FIXED 1 BUG. CREATED 3 MORE.
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 focused on reviewing a Shopify app before launch, combining chat and merchant-UX refinement, fault diagnosis, and conversion of review findings into a debugging and testing queue inside a custom command center. Practical work covered crashes caused by products without images, disappearing text or attachments, slow replies, alignment problems, contextual actions, and confirmation states. A recurring conclusion was that a local fix is not proof of system readiness: human review and realistic scenario testing remain necessary. Luqman also explored agent-driven debugging through backend tests and log inspection, context management through documentation, state files, and fresh sessions, plus confirmation gates and Git checkpoints to constrain autonomous changes. Audience questions led to substantial discussion of Claude, Codex, OpenCode, OpenRouter, local models, and agentic coding tools. The practical view was that model choice depends on the task, quality, cost, and familiarity with the project, while benchmark tables must be interpreted through their methods rather than a single ranking. Many multi-model and cheaper-routing ideas remained proposed experiments rather than completed findings. The stream also covered validating product ideas with landing pages, taking products to market instead of repeatedly building, and moving from a useful first release toward scaling based on real usage. Visual brainstorming and interface design, SEO and GEO, precise product descriptions, and an Arab knowledge-sharing community were additional useful threads. Open questions remained around some latency and timeout causes, classifier-context cost trade-offs, automation reliability, and several fixes and tool trials that were not conclusively validated.
155 moments from the stream
- 3:10:18
Usage Limits and Their Effect on Debugging Workflows
The speaker says usage limits had previously been a major obstacle to working, but were no longer the main issue during the current debugging phase. This is a workflow-specific opinion, not a general performance test.
- 3:11:42
Slow Batch Processing Ending in a Crash
During the live work, the speaker observes that a product batch is taking a long time to process, attributes the delay to action execution through a model whose name is unclear in the captions, and ends the segment by noting a crash.
- 3:12:51
Diagnosing a Crash Caused by a Product Without an Image
The interface was found to crash when it encountered a Shopify product without an image. Luqman discussed handling this edge case, rerunning the fix, and reviewing the coding assistant’s work rather than assuming task completion meant the UI worked.
- 3:17:23
Prioritizing Core UX and Logic Before Speed Optimization
Luqman treated slowness as a separate concern and said the immediate priority was establishing the core logic and user-experience flow, with possible assistant-configuration improvements deferred until later.
- 3:19:50
Observing Late-Night YouTube Audience Activity
Luqman noted that YouTube analytics suggested audiences for this field in Egypt and the Arab region tend to be more active during the evening and late at night, matching the arrival of more viewers after dark.
- 3:21:59
Using Multiple Coding Models Alongside OpenCode
Luqman said he was considering using multiple coding models alongside OpenCode. The identities of the other models are unclear in the captions.
- 3:23:57
Exploring the Content Ops Scope Before Creating Tasks
Luqman reviews the newly added Content Ops domain, noting that most of it has not yet been implemented. He corrects the approach from immediately proposing tasks to first using explorer and researcher agents to identify the required work, then turning it into specific, consolidated tasks without excessive fragmentation.
- 3:37:07
Balancing Exercise With a Streaming Schedule
The speaker says exercising several times a week can positively affect work and considers how to arrange morning or evening streaming around a consistent gym routine.
- 3:38:31
Understanding and Verifying a Task Before Execution
The speaker demonstrates a task-review workflow: separate critical findings from deferrable ones, ask the agent to understand the task and inspect the code for an existing implementation, and explicitly avoid executing anything until the review is complete. The spoken result indicates that the task had already been implemented.
- 3:48:51
Freelance Platforms as a Side Income
A discussion of exploring work through platforms such as Upwork and Fiverr, including using an AI assistant to research whether a platform is suitable. The speaker presents this as a possible side-income option based on client-side experience, without guaranteeing earnings.
- 3:51:54
Perceived Improvement to a Rate Limit
During use, the speaker felt that a rate limit might have been fixed, but framed it as a personal impression rather than a verified test result.
- 3:53:42
Misleading Action Buttons at a Confirmation Gate
A UI issue is described in which quick-action buttons appear while the agent is waiting for a yes-or-no answer about a specific action, potentially making the confirmation state misleading. The task is then handed off for closer analysis.
- 3:55:06
Verifying Frontend Changes in the Browser
The notes say that a frontend change should be checked in the browser. The speaker then asks whether any server is open and requests that all open servers be stopped. The captions do not show a test outcome.
- 3:58:36
Linking Usage-Limit Messages to the Upgrade Path
A proposal to make “upgrade” in usage-limit messages link directly to billing, and to apply that pattern consistently wherever an upgrade is required.
- 3:59:19
Redesigning Quick Actions Around Conversation Context
A review of misleading quick-action buttons appearing while an agent awaits yes-or-no confirmation. The buttons should be prominent, tied to each conversation’s logic, and offer relevant actions instead of generic placeholders, following inspection of the current implementation and a proposed improvement plan.
- 4:04:15
Guiding Merchants Through Chat-Driven Actions
An explanation that merchants need clear guidance, especially when taking actions through a chat box is unfamiliar; contextual action buttons therefore play an important role in clarifying the next step.
- 4:10:10
Investigating an Assistant Tone and Emoji Issue
A proposed review of the task and codebase to identify whether an assistant personality, bubbly-tone, emoji, and brand-voice issue had already been addressed before making changes.
- 4:12:10
Context Limits in a Multi-Turn Failure Classifier
A potentially important issue is identified: a multi-turn failure classifier may see only the latest user–assistant message pair, limiting the context available for classification.
- 4:14:04
When Should an Agent Act or Ask for Clarification?
A discussion of handling vague requests proactively: infer the merchant’s intent and proceed with or propose a plan when appropriate, while defining when follow-up questions are necessary instead of becoming overly proactive.
- 4:16:00
The Gap Between System Instructions and Model Behavior
A qualified observation that the bubbly-tone issue may be model behavior: system instructions may already forbid it, yet the language model may still produce it.
Got a question about something in this stream?
Ask in the group