26 AI Employees → 9 Stronger Ones
Stream summary 74This stream in minutesWant 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 documents broad practical work and discussion around building, marketing, and measuring AI-assisted products. Its strongest working thread was a Shopify app: reconsidering free-plan versus trial pricing, reviewing onboarding, code, and testing, and developing GA4 analytics that connect visibility, traffic, and orders. Deployment and test outcomes and the pricing impact were not settled, while Luqman viewed acquiring enough traffic as the main remaining challenge. Growth discussions covered inbound and outbound channels, demos, partnerships, advertising, separating traffic problems from conversion problems, demand validation and distribution, and possible later expansion into more languages, platforms, and markets such as Salla. They also included general opinions on bootstrapping, investment, when to raise funding, avoiding excessive early equity dilution, the trust barrier in government sales, and market-dependent YouTube monetization. The stream also explored smaller AI-agent teams with defined roles, dependencies, and human oversight; a permission-limited central executive agent; and voice-assistant tests that exposed latency, interrupted conversations, conflicting records, and incomplete integrations. Luqman compared coding tools and models including Codex, Claude, Gemini, and Grok, arguing that token prices and published benchmarks are insufficient without task-specific efficiency measurements and consistently framing his assessments as personal observations. Codex slowdowns and context degradation remained unresolved. The attempt to locate Gemini 4 was specifically a historical snapshot dated October 1, 2026 and did not establish its availability or final identity. Further discussion covered goal engineering, data and website security and AI’s use by both attackers and defenders, hardware constraints on local models alongside the opinion that local compute may matter long term, practical AI learning while continuing computer-science study, and whether AI weakens thinking or improves efficiency. Other topics included remote work and the possible advantage of lower-cost locations; general distinctions among work or study migration, asylum, and visits; non-personal opinions about financial independence, investment funds, and dollar-cost averaging; and workspace ergonomics through standing desks or walking pads. Luqman also discussed content and livestream strategy, including consistency versus repetition, streaming for accountability, useful standalone videos, and background-music copyright risk. The music work went beyond a general focus-music experiment: it covered tempo, layering, and Arabic maqamat, plus a Suno Studio workflow in which a musical scale is created with Python, uploaded, arranged in layers, and supplemented with sounds. This work and the assistant’s visual and vocal identity remained experiments rather than demonstrated successes.
180 moments from the stream
- 0:01:15
An Idea for Reviewing the First 100 Sessions
Luqman considered making a concise video at session 100 to review the preceding sessions, possibly accompanied by a competition.
- 0:02:35
An Unsuccessful Attempt to Verify Gemini 4 Availability
In a historical snapshot dated October 1, 2026 in the captions, Luqman wanted to try Gemini 4 but could not find it in the tool he checked and remained unsure whether it had actually become available. The attempt does not establish the model’s availability or final identity.
- 0:06:21
Reviewing Code Before Deployment and Designing Product Trials
Luqman said a pricing-structure change had not yet been deployed and that the code first needed review, smoke testing, and polish. He outlined a general distinction between users ready to pay and an option for people who want to evaluate the application, without claiming the change was shipped or successful.
- 0:07:04
Reducing Sarmad from 26 AI Employees to 9 Stronger Ones
Luqman described reducing Sarmad’s AI workforce from 26 employees to 9, saying the smaller group was stronger and had a more focused workflow, especially for outbound work. He presented this organization and improved pricing as steps he expected to help, not as a proven outcome.
- 0:07:52
Using a Demo in Customer Outreach
Luqman discussed offering a demo requested by email and arranged through a meeting, agreeing that it could serve as a useful part of customer outreach and longer-term lead generation.
- 0:08:50
Continued Uncertainty About Gemini 4 Access
Luqman returned to Gemini 4, saying he had thought people were already using it but could not find it in his access, and raised OpenRouter as an unverified possibility. The discussion remained a historical attempt to locate access, not confirmation of availability.
- 0:13:02
Comparing AI Coding Tools by Practical Efficiency
Luqman argued that choosing between Codex and Claude should depend on the efficiency each user actually achieves. He explained that the listed price per million tokens may be insufficient on its own because different token consumption patterns could change practical cost, presenting this as something to measure rather than an established fact.
- 0:15:53
Testing Personal Efficiency Instead of Relying on Price Alone
Luqman acknowledged that he did not know a specific benchmark for token-consumption efficiency and suggested testing one’s own performance with GPT and Claude. He added that, at the time of recording, he personally felt more efficient with GPT, explicitly leaving this as an opinion rather than a benchmark result.
- 0:17:56
Questions for a Bird’s-Eye View of a Project
Luqman formulated a practical request for a high-level project review: what has been completed and built, what remains, what is being merged, how many pull requests have been merged, and whether the code is in production or still unmerged.
- 0:18:37
Prompt Engineering’s Declining Importance Depends on Task and Model
Responding to a claim that one prompt produced the same result from two models, Luqman said prompts still matter depending on the task and model, but in his view mattered less in 2026 than in 2024. He attributed the expectation that prompting may eventually disappear to what others say, rather than presenting it as certain fact.
- 0:22:51
Comparing Coding Models: Efficiency Matters Beyond Price
The speaker argues that leading coding tools are increasingly similar and that practical differences depend on efficiency and token consumption, not price alone. He discusses tentative opinions about Codex, Claude, Gemini, Grok, and other models, asks whether a benchmark measures token efficiency, and prefers testing models himself rather than waiting for benchmark results or relying on hype.
- 0:27:13
From Prompt Engineering to Goal Engineering
The speaker outlines a historical view of how AI systems are directed: from carefully refining prompts, to assigning repeated work in loops, then organizing work as graphs, and eventually giving the system a goal while leaving it to determine the execution. This is presented as his view of the field’s direction, not as an established result.
- 0:31:17
From prompts to goal-driven agents
The discussion describes moving from detailed prompts to giving an agent a clear objective and letting it determine how to accomplish it, while ensuring the objective is understood.
- 0:32:24
When should you set TikTok ad targeting yourself?
The advice is to choose the audience and location manually when an account lacks enough followers or engagement data. With richer signals from followers, likes, and engagement, automated targeting may become more useful.
- 0:33:18
Ask the agent to act or provide clear steps
The workflow asks the agent to perform feasible actions, provide ordered steps when direct execution is impossible, then transcribe and understand a mixed English–Egyptian Arabic recording and add any newly identified angle not already in the MVP to an ideas repository.
- 0:34:32
Transcribing and summarizing long voice messages
A suggestion to use an AI tool to transcribe and quickly summarize a long voice message; the tool name is unclear in the captions.
- 0:35:42
A smaller agent team with clearer roles and focus
The discussion outlines consolidating many agents into a smaller, better-equipped and more focused team, separating inbound from outbound work, assigning specialized outreach roles, and using a management agent as the single coordination interface. The expected improvement is presented as a goal, not a proven result.
- 0:38:35
Chrome features: PDF annotations and vertical tabs
A brief discussion mentions Chrome capabilities for interacting with or annotating PDFs and using vertical tabs, while noting that split view had already existed.
- 0:39:27
Promotion costs vary by country
The discussion notes that TikTok promotion costs vary by geographic market, so the campaign should specify its intended country and audience.
- 0:40:07
Showing Mistakes as Part of the Building Journey
Luqman says that focusing heavily on inbound traffic was not the right fit for him, and explains that the stream presents mistakes and setbacks alongside successes, especially the challenge of finding the right traffic source.
Got a question about something in this stream?
Ask in the group