Building Intelligence + Manifesto
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 stream centers on hands-on review of Burnish before V1: auditing the first-time-user experience, documenting bugs in product-readiness states, publishing, links, and image alt text, and testing autopilot batches, timing, loading, and pause behavior. It also traces failures in manifesto generation and PDF downloading and a missing FAQ presentation, using a workflow of visual verification, issue logging, root-cause research, separate implementation handoffs, and incremental merging and deployment. The work remained unfinished: autopilot, the Intelligence page, and downstream manifesto integration still required testing and polish, while several bug causes were unresolved. The discussion also develops an exploratory Telegram-based virtual office for collaboration and project progress, including prototypes, character movement, and collisions, although its lasting purpose, implementation scope, and privacy model remained open questions. Audience exchanges cover launching mobile apps, choosing B2B or B2C and web or mobile, competitor research, unit economics, and marketing, alongside qualified opinions about agentic coding tools, AI models, infrastructure, and token pricing. The speaker’s recurring view is that workflow fit matters more than broad model claims and that building a product is only the start of market testing, marketing, and sustained execution.
82 moments from the stream
- 2:20:26
Early Work and the Challenge of Routine Consistency
The speaker says that starting work early can make the day feel longer and more productive, while emphasizing that consistency is the harder part. Late-night activities are cited as something that can disrupt sleep and make an early start the next day difficult.
- 2:20:56
Building a Specialist Presence on X for Arabic Audiences
Discusses X monetization eligibility as understood at the time of recording, then suggests that the relatively limited presence of specialist Arabic content on the platform may create a growth opportunity. The practical idea is to adapt useful contributions from a Telegram community into regular X posts, particularly around AI-assisted coding.
- 2:26:05
Reviewing Autonomy Across System Tiers
A technical discussion questions whether “presence” stops at level two while “authority” reaches level three, then proposes logging the issue and verifying that autopilot works across all tiers rather than only one path.
- 2:33:16
From IDEs to Agentic Coding Tools
A personal account of moving from VS Code to Antigravity soon after what the speaker believes was its late-2025 release, then moving around early or mid-January to Warp and using it until April. The speaker argues that Antigravity and Cursor later shifted from traditional IDEs toward Codex-like agent interfaces, and says Warp’s CLI experience had begun to feel constraining.
- 2:44:26
Cleaning Up and Polishing Level-Two and Level-Three Behavior
A brief closing note says the remaining work is cleanup and polishing around levels two and three; the terse captions do not establish implementation details or a final outcome.
- 2:45:21
Cleaning Up Pull Requests and Preparing a Multi-Agent Handover
After verifying that the intended pull request was merged and deployed, the remaining requests were to be cleaned up and the discovered bugs assembled into a handover prompt for a fresh session, structured so an orchestrator could distribute the work across multiple sub-agents.
- 2:55:47
A Debugging Workflow Before Starting an Implementation Session
The debugging phase was described as recording issues, verifying them, tracing their sources, and then writing an implementation prompt. When one issue was said to have fixed itself, it was checked first; the FAQ section was still missing, so its repair was made a prerequisite for the next implementation session.
- 2:59:26
Diagnosing an Optimized FAQ Block That Is Not Rendering
Visual mockups were requested so the result could be inspected directly. The diagnosis clarified that an FAQ was technically present on the product page and the app was enabled in the theme, but the previously created optimized block was not rendering, prompting a deeper search for the merged work and the cause of its absence.
- 3:06:20
A Collaborative Virtual Office Inside Telegram
An early concept for a 2D virtual office inside a Telegram group, where participants have movable avatars, can see what others are working on, and communicate. The idea remains exploratory, with its scope and implementation difficulty still undecided.
- 3:09:11
Guiding a Prototype with a Visual Reference
Implementation feedback asks for close study of a reference image and a more colorful 2D office with work desks and a possible lounge. The passage illustrates refining a build by comparing its output with a clear visual standard while remaining mindful of exploration cost.
- 3:13:20
Checking Fix Durability and Handing Off a Work Session
After a change works, the speaker follows up by asking whether the fix is permanent or the issue may recur, then indicates readiness to hand off a previously prepared coding session. The exact technical context is not visible in the captions.
- 3:14:56
Exploring Collaborative App Development
The speaker considers experimenting with building the small Telegram-related app during the stream and possibly opening a shared repository if the idea is well received. The implementation and collaboration model remain undecided.
- 3:16:58
Opinions on Model Restrictions and Jailbreak Handling
The speaker speculated that a returning system might face tighter restrictions and regulation without necessarily becoming less capable, and argued that the main need was to fix the jailbreak path. He also discussed, with explicit uncertainty, claims about jailbreak reporting and specialized teams or bounty-style efforts to discover such issues.
- 3:22:42
Handing Off a Visual QA Workflow to a New Session
The speaker requested a handoff prompt for a separate session that would continue an established visual walkthrough QA workflow, and asked whether the current coding session should finish before the app is reset.
- 3:33:14
Auditing the Page for a First-Time User
The page is to be inspected end to end to distinguish working connections from UI-only elements, evaluate the first-time-user flow, and concisely identify what is missing, with the backend described as weak and substantial testing still expected.
- 3:37:45
Clarifying the AI Article Topics Workflow
When a user lands on the page, it should immediately communicate that Burnish is generating AI article topics. The work also calls for recovering the previously discussed article and autopilot workflow, then explaining the topics tab in plain language and defining its behavior and topic count.
- 3:44:53
Prioritizing the Intelligence Page
The priority shifts from articles and visibility to completing the Intelligence page first because it is considered foundational to the rest of the application. The proposed next step is to investigate how the page is built and wired, then capture requirements and prepare a prompt for subsequent work.
- 3:58:39
The Manifesto’s Role and Verifying Generation Readiness
The speaker explains that manifesto generation is available across tiers and recommended for users because it is intended to guide downstream parts of the application, including the storefront, writing, visibility, prompts, article topics and language, and market scouting. Although the implementation appears locked in, he has not yet inspected the generated manifesto, so he wants to generate it and examine the PDF, page count, and completeness before considering downstream backend wiring.
- 4:02:21
How Much Can a Generated Design Be Customized?
A question was raised about fitting the design to a mobile screen and whether multiple rooms or layout improvements could be added after download. The page does not provide a definitive answer about those customization limits.
- 4:08:23
Checking the Local Worker When Manifesto Generation Breaks
Unwanted jobs were cleared and a local worker was started for generation. When the Manifesto generation popup appeared broken, the speaker asked whether the local worker might be involved; these units establish neither a confirmed cause nor a fix.
Got a question about something in this stream?
Ask in the group