Optimizing Sarmad
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 developing Sarmad as an interconnected agent system: defining roles and delegation paths, monitoring token usage, experimenting with Telegram-based control and reporting, and reviewing the interface to improve visual separation. It also demonstrates automated content creation and cross-platform publishing; only the social workflow had been verified, while other channels still needed refinement. A major theme is that building a product is insufficient without distribution. Luqman discusses traffic acquisition, measuring content impact, competitor research, gradual ad testing, and a possible shift from Sarmad’s free trial to a free tier. He stresses that limited traffic and data prevent firm conclusions about pricing or conversion, and that cost reductions should not override privacy or healthy margins. Audience questions cover choosing tools such as Claude and Codex based on practical efficiency and usage limits, learning through real projects, AI-engineering paths, marketing, freelancing, and workstation choices. Several ideas—including open-sourcing Sarmad, proactive agent hierarchies, traffic projections, and revised pricing—remain proposals or early tests rather than completed outcomes.
116 moments from the stream
- 1:58:03
Build for AI, Not Only with AI
The speaker argues that future opportunities may lie in building tools or services for the AI ecosystem rather than merely consuming AI, and offers the personal opinion that AI engineering will be highly sought after.
- 2:00:35
Vision and Graph Engineering Before Execution
Advice that execution begins with a precise vision of what to build, alongside studying and understanding graph engineering, then using AI to realize that vision.
- 2:01:27
A Personal View on Learning with AI Instead of Courses
The speaker gives a personal view against paid courses in general and suggests AI-assisted self-learning, while cautioning that learning without clear boundaries or direction can carry risks.
- 2:02:57
A Practical Path for an Aspiring AI Engineer
Advice for a student to build tools, websites, and Micro-SaaS projects useful for study or daily life, developing a portfolio through consistent AI-assisted practice. The speaker presents this as a suggested path, not an official standard for the professional title.
- 2:05:33
Open Source for Promotion and Portfolio Building
The speaker considers open-sourcing the tool and promoting it through platforms such as X. It could attract traffic and serve as a useful portfolio example, though the marketing benefit is presented as a possibility rather than a proven result.
- 2:06:47
Pre-Build Product Lessons: Competitors, Market, and Distribution
The speaker stresses close competitor research, explicit criteria for choosing a business area, and early study of monetization and distribution. In his view, a crucial question is how the finished product will attract traffic and users, since ignoring that can leave a built product with no traction.
- 2:08:32
Automating Provider Integration with an Orchestration Layer
A participant briefly describes an automation tool that works with a service provider and API, then connects it to an orchestration layer, and says it will become open source soon. The captions do not establish testing results or final readiness.
- 2:10:59
Can Courses Keep Up with Fast-Changing Technical Fields?
Responding to a question about a machine learning, deep learning, and NLP course, the speaker expresses skepticism about selling theoretical knowledge without applying it to a practical product. He adds that the pace of change makes comprehensive, current courses difficult to produce; this is presented as his opinion, not a verified assessment of any specific course.
- 2:14:49
The Rapid Evolution of AI Engineering
The speaker argues that the field moves very quickly and that topics described as prompt, loop, and graph engineering can shift in importance within months, so learning paths should be treated as evolving rather than fixed.
- 2:17:26
When Theory Goes Beyond Your Practical Needs
The speaker advises someone who already paid for a course to learn what they can from it, but argues that deep study of deep learning, NLP, computer vision, and model training may be unnecessary when the goal is to use AI to build products. In his view, practical projects and a portfolio deserve priority over prolonged theory.
- 2:20:23
Learning by Building with Claude or Codex
The discussion distinguishes learning to train machine-learning models from learning to use AI tools. The speaker suggests choosing a useful problem or small tool, building it with Claude or Codex, and iterating repeatedly, arguing that deeper learning happens during the act of building.
- 2:22:57
Does Vibe Coding Need to Be Studied?
Answering a question about graph engineering and vibe coding, the speaker recommends continually following current AI trends. He describes vibe coding as building things with AI even without a programming background—while acknowledging that programming experience is better—and frames it as a practice of building rather than a separate subject to study.
- 2:24:40
Define Why You Are Building Before Investing Your Time
The speaker recommends clarifying whether a project is for personal use, open source, portfolio building, or income, then judging whether its potential return justifies the time invested. He stresses that the decision depends on the builder’s own goals rather than a universal prescription.
- 2:25:22
Automating Cross-Platform Social Publishing
The speaker demonstrates a system that prepares scheduled posts, provides previews, and automatically publishes content and videos across several social platforms, with calls to action leading to a website. He says the system is operational but still needs refinement.
- 2:28:51
Why the First Revenue Changes How a Project Feels
The speaker argues that earning the first revenue, however small, is not by itself a final verdict on a project’s success. It can nevertheless make the idea feel tangible and give its creator a strong incentive to continue.
- 2:34:36
A Multi-Channel Plan for Solving Traffic Acquisition
The speaker distinguishes general traffic from targeted traffic, noting that targeted visitors may come through ads, data signals, or direct outreach. His broader traffic plan combines persistent social publishing, search and answer-engine optimization, programmatic pages, affiliate marketing, prospecting, and outreach, with paid advertising scaled only after early signs of return appear.
- 2:37:33
Choosing a Coding Assistant Under Usage and Budget Constraints
Answering a question about usage limits, the speaker compares several coding assistants and subscription options, tying the choice to the desired quality and value available under constrained resources. These are his opinions, including things he had heard or recommendations he received, not the result of a documented test in this excerpt.
- 2:40:52
Understand the Core Engine Before Proposing AI Improvements
The speaker instructs an assistant to study the Burnish application and its core engine deeply—including catalog automation and smart posting—before proposing improvements. The aim is to produce tangible ideas grounded in existing capabilities instead of fabricating suggestions merely in response to a prompt.
- 2:44:02
Efficiency With Limited Resources Requires Resourcefulness
The speaker argues that limited resources demand greater resourcefulness and active searching for methods that improve output and efficiency. He presents this as a general principle, not a guarantee of a particular result.
- 2:46:08
Subscription or API Access?
Responding to a general question about buying usage credits or subscribing, the speaker recommended a subscription instead of relying on APIs. This was practical advice, not a detailed cost comparison or tested benchmark.
Got a question about something in this stream?
Ask in the group