Social Done. Search Next.
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 experiments with a multi-agent system and an automated, multi-platform social-content pipeline. Luqman discusses a possible shift from fixed loops and assigned tasks toward graphs and goal-driven agents with roles, contracts, memory, and hierarchy. He presents these as evolving ideas and experiments, not proven capabilities: agent communication cost, usage limits, stop controls, review, and reliability remain open problems. The practical work includes plans to generate video, text, voice-over, and captions, run the pipeline overnight with scheduled checks, and move gradually from human review toward automation while keeping personal posts under the account owner’s review. The stream does not establish that this end-to-end run was completed successfully. Audience questions broaden the discussion to connecting WhatsApp with AI tools, selecting models by total task time, retries, and usage rather than names alone, and protecting human judgment from model flattery or excessive autonomy. Other substantial topics include validating user problems, pricing around measurable usage costs rather than outcomes that cannot be guaranteed, the difficulty of acquiring traffic, and learning marketing, programming, and practical skills in the AI era. Additional discussions cover app-publishing and incorporation considerations in general, non-definitive terms; fair and secure agent competitions; visual summaries for long texts and streams; a proposed experiment with AI-generated musical variations; and lighter sports and workspace topics. The clearest conclusion is a preference for hands-on learning and building with human oversight and explicit operating limits. GitHub releases, autonomous operation, the virtual-team architecture, and several proposed tests and integrations remained plans or unresolved questions rather than confirmed results.
82 moments from the stream
- 0:48:35
Gradually Moving from Manual Review to Automated Publishing
The speaker described a gradual automation approach: review clips during an initial trial period, then allow the workflow to publish automatically if its output proves consistently acceptable. This was a future goal, not a confirmed result.
- 0:49:31
Beware of AI Sugarcoating
The speaker warned against treating a model's agreement as proof that an idea is good: it may contribute a useful suggestion while framing its response too approvingly. This was presented as a general opinion based on his experience with capable models.
- 0:51:02
Codex 20x availability and its use for planning
The speaker discusses an unverified report that the Codex 20x option was full or closed to the public at the time of recording, suggesting load or marketing as possible explanations without certainty. He also argues that reserving the Codex 20x option only for planning may be excessive and recommends checking availability first.
- 0:53:14
Can an agent open a separate session for an integration task?
The discussion raises a practical question: whether an agent can independently open a new session to work on a newsletter-publishing integration. The passage records the question and an intent to test the capability, but gives no confirmed outcome.
- 0:54:38
Taking a step back to move further forward
Citing an article, the speaker suggests that a temporary step backward can be a courageous choice that enables greater progress later.
- 0:55:02
A dedicated prompt-engineering agent
A dedicated prompt-engineering agent is proposed: it would receive the latest information and updates about a specific model and take raw ideas as input, but the sentence cuts off before explaining what it would do with them.
- 0:56:02
From prompt engineering to agent-system engineering
The speaker describes prompt handoffs between models as useful, but argues that the work is evolving toward designing systems of agents with explicit roles, responsibilities, contracts, shared goals, and workflow dependencies. He presents this as an idea still being refined and validated, not as a proven success.
- 0:59:22
Agent communication costs and a simple multi-agent explanation
The speaker warns that back-and-forth communication among agents can consume many tokens, making coordination efficiency important. He then gives a beginner-level explanation: a group of AI agents can communicate and work together toward a defined goal.
- 1:01:13
Reflection Layers Before Asking Questions
A discussion of a system that asks people to review a question through several layers and confirm that the message has a real purpose before sending it. The response frames this as encouraging private reflection first, although implementation details are not established here.
- 1:02:25
App Publishing and Organization Requirements
A question about differing app-store publishing requirements, particularly the possible need for a D‑U‑N‑S number for an organization account. The response was tentative and recommended checking with developers who have direct publishing experience.
- 1:04:40
Business Registration Does Not Ensure App Success
An opinion that forming a company or obtaining a D‑U‑N‑S number should not be treated as the central obstacle, focusing instead on the founder’s ability to execute. The speaker explicitly lacks detailed knowledge of the app, so the advice remains general rather than a product assessment.
- 1:07:46
Planning the Next 24 Hours of Content Output
A prompt asks for a specific forecast of content output over the next 24 hours, including how many posts will be produced and where they will come from. No answer appears in this excerpt.
- 1:08:36
Tracking AI-Agent Usage and Cost
A walkthrough of a per-agent section showing settings such as the model and reasoning configuration, plus records of what the agent does and how much usage or cost it incurs. The interface or product name is unclear from the captions.
- 1:09:27
Building Agents and Automation for Marketing Operations
A question about learning marketing and sales fundamentals while building coded agents with Codex to operate marketing functions and serve businesses. The speaker says the current work similarly focuses on agents and a marketing automation system, motivated by the difficulty of generating traffic.
- 1:10:28
Programming or Marketing in the AI Era?
As a personal opinion, the speaker argues that AI tools have broadened the ability to build, making sales, marketing, and market understanding more differentiating skills. He qualifies this by noting that learning programming still makes sense for people drawn to software development and problem-solving.
- 1:13:45
Combining University with Building AI Projects
Responding to a question about succeeding in AI without a degree, the speaker advises against turning the choice into a family conflict and suggests continuing university while building, learning, and working. He presents this as practical, context-sensitive advice rather than a universal rule.
- 1:15:22
Memory and Hierarchy for a Multi-Agent System
The discussion considers giving a supervising agent substantial memory instead of repeatedly querying worker agents, possibly using a graph to reduce token usage. It then develops a supervisory layer for coordinating several agents and compares the architecture to building a company from operating roles upward before adding senior management. Agent activation, review, and token usage remain unresolved.
- 1:22:35
Education and Practical Qualification in the AI Era
The speaker argues that AI is creating major challenges for universities while expanding opportunities for self-directed learning, but explicitly says he is not encouraging anyone to leave formal education. He recommends questioning the practical purpose of a credential and career path, and prioritizing real capability over certification alone.
- 1:24:30
Planning an Overnight Automated Run
The speaker outlines requirements for testing an automated process overnight: schedule the recurring run, monitor whether it proceeds smoothly, anticipate possible failure points, and ensure the machine stays awake. The captions describe a testing and monitoring plan, not a successful result.
- 1:27:47
Traffic Acquisition as a Core Product Challenge
The speaker describes traffic acquisition as one of the hardest challenges in building products and notes that the process involves mistakes and ongoing learning. He also argues that AI tools expand a developer’s capacity to work on the problem, without claiming a proven outcome in this segment.
Got a question about something in this stream?
Ask in the group