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
- 0:08:56
Coordinating Claude sessions and control limits in Warp
Luqman describes a coordinating agent managing parallel Claude sessions through browser and desktop interfaces. He reports that Warp security restrictions hinder automated control and likes the workflow while acknowledging that the coordinator can lose track.
- 0:15:16
Small-tool longevity when platforms absorb their features
Luqman argues that some GitHub tools fade when major platforms absorb their functionality, even when the original idea is useful. He suggests complementary services as a possible path to longevity.
- 0:16:49
Requesting a landing-page draft with AI-generated assets
Luqman requests an initial landing-page draft using AI alone, accepting some compromises to obtain a concrete version he can review and refine.
- 0:18:21
From IDEs to ADEs and goal-driven work
Luqman presents his understanding of the shift from traditional IDEs to agentic development environments and terminals. He explores an additional kind of goal-driven work environment while expecting ADEs to remain relevant long term; the proposal is still exploratory.
- 0:21:41
Testing lasting usefulness before pursuing revenue
Luqman asks whether a new tool would provide lasting usefulness or be abandoned after a brief trial. He argues for delivering substantial user value before pursuing revenue.
- 0:25:40
Multiple agent sessions and usage tracking
Discusses coordinating Claude and Codex sessions, suggesting separate Warp environments as a possible way to manage multiple accounts. Proposed features include weekly usage tracking and approximate token estimates by plan; implementation is not established.
- 0:27:21
Requesting a resource audit during a stream
Asks an agent to inspect disk usage and processing load after stream lag raises concern about an overloaded computer. This is a diagnostic request, without a confirmed cause or demonstrated fix.
- 0:28:55
From IDEs to agent coordination environments
Explains a shift from IDEs toward agent-centered environments, describing a tool that brings Claude and Codex together and allows agent selection, with comparisons to Warp.dev. The speaker favors it for relatively hands-on work and describes it as open source and apparently free, while acknowledging uncertainty about internal charges.
- 0:31:35
Reducing computer load while preserving coding tools
Discusses reducing load during a coding stream: checking whether Docker Desktop is in use before requesting its closure, preserving coding processes, and reducing other activity. Cache files are also discussed, alongside the distinction between disk space and processing load. No successful cleanup or performance improvement is established.
- 0:37:29
How should beginners build continuously working agents?
Answering an audience question, the speaker recommends reading about graph engineering, preparing an environment for agents, and starting with one or two agents handling a single simple, recurring task. Token usage should be monitored as the system develops gradually.
- 0:39:25
Agent management of Chrome sessions
Describes an agent managing several Chrome sessions through direct interaction.
- 0:40:27
Handling high CPU load
Reports closing applications and Docker to address CPU load and mentions needing to restart the Mac. The captions do not confirm a restart or improved performance.
- 0:41:33
Questions about remaining tasks and parallel execution
Asks how many missions and tasks remain in the second milestone, whether they can run in parallel, and how to organize them. No answer appears on this page.
- 0:42:17
Planning customer acquisition and monetization early
Answering a general app-development question, Luqman argues that reaching customers is harder than coming up with the idea or building it. He recommends clarifying monetization and customer acquisition early to save time and effort, while acknowledging that he does not know the app's details.
- 0:43:05
Background music question and a proposed playlist
Cannot identify the track and suggests searching for qanun or Arabic oud music to find something similar. He proposes creating a playlist command for the recurring question and clarifies that the Dune ambient music mentioned is different.
- 0:46:31
An orchestrator for coding tasks and reviews
Luqman describes a recent workflow in which an orchestrator manages independent sessions, product requirements, task allocation and reviews, involving additional agents when complexity increases.
- 0:48:24
Revising the hoopoe scene and animation loop
He discusses grass and framing problems and requests removing the grass and adding a branch on the right, where the hoopoe occasionally hops before returning to its usual loop. He asks for the requirements to be restated before implementation; this passage does not establish completion.
- 0:50:59
Discussing «سرب» as a proposed name
Luqman likes the proposed Arabic name «سرب» because it describes the idea, while leaving room for the audience's choice. The discussion expresses a preference without confirming adoption.
- 0:52:39
Hoopoe character feedback and prototype priorities
Audience feedback addresses crest proportions, surroundings and colors. Luqman explains that the current character is a temporary cartoon prototype; commissioning a 3D artist would depend on the idea developing and being used, while visual refinement is a lower priority for now.
- 0:54:26
Claude–Codex communication and durable delegation
Luqman says Claude can communicate with Codex through command-line tools or terminals that are not visible to the user. He connects this to a lesson from an earlier delegation mechanism ending: choose approaches that can remain useful longer.
Got a question about something in this stream?
Ask in the group