Four Projects. One Flow.
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
Luqman and the audience explore a platform that would turn conversational or spoken goals into teams of agents for building products and running operations. Topics include roadmaps, role contracts, architectural oversight, and knowledge and context graphs; the concept remains under development and its feasibility unresolved. Token consumption, repeated file reading, review loops, and blocked parallel work recur throughout, alongside proposals for event-triggered supervision without a practical comparison establishing its advantage. Practical work includes reviewing an animated hoopoe website, exploring 3D and reverse-engineering tools, and demonstrating keyword search in a partially indexed stream library, with semantic search proposed. Content automation, reply approvals, and scheduling are also reviewed, including posting that reportedly continued after a pause request and unresolved performance and storage problems. Audience questions cover learning AI, choosing a specialization, and reaching customers; Luqman recommends planning distribution and using direct outreach for B2B. His predictions about voice interfaces and increasingly autonomous agents remain forecasts, while the work presented does not establish a completed platform or successful end-to-end automation.
153 moments from the stream
- 3:54:35
A brief view on opportunities for GTM engineers
Luqman briefly identifies go-to-market engineering as a field he expects to offer strong opportunities, without detailing its responsibilities or required skills.
- 3:56:49
Does a graphics tool wrap a model or train one?
Luqman speculates that an unclearly named graphics tool may wrap existing models, then acknowledges that its team might have trained models and that he does not know. A later audience comment mentions a strong training team, but the page provides no technical resolution.
- 3:58:02
Planning animations and reviewing an element’s position
The work review calls for identifying motions and animations in advance, then flags an animated box confined to the screen’s center instead of performing the intended entrance. Accepting the current behavior is considered, without a demonstrated fix.
- 3:59:51
Reading a brief app progress report
A report readout describes accepted and merged tasks and an app foundation with a frontend and secure sign-in. The excerpt ends during the feature description; these are reported claims rather than visibly verified behavior.
- 4:01:14
A question about training a model on specialized data
An audience member asks about building a model for domain-specific data or training an open-source model on it. Luqman requests a concrete example to clarify the data and goal; this page gives no training procedure or specific technical recommendation.
- 4:02:10
Using Command Center for marketing, research, and task coordination
Asked about uses beyond app development, Luqman recommends a clear roadmap, PRDs, and task breakdowns, then bringing that plan into Command Center to display swimlanes, agent assignments, and expected progress. He says the version discussed had not been updated since its April launch and advises adapting it, while mentioning a replacement under development at recording time.
- 4:03:49
From AI wrappers to harness engineering
Luqman advocates moving from AI wrappers toward harness engineering, describing it as powerful but double-edged without detailing implementation or risks here.
- 4:04:30
Prospect classification and an exploratory RAG discussion
An audience question proposes using labeled company data to assess whether a new company fits as a customer. Tentative RAG tool suggestions follow, with Luqman noting limited experience; no technical solution is established.
- 4:05:00
Building a roadmap with AI
An audience member asks about building a roadmap with AI. Luqman strongly endorses the approach but provides no steps or roadmap example here.
- 4:07:56
Searching a stream library and jumping to mentions
Luqman demonstrates keyword search across a stream library, using Codex to locate recordings and jump to mentions. He notes that indexing is incomplete and presents the library as a way to find relevant parts of long recordings.
- 4:09:26
A plan to turn streams into shorter videos
Luqman describes a planned workflow for extracting stream summaries into YouTube videos, including thumbnail adjustments. Completed results are not demonstrated.
- 4:10:08
Automating clips and comment replies
In response to a question, Luqman says Instagram clips use automation, as do comment replies on Instagram and TikTok. The technical setup is not explained here.
- 4:10:35
Adding semantic search alongside keyword search
An audience suggestion introduces embeddings for semantic search. Luqman plans to add them and organize the library into sections while retaining exact-word search; these are planned additions rather than demonstrated features.
- 4:15:43
Vectorizing the search query
An audience comment notes that the query must also be converted into vectors for semantic search. Luqman acknowledges the concept; no implementation or test is shown.
- 4:21:36
Agreeing on website motion before further generation
Luqman requests a section-by-section website walkthrough and agreement on motions and transitions before generating more. These units contain no response or implementation result.
- 4:22:45
Online learning and portfolios for AI engineering
Luqman tentatively agrees that learning online can remove the need to travel. In his view, a strong portfolio of AI projects may qualify someone for some AI engineering roles, depending on company size and role requirements.
- 4:23:43
Telegram or Discord for an Arabic-speaking community?
The speaker personally prefers Discord but considers Telegram better suited to the current audience because of joining friction and perceived Arabic-speaking audience activity. Expansion to Discord is presented as a possibility if the community grows.
- 4:27:25
Running long tasks while gaming
A suggestion to run lengthy tasks while gaming to feel that progress is being made during the wait. No execution result is presented.
- 4:28:04
Community growth through content and competitions
The speaker suggests that community growth could come from viral content or investment in growth, and proposes a possible competition as another way to expand the community. No results are established.
- 4:32:09
A concern about GPT over-researching simple tasks
A humorous observation suggests that GPT may gather information far beyond what a simple task needs, within a discussion of AI usage. This is an opinion expressed in the conversation, not a comparative test result.
Got a question about something in this stream?
Ask in the group