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
- 5:24:00
Ideas, execution, and reaching an audience
The speaker argues that ideas alone are insufficient and that, as execution becomes easier, reaching an audience and being able to market and sell to it become important factors.
- 5:29:54
An automatic draft for requesting mission progress
The speaker describes a draft that anticipates a request for a mission update and a simple completion percentage, saying it can be sent with Tab. He presents it as an example of reducing detailed manual prompting; these captions do not provide the progress answer or verify its accuracy.
- 5:32:08
Voluntary support platforms for creators
An unclearly named platform is described as allowing creators to receive tips and paid messages, and is compared with Patreon. Participants discuss voluntary support as appreciation for content and suggest placing its link in the description or alongside social links.
- 5:37:37
Coordinating and periodically reviewing Claude sessions with Codex
Luqman describes using Codex to prompt Claude sessions and review work and code through a scheduled job every fifteen minutes. He sees his role shifting toward preparing system designs, requirements documents, and goals, then monitoring the agents. This is his account of the workflow, rather than independent verification of its results.
- 5:40:14
Comparing Claude and GPT costs: token prices and usage per task
After describing Codex access as paid at the time of recording, Luqman discusses a question about choosing a Claude or GPT subscription on a limited budget. He uses API token prices and output-token usage per task to explain cost, concluding that Claude is more expensive in the comparison presented. This segment does not settle the subscription choice; unclear model names in the captions are omitted.
- 5:42:16
Codex and Claude: consumption and usage-limit resets
At the time of the discussion, the speaker favors Codex for more frequent usage-limit resets, while acknowledging that it can sometimes consume more. This is a qualified personal impression, not a controlled efficiency comparison.
- 5:42:50
Speculation about an Anthropic IPO and its service impact
The speaker speculates that IPO funding could improve efficiency, cites mid-November 2026 as an expected date, and anticipates access to profit-and-loss information. The passage does not establish the IPO timing or its effects.
- 5:48:14
What is a token? An initial explanation and OpenAI Tokenizer
In response to a question, the speaker tentatively describes tokens as units of model usage and distinguishes input from output. The discussion then suggests trying text in OpenAI Tokenizer and reads a token-and-character count, without establishing a fixed word-to-token ratio.
- 5:49:56
A checkpoint for resuming agent work
The speaker requests a safe checkpoint so work can resume after system-design and product-requirements discussions have expanded. This passage proposes the pause without detailing a persistence mechanism.
- 5:51:03
A workspace that adapts its tools and interface to each project
A proposed workspace starts with baseline MCP tools and skills, then adapts its capabilities and interface sections to project needs, such as adding analytics for tracking. The speaker presents a project-shaped interface as a design direction, contrasting it with conventional workspaces that feel relatively fixed.
- 5:54:16
Slowing down amid AI fear of missing out
Advice to slow down amid rapid AI developments, with reassurance that keeping up with every release is not necessary to avoid missing an opportunity.
- 5:59:01
Planning a large task and supervising a visible Claude session
Luqman describes preparing requirements, a roadmap and system design, then having an agent open and supervise a visible Claude session through computer control. He suggests resolving human dependencies upfront to pursue autonomy, with code feedback and review as possible additions.
- 6:05:14
A five-round debate experiment between agents
Luqman proposes opening a visible Claude session for a short five-round debate between agents to observe their interaction and computer control, recalling a similar experiment on an earlier stream. He reports fast responses in his setup but remains uncertain about why others experience different performance.
- 6:06:53
Visible delegation versus a background delegation skill
Asked about a delegation skill, Luqman prefers seeing the exchanges between agents rather than having them happen in the background, while acknowledging limited experience with the skill. The discussion raises a token-use objection, and he emphasizes communicating the intended vision clearly; his favorable assessment is personal opinion.
- 6:10:29
A recommendation to use Codex for the delegation workflow
Responding to a question about whether using GPT chat instead of Codex explains different behavior, Luqman recommends Codex for this workflow and GPT chat for personal uses. This is his recommendation; these pages provide no documented comparison.
- 6:11:15
Automated handoffs and visibility into agent exchanges
Luqman likens agent coordination to automated handoff prompts that he considers more precise and timely. The audience discusses feeding an agent’s outputs back to an orchestrator, while allowing that the skill may have been misunderstood. He prefers visible exchanges but admits that planning documents give him direction without understanding every prompt; his satisfaction with the results is subjective.
- 6:11:47
Opinions on changes to Claude’s session interface
Luqman discusses Claude interface changes: he likes the new session rail but dislikes the left-side session arrangement, allowing that he may need time to adjust. A participant mentions an Overview section for each project, but he does not confirm finding it and notes that he rarely uses projects.
- 6:14:23
Monitoring agents and qualifying judgments of work quality
Luqman describes tracking active agents and open sessions through a monitoring dashboard. He sees rapid progress and considers the work promising, while explicitly saying he has not yet verified code structure or bugs.
- 6:15:40
An opinion on agent delegation and Codex computer use
Luqman recommends trying delegated management of other agents, including through their web apps. He describes Codex computer use as more human-like and better than Claude in his opinion; the passage offers advice and impressions without documenting a comparative test.
- 6:17:19
Requesting a session pause with a checkpoint
He requests a pause and a checkpoint, asking that the session, which appears already paused, not be restarted. The captions document the request without confirming its execution.
Got a question about something in this stream?
Ask in the group