Meet the Swarm
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
A wide-ranging, practical stream about building AI products in public. Its strongest thread is the shift from short tasks to long-running “missions” pursued by agent swarms, supported by a control center and living roadmap that expose progress, quality, and blockers. Luqman discusses requirements documents and system design as foundations for greater agent autonomy, splitting work between Codex, Claude, or specialized models, and measuring real cost per completed task rather than advertised token prices. These are presented as concepts, early experiments, and personal judgments—not settled results. Practical activity includes reviewing an interface and identifying a scrolling bug, requesting an interactive mock to clarify the design direction, and beginning an application walkthrough by granting access to a sample project and requesting a quick capability test. The stream also discusses web testing and continuous monitoring, with the explicit caution that agent-driven UI checks do not prove scalability. Audience questions broaden the stream to RAG and learning resources, context compaction, usage limits and tool handoffs, handwriting OCR, reviewed data-entry automation, and disclosed AI assistants for social engagement. It also covers app frameworks and community platforms, product validation, free-plan pricing, distribution and sales, SEO, ads, direct outreach, freelancing, and uncertain effects of AI on jobs and SaaS. Open questions remain around Mission Control’s interface architecture, subagent concurrency, the value and cost of bridging models, and whether several proposed integrations and workflows will work in practice. The supported conclusion is that documentation quality, low-friction onboarding, human oversight, realistic usage measurement, and distribution matter at least as much as model capability; many demonstrated products and workflows remained exploratory or unfinished.
124 moments from the stream
- 1:46:03
Using Computer Control for a Codex–Claude Discussion
A suggested method has one assistant use a computer-control tool to open a session with the other and continue discussing a topic until they reach an agreement. The speaker noted that the second assistant may treat the messages as coming from the user unless the context is made explicit.
- 1:48:50
Why System Design and Project Documentation Matter Before Building
Answering a question about architecture challenges, the speaker qualified that he does not have a programming background and does not deeply inspect the architecture himself. His takeaway was that strong system design, requirements documents, and diagrams provide a clearer foundation, while acknowledging that early problems may have occurred without being noticed.
- 1:53:31
AI and Competition in Freelancing
The speaker discusses the view that freelance platforms have become more crowded as both clients and freelancers adopt AI, while arguing that this is not a reason to abandon them: a strong portfolio and the ability to use AI for quality work may still be valuable. With explicit uncertainty, he predicts that prompt-focused and surrounding engineering specialties could diminish if AI systems become substantially more capable.
- 1:56:56
Why Public Models May Not Reflect the AI Frontier
The speaker says he is interpreting industry statements cautiously and has not read all of the referenced material. He argues that the public typically receives more tested and constrained systems, while company insiders may see newer capabilities that are not publicly available; their warnings may therefore reflect systems the public has not used. Several referenced names and system details remain unclear in the captions.
- 1:59:22
From an Idea to Executable System Design and Documentation
A concept is developed for a tool that can understand a project idea or accept existing documents, then produce a structured foundation for software developers, such as a PRD, ERD, and roadmap. The speaker argues that strong requirements documents and a robust onboarding phase will matter as development tools become more autonomous; this is presented as a product goal under development, not a demonstrated result.
- 2:01:56
Can AI Create Jobs While Displacing Others?
The speaker argues that AI may displace some work while creating other jobs. He also relays, without identifying a source, an unusual prediction that economic growth and unemployment could rise together; it is therefore presented as a reported forecast rather than an established fact.
- 2:03:02
Automating Workflows with Coding and Communication Tools
A participant described using Codex in Visual Studio alongside Slack workflows, Google Apps Script, chat tools, and a dashboard to develop ideas and automate work, while remaining unsure how to extend the benefit more broadly.
- 2:09:16
Why Is It Hard to Make Large Profits from Dropshipping?
Drawing on his experience, the speaker says dropshipping can be profitable but is harder than commonly portrayed. Thin margins require substantial sales volume, which brings higher operating costs and expenses; he also notes the field’s popularity from 2015 to 2018.
- 2:10:59
The Project Manager’s Role in Initial Discovery
The discussion describes a software-company model in which a project manager receives an idea, explores its scope and prior discussions, and identifies additional questions that may need to be asked of the user.
- 2:13:26
Starting a First AI Agent with a Ready-Made Platform
The advice is that the quickest entry point for building a first AI “employee” is a ready-made agent platform. The tool names are unclear in the captions, so the recommendation is not attributed to a specific product.
- 2:21:00
Building a Custom Automation System for Advanced Users
A personal view that an advanced software engineer may benefit from understanding the system being discussed and building custom automation instead of relying only on a packaged tool, provided they first understand the concept and required resources. The system’s name or underlying technology is unclear in the captions.
- 2:21:47
Testing Your Skill in the Freelance Market
Someone seeking side work is advised to search for their existing skill on freelance platforms, study competing offers and market rates, and gradually build a portfolio. The speaker cautions that progress may be slow until credibility is established on the platform.
- 2:22:36
A Comparative Opinion on Claude and Codex
The speaker considers Claude slightly better in quality and speed and prefers its overall philosophy, while stressing that the difference is very small. In his experience, usage limits affect both tools, and Codex required more retries.
- 2:24:30
Usage-Limit Resets in an OpenAI Challenge
The discussion notes that participants in a 28-day OpenAI challenge received several usage-limit resets early on. The captions do not establish the challenge terms or why the resets occurred.
- 2:27:06
A Workflow Question About Overlapping Content Batches
The discussion raises an unresolved workflow question: what should happen when a batch tied to one stream has not been reviewed and published before another stream arrives the next day? It identifies the case as needing an explicit workflow rule.
- 2:28:13
Health Before Money
A brief statement of the view that health matters more than financial gain.
- 2:33:35
From an Idea to Requirements and System Design with AI
A proposed onboarding stage would turn a non-specialist’s idea into requirements documents, a system design, and a suggested technology stack before implementation. AI can prepare much of this material, but the speaker notes that suitability still requires judgment; giving the system more decision-making authority would demand strong intelligence and could still produce mistakes.
- 2:37:20
Building an MVP for Use During Development
A practical approach is to build a quick MVP that can be used while the application continues to evolve, prioritizing work-management and roadmap functions for now and deferring onboarding. It is presented as a way to avoid delaying progress, not as a claim that the MVP was completed or tested.
- 2:42:21
Can the Context Window Be Configured?
An unresolved question asks whether context-window size can be changed through configuration files or a command. The remarks are tentative recollections, so they do not establish that such control is actually available or reliably identify a specific tool.
- 2:44:00
Trading Off Usage Limits and Weekly Allowance
The speaker discusses a poll result he found surprising: most respondents preferred removing the five-hour usage limit while keeping the current allowance, rather than restoring the limit in exchange for double weekly usage.
Got a question about something in this stream?
Ask in the group