Back to Warp.dev? 2x Claude + Codex
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 indexed portion of the recording centers on coding-agent orchestration. Luqman describes using Codex to supervise Claude sessions through computer use and scheduled checks, allocate tasks, hand off context, and review work. Audience discussions raise workflow visibility, lost details, and trust in reviews, particularly for scientific code; reliability and token efficiency remain unresolved. He explores a goal-driven workspace combining agent teams, memory, roadmaps, and an interface that adapts to each project. Practical work includes reviewing system architecture, extending plans beyond MVP, and requesting landing-page and hoopoe-animation revisions. The concept and revisions remain in development, alongside discussion of open sourcing; completed implementation and fixes for computer-performance problems are not established. Audience questions cover getting started with AI, programming fundamentals, hardware requirements, security, and developing skills. Other discussions examine task cost versus token pricing and model speed without a measured comparison. On product strategy, Luqman emphasizes lasting usefulness, discovering problems for a focused audience, and planning marketing and distribution. He argues that easier building does not guarantee sales and that marketing outcomes remain uncertain. Secondary topics include agents that read and reply to Facebook and TikTok comments while identifying themselves as AI. Luqman reports workflow problems, local slowness, and agents stopping when the application or computer closes, without documenting fixes. Participants also discuss voluntary creator support through tips and paid messages, affiliate marketing, sponsorships, personal branding, and professional visibility through building in public; the question of how personal branding relates to technical competence remains unresolved. Discussion also covers AI-company IPOs and profit-and-loss disclosure. At recording time, Luqman offers unverified forecasts about IPO timing and funding improving pricing and efficiency, while audience contributions suggest funding could support data centers and research. He also discusses his trust in expert forecasts and a warning attributed to Geoffrey Hinton about superhuman-AI risks, presented as a risk forecast rather than an established outcome.
160 moments from the stream
- 0:55:09
Building backend architecture alongside the website
Luqman explains that the main work concerns the backend and agent harness architecture, while the website is developed alongside it. He mentions setting milestones while acknowledging that planning remains exploratory.
- 0:56:44
Computer-use orchestration and automatic context handoffs
He explains assigning an orchestrator to open and manage implementation sessions for defined requirements or milestones, then receive and review the work with its context. He compares this to automating manual handoff prompts and notes that the operator may lose track of details despite contributing to the design and requirements.
- 0:59:48
An observation on marketing tasks and token usage
Drawing on his experience, Luqman describes marketing tasks as relatively light on token usage in a discussion of how workloads differ.
- 1:00:55
A question about high-resolution AI advertising images
An audience question asks about creating high-resolution advertising images with additional tools. Luqman mentions a tool he considers expensive, but its name is unclear in the captions, and this passage provides no detailed implementation steps.
- 1:01:44
Evaluating tools through financial or practical returns
Luqman argues that spending on tools should produce a return, whether financial or indirect, such as completing work and creating useful outputs.
- 1:04:31
Envisioning a goal-driven agent environment
Luqman describes an early serious exploration of harness engineering and an environment that differs from existing terminals and interfaces such as Warp.dev. He presents goal-driven engineering as an exploratory direction, without claiming a finished replacement.
- 1:05:47
Paid advertising and organic growth
Luqman argues that paid advertising can generate revenue quickly without guaranteeing profit. He explains that organic growth usually requires time, effort, and patience, especially against funded competitors.
- 1:06:07
Stream delay and computer resource pressure
Luqman notices stream delay and reports processor and storage-space issues. These are preliminary observations, with no confirmed diagnosis or documented fix.
- 1:10:55
An agent launching and managing a session
Luqman reports that an agent launched a session in front of him and was managing it. The observation concerns this session and does not establish general system reliability.
- 1:12:28
Highlighting comments with Super Chat
Luqman offers a tentative understanding of Super Chat as a way to highlight a comment in a busy chat, while acknowledging limited familiarity with the feature.
- 1:12:55
Reviewing a regression in hoopoe quality and motion
Luqman criticizes a revision that he believes degraded the hoopoe and scene. He requests a smaller, higher branch and flight to it instead of hopping, then asks for a fix or reversal; the pages do not document a completed correction.
- 1:14:17
Comment-reply agents and local execution limits
Answering a question about Facebook and TikTok comments, Luqman describes agents that read and reply while identifying themselves as AI, and mentions a supporting tool. He reports workflow problems, local application slowness, and agents stopping when the application or computer closes, without documenting fixes.
- 1:17:34
Choosing a project name and comparing domains
Luqman confirms an Arabic project name and compares domain variants, proposing an audience vote. The captions do not establish a reliable Latin spelling or clear verification of domain availability.
- 1:19:55
Tentative release timing and early open sourcing
Asked about availability, Luqman tentatively estimates the end of the following month relative to the recording. He also considers open sourcing the unfinished project and inviting contributors to help develop it.
- 1:22:05
Planning a flying envelope delivery
Requests a character flying in with an envelope in its mouth and then dropping it, with a clearer motion plan before continuing generation.
- 1:22:46
A workspace that proposes an agent team for a goal
Describes a tentative workspace starting from a GitHub repository, local folder, or new project, then studying the goal and proposing an appropriate agent team. Envisions a minimal interface using harness engineering, context and knowledge graphs, and workflow loops, with an intention to enable open-source contributions.
- 1:26:59
Planning token usage across accounts
Proposes combining subscription and API information to estimate weekly capacity and allocate work around goals and available limits. Says operating costs and their relationship to context and memory need further study, and token savings need evidence.
- 1:29:13
Harness modification risks and studying agent memory
Compares a harness to limbs that let a model act, warning that poorly understood modifications can reduce output quality. Speaks favorably of Hermes memory and says he is studying its design, including looking into reverse engineering, without presenting test results.
- 1:30:54
How do you start learning AI?
In response to a broad question, suggests choosing an AI platform and discussing background and goals to find a practical direction. Notes that a precise answer requires more context.
- 1:32:53
Managing Claude sessions with scheduled tasks
Suggests trying Codex to manage Claude sessions through scheduled tasks or cron jobs. Later describes centrally managing four sessions and their prompts, although the coordinator's identity is unclear in the captions.
Got a question about something in this stream?
Ask in the group