Video summary

28 Insane Things Astra Can Do (2 Hour Course)

Main summary

Key takeaways

Technology

Summary

Riley Brown presents Codex as a desktop application for using GPT-6 Astra and other models on coding and broader work tasks. He distinguishes the Codex app from the AI model running inside it, and explains that Codex’s capabilities depend on the selected model and the apps or data sources connected through plugins.

The tutorial demonstrates features through example projects. The subtitles end during item 24, so the remaining topics in the video’s advertised list are not covered here.

Product features and demonstrations

  1. Build applications and websites

    • Codex can create and edit web apps, games, desktop software, and iOS apps from natural-language instructions.
    • The presenter demonstrates editing a local website, checking the changes in a browser, then asking Codex to publish the update.
    • He also shows GPT Sites, which he describes as supporting hosting, databases, file storage, login authentication, and custom domains. His example is a private real-estate calculator for his team.
  2. Desktop and iOS app development

    • A desktop utility is built to find images, copy them into a workflow, and remove their backgrounds.
    • An iOS app mock-up is generated as a Swift project and tested in an iOS simulator.
    • The presenter says he plans to cover the App Store publishing process in a separate guide.
  3. Built-in browser and testing

    • Codex includes a browser that can open multiple tabs and, with connected accounts, access services such as Notion and social platforms.
    • The agent can use the browser to inspect sites, edit content, and test applications. Demonstrations include checking calculator behavior, input limits, layouts, and mobile views.
    • The presenter says browser cookies and passwords can be imported from Chrome, allowing the agent to work within logged-in sites.
  4. Documents, spreadsheets, presentations, and annotations

    • Built-in tools and templates can generate documents, spreadsheets, and slide presentations.
    • Annotation mode lets users mark a specific element or area and describe the requested change. Codex can then apply multiple targeted edits, such as changing a slide background or adjusting a website component.
  5. Applications ready for deployment

    • For more production-oriented projects, the presenter demonstrates a workflow using GitHub for source control, Vercel for hosting, and Convex for a database.
    • His advice is to ask Codex for help selecting and configuring a technology stack, creating a repository, deploying the app, and checking security and maintenance practices.
    • He contrasts this with GPT Sites, which he presents as a simpler option for personal or internal team sites.
  6. Chats, projects, multitasking, and sections

    • Recent chats are suited to quick, separate tasks; projects group related chats and files for a larger application.
    • Multiple tasks can run in parallel. A temporary side chat can be used to ask a question while another task continues.
    • Sections organize chats, including research or scheduled-task conversations, without requiring them to belong to a project.
  7. Plugins and connected apps

    • Codex can connect to services such as Gmail, Slack, Notion, Google Docs, and iMessage, as well as creative tools such as Blender and Paper.
    • The presenter demonstrates asking Codex to search across connected sources, summarize information, create a document, or produce draft work.
    • He also shows a workflow in which Codex reviews business communications and proposes tasks the user could delegate or complete.
  8. Blender, 3D assets, and games

    • Codex is shown creating and revising 3D models in Blender, including customized mouse-shaped vehicles.
    • Those assets are then added to a browser-based Rocket League-style game, which is deployed to Vercel and tested.
    • The presenter also shows interactive 3D educational sites, such as models of the human body and a Tesla that users can explore or disassemble.
  9. Scheduled tasks and communications

    • Scheduled tasks can run at specified times, such as drafting useful tasks several times a week. The presenter notes that locally run work depends on the computer being on, while some tasks can be cloud-based.
    • Codex can summarize email, draft replies for approval, and—with permission—send email or iMessages. The demonstration shows the user reviewing and approving a message before it is sent.
  10. Skills, recordings, and external APIs - Skills package repeatable workflows, such as creating video outlines, thumbnails, or social-media research. The presenter demonstrates asking Codex to turn an existing format into a reusable skill. - Record and replay captures a user’s on-screen workflow so Codex can analyze it and turn it into a skill. - Skills can also call external APIs. Examples include social-media data extraction, background removal, and video analysis using Google Gemini. The presenter describes supplying API documentation and a key so Codex can use the service.

  11. Video analysis and visual outputs - A video-understanding workflow sends a video to Gemini for analysis and returns a detailed critique, including transcript structure, key frames, and suggested edits. - Codex can also create inline flowcharts and other diagrams. The presenter prefers these visual explanations to long blocks of text and asks Codex to remember that preference.

  12. Remote control from a phone - The presenter demonstrates connecting a phone to the desktop Codex app for remote control. He says the computer must remain on and shows creating a chat and viewing generated diagrams from the phone.

Guides and tutorials highlighted

  • Downloading and setting up the ChatGPT desktop app for Mac or Windows, then opening Codex.
  • Creating a first hosted site with GPT Sites, including a private team application.
  • Building desktop and iOS applications; a future App Store publishing guide is mentioned.
  • Using the built-in browser to inspect and test apps, including mobile layouts.
  • Creating documents and making precise edits with annotation mode.
  • Setting up a production workflow with GitHub, Vercel, and Convex.
  • Creating Blender assets and incorporating them into a game.
  • Building reusable skills, recording workflows, and connecting external APIs.
  • Using Google Gemini through an API for video analysis.
  • Connecting a phone for remote access to the desktop app.

Main speaker and sources

  • Main speaker: Riley Brown, the video’s presenter.
  • Tools and services demonstrated or discussed: OpenAI Codex and GPT-6 Astra; GPT Sites; GitHub; Vercel; Convex; Blender; Notion; Slack; Gmail; Google Docs; Paper; WhisperFlow; Google Gemini; and external APIs including Scrape Creators and Serp API.

Rate this summary

Your feedback will help improve summaries.

Improve this summary

Reprocess with a stronger model when the summary feels incomplete or inaccurate.

Pro

Translate summary in another language

Pro

Ask questions to this video

Chat for follow-up questions, clarifications, and source-backed answers.

Coming soon

Share this summary

Original video