the push. The cause remained unknown at this point.
Got a question about something in this stream?
Ask in the group