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
- 0:32:53
AI assistants for different operational roles
The speaker compares AI assistant tasks with virtual assistant roles, including replying to comments and messages, editing short videos and YouTube content, and project-based work.
- 0:33:42
Turning a stream archive into a searchable knowledge base
The speaker describes work to transcribe streams into a searchable library, with a plan to move from processing the backlog to periodically adding new streams.
- 0:35:38
An audience suggestion to use an ad blocker
An audience member suggests using an ad blocker to improve browsing. The captions do not establish that a particular tool was installed or its effect measured.
- 0:35:58
Reverse engineering, assembly, and uncertainty about AI's role
An intermittent discussion covers the difficulty of reverse engineering before widespread AI use and the role of assembly. The speaker acknowledges uncertainty about tool differences and offers a tentative account of processing binary code, with reservations about output quality.
- 0:36:50
A community for people coding with AI
An invitation to a Telegram community for programmers and people using AI to code.
- 0:38:12
Requesting a cleaner, simpler design reference
The speaker likes the strategy but asks for a different design reference that is cleaner and simpler.
- 0:40:16
An Alexandria example in the style of GTA
The speaker refers to an example described as Alexandria in the style of GTA, featuring a yellow taxi, while explicitly questioning whether it relates to reverse engineering.
- 0:41:33
A prediction of broader open-source software adoption
After reacting positively to a taxi-game example, Luqman expresses a personal expectation that software is approaching a broad shift toward open source. The discussion does not establish that prediction.
- 0:43:21
Metal Slug and its local “Gulf War” nickname
The discussion compares an Egyptian-themed game with Metal Slug, then clarifies that participants use “Gulf War” as a local nickname for Metal Slug. Arcade nostalgia follows, along with an unanswered question about the nickname’s origin.
- 0:45:27
A general view of cybersecurity opportunities in the Gulf
Luqman suggests that Gulf companies could broaden cybersecurity opportunities beyond the Egyptian market, and tentatively mentions the possibility of remote work.
- 0:47:46
A coding-agent roadmap with periodic supervision
Luqman describes organizing a roadmap into milestones, missions, and smaller tasks, then assigning a milestone for execution while another agent periodically checks progress through a scheduled job. The passage describes the setup without establishing completion or success.
- 0:50:13
Is splitting tasks across sessions more efficient?
Luqman asks whether running a task set in one session or splitting it across several sessions would use less of the available allowance. The page ends during the question, with no answer or practical comparison.
- 0:51:20
Preparing task prerequisites to reduce human intervention
Discusses preparing permissions, keys, and other prerequisites so tasks can proceed without human intervention, while noting one remaining blocker that is not identified in the discussion.
- 0:52:52
Repeated file reading and cached tokens
Discusses how rereading project files and repeating comprehensive checks can increase consumption, explains the savings associated with cached tokens, and supports organizing files and sessions to reduce repeated work.
- 0:53:55
Task dependencies and periodic agent supervision
Acknowledges that some tasks block later ones and suspects that some enter review loops. Describes periodic supervision through computer interaction with the application interface, followed by audience discussion of longer checking intervals; the discussion does not establish that supervision resolves the blockers.
- 0:54:48
Reasoning effort, multiple agents, and multiple models
An audience member suggests that lower reasoning effort with multiple agents could improve speed and reduce consumption. Luqman explains his expectation that distributing work across different models could help; these are suggestions and expectations, not verified comparison results.
- 0:57:04
A platform concept that builds an agent team from a goal
Luqman proposes an open-source concept combining work management and agent monitoring. He envisions users stating a goal by voice and having a team assembled to pursue it, reducing direct interaction with commands and interfaces. He stresses the need for strong architecture and orchestration, while leaving feasibility unresolved.
- 1:00:13
Triggering Codex review with a signal from Claude
A proposal has Claude signal the completion of a defined block of work so Codex can review it, replacing fixed-interval checks with completion-triggered review. The idea receives support in the discussion, without a demonstrated implementation or test result.
- 1:01:16
Proposing GitHub event triggers for Codex
The speaker asks whether hooks could be more efficient than scheduled tasks and proposes triggering Codex through events such as opening a pull request or CI activity. The proposal remains exploratory, and the speaker is unsure what the existing follow-up process does.
- 1:03:30
Subtle continuous motion for a 3D element
The speaker requests subtle, continuous movement to make a 3D element feel alive, along with work on transitions. The captions do not clearly identify the element.
Got a question about something in this stream?
Ask in the group