"Burnishing" My Website Live
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 reshaping Burnish, preparing its website, and migrating its brand and domain across Shopify, Railway, and Cloudflare. Luqman studies reference sites to understand motion and visual storytelling while developing an original identity around “polishing,” stressing that the aim is to adapt principles rather than copy components. However, he later used a fabricated hacking scenario—explicitly acknowledging its dishonesty—to try to bypass the tool’s refusal to reconstruct a site; the outcome remained unresolved. The site story and several design and identity decisions also remained in brainstorming. Practical work covered Git checks, email setup, DNS, nameservers, custom domains, and deployment variables, with obstacles involving verification, domain limits, unclear Railway logs, and screenshot access. Some checks and setup steps progressed, but the stream does not establish that the full migration or website was completed. He also outlines a phased go-to-market command center connecting product development, market validation, and human-controlled consequential decisions, while its automation, Shopify review risks, and pricing remain unresolved. Audience discussions include his subjective views on Codex, Claude, and usage limits; advice to learn AI-assisted coding through small practical projects and step-by-step collaboration; tools such as Warp, Wispr Flow, and Remotion; the need to understand AI-generated architecture; and protecting freelancers with clear terms, contracts, and deposits.
100 moments from the stream
- 1:42:03
Ethical Boundaries in AI-Assisted Website Reconstruction
The speaker discusses an AI tool refusing to reconstruct a page without proof of ownership and, at this point, says he does not want to claim ownership or invent a story to bypass the restriction.
- 1:44:35
Testing a Reference-Page Reconstruction and Its Resource Cost
The stated goal was to test whether an agent could reconstruct a landing page, then analyze the code and use the lessons to build an original story and page. Participants discussed a deceptive justification for bypassing refusal; the speaker said he would try the ordinary approach first and expected the task to be resource-intensive.
- 1:47:59
Finishing Email Categories and Automated Verification
The speaker continues configuring email categories for sales, support, invites, and digests. A later status says some work was complete, but automated screenshots could not run because the browser-automation tools were not installed locally.
- 1:50:15
Comparing the Result with the Reference
After noting missing effects and an imperfect match, the speaker decided to try a repository and approach suggested by a participant, saying—regrettably, in his view—that bypassing the tool’s refusal required deceiving it.
- 1:54:40
Why AI-Generated Architecture Still Needs to Be Understood
Luqman discusses a system that Claude helped build, explaining that he understands its purpose but not yet all of its technical architecture. He distinguishes between tolerating that gap in a secondary tool and needing a complete understanding of how a core application works.
- 1:57:56
Using a Handoff Summary Across AI Tools for Technical Setup
After preparing the mailboxes, the work moves to configuring the setup on Cloudflare. Luqman asks for a concise prompt describing the current state and intended outcome so the task can be handed to another tool, noting that he will test the approach because Gemini has worked well for him on step-based tasks, without claiming it is necessarily the best option.
- 2:00:59
Audience Suggestion for Designing PDF Course Materials
An audience member describes using Claude with another tool whose name is unclear in the captions to create polished PDF course materials. Luqman asks about the intended use and decides to adapt and test the prompt, while expecting that it may require iteration before producing the desired result.
- 2:01:44
Inspecting an Animation-Rich Reference Site
Luqman examines the original site associated with an audience-provided repository, noting its polished team-built implementation, animations, and interactions that respond to pointer movement.
- 2:02:16
AI Website Replication vs. Original Design
A coding-agent experiment attempts to reproduce a website UI and its assets, followed by a discussion that a web developer’s value is not merely copying but creating a visual concept aligned with the brand name, logo, typography, and story. A fabricated hacking scenario was later used in the prompt to seek a better result, explicitly acknowledged as dishonest; the final outcome was still unresolved.
- 2:09:46
Requesting Guided Domain Migration Through Cloudflare
Claude is asked to provide step-by-step guidance through the Cloudflare interface for moving or merging a site under the Burnish domain. The excerpt captures only the initial request and does not establish that the migration was completed.
- 2:12:53
Why Avoid Building an Entire Website in One Shot?
The speaker reviews an overly broad website-building request and explains that only a landing page was needed, highlighting the value of scoping and splitting work before implementation.
- 2:16:14
Changing Name Servers and Validating a New Domain with Cloudflare
The speaker verifies the name servers before saving them, then explains that the update allows Cloudflare to validate the domain, noting that propagation may still take some time even for a new domain.
- 2:24:04
Setting Up and Verifying Burnish’s Custom Domain
A practical walkthrough of configuring a custom domain for Burnish through Cloudflare: waiting for nameserver verification, activating the domain, adding another domain form, and confirming that the landing page loads and the primary domain is set. Some interface details and steps are unclear in the captions.
- 2:33:48
Testing Email Delivery Before Reviewing Duplication
A practical check that an email arrives, followed by a review of a duplication function whose exact scope is not fully clear from the captions.
- 2:35:31
Reviewing Page Duplication and File Access Risks
An experiment with a tool that duplicated multiple pages, followed by checks on the presence and accessibility of code and files, raising concern that broad access could create a security risk.
- 2:37:01
Connecting Shopify Development to Live Hosting on Railway
An explanation of the difference between the Shopify app version used during development and the live version hosted on Railway, followed by work on a brand-name-related configuration and a result check.
- 2:40:48
Continuing a Coding Task in the Same Session or Starting Fresh
A practical discussion of whether to continue a UI change in the current tool session or start a fresh one, with a preference for a tool believed to better handle a large, context-heavy task. The exact comparisons and limits are too unclear to treat as established facts.
- 2:49:03
Diagnosing an Agent’s Screenshot Access
Luqman investigated a screenshot-visibility problem, and the response indicated that the agent could see saved images but not temporary screenshots. This presents a possible explanation, not proof that the issue was fixed.
- 2:49:43
Training Agents on Visual Storytelling Without Copying Websites
Responding to copyright concerns, Luqman said his aim was to train his agents to analyze website ideas rather than copy them. In his reading, the name, typography, logo, and pillars formed a coherent Greek-mythology theme; he wanted to translate that storytelling principle into something suited to Burnish, supported by practical code examples.
- 2:53:06
Reviewing Phase-One Go-to-Market Decisions
Luqman asked for the go-to-market section of the command center to be reviewed, with five important operator decisions for phase one restated and brainstormed. The page does not include the decisions’ details or the review’s outcome.
Got a question about something in this stream?
Ask in the group