SPEEDRUNNING TALKSTORE
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 building TalkStore, envisioned as a Shopify agent that can take store actions rather than merely return text. Luqman tests the app incrementally, from fixing product access, enriching data, and preparing human test checklists to creating metaobjects and trying prebuilt storefront blocks after a write_themes exemption was rejected. The work exposes an unresolved architectural problem: deep links do not automatically bind a metaobject entry to a theme block, while alternatives involving an app embed, an app proxy, and how to decompose their implementation remain under consideration without a clear final decision. In one inspection, the feature grid did not appear even though no console errors were reported; in a separate post-deployment test, some UI content still failed to render as expected. He also discusses structuring debugging as its own milestone with specialized agents, using Command Center plus parallel sessions and agents to accelerate work, and exploring a multi-model strategy for AI-tool usage limits, while deferring UI/UX refinement until core logic is validated. Project planning includes reviewing task dependencies, blockers, and priorities, as well as weighing early submission against further polish amid uncertain Shopify review and opportunity cost. Audience discussions cover livestreaming as a focus and progress log, cautiously building a side project, validating demand before development through landing pages and early-commitment signals, and the speaker’s view that clear vision, strong goals, and self-confidence support commitment. A future commerce-intelligence milestone could answer merchant questions about sales metrics, while product scope, the post-launch workflow, and workable paths around Shopify’s constraints remain open.
47 moments from the stream
- 2:16:09
A Critical Storefront-Domain Issue
During debugging, the speaker explains that broken selection or URL logic in the storefront domain is critical because leaving it unresolved could make that functional area unusable. The discussion ends with the cause still unclear, not with a confirmed fix.
- 2:16:55
A Multi-Agent Workflow Specialized for Debugging
The speaker believes the current building workflow accelerates progress, while debugging can remain the bottleneck. He proposes a separate workflow with agents specialized in finding bugs and handling specific project areas, beyond general research and review roles.
- 2:22:03
Shopify Theme Alternatives Without the write_themes Scope
After an exemption request for the write_themes scope was rejected, the speaker proposes researching Shopify’s documentation for viable alternatives before executing anything. One possibility is providing prebuilt blocks through Metaobjects, though this is described as substantially weaker than the original idea.
- 2:27:48
Investigating Rapid Claude Usage-Limit Exhaustion
The speaker asks for research into user reports and Anthropic’s statements about unusually rapid session-limit exhaustion, seeking to determine whether it is a temporary issue likely to be fixed or a new usage standard. This page does not provide the research outcome.
- 2:34:24
Claude Usage Limits and Their Workflow Impact
A discussion of usage limits being reached quickly and interrupting work sessions. A cited explanation attributes peak-hour throttling to demand exceeding capacity and notes that a large context window makes conversations heavier. Some issues were said to be potentially fixed later, but this was presented as a reported expectation, not a confirmed outcome.
- 2:41:20
Evaluating App Embed vs. App Proxy and Task Decomposition
An initial review of whether an app embed block could replace an approach involving an app proxy and large blocks. The discussion asks why one option was recommended and whether implementation should happen in one pass or be split into multiple tasks. The exact problem and final decision are unclear from the captions.
- 2:43:39
A Multi-Model Strategy for More Reliable Coding Sessions
A proposal to research using multiple models instead of relying on one, distributing work between them to reduce session interruptions when token limits are reached. The discussion links higher output quality with heavier consumption and suggests improving token management, along with creating a dedicated debugging agent after the project is completed.
Got a question about something in this stream?
Ask in the group