28 Insane Things Astra Can Do (2 Hour Course)

Summary

Riley Brown presents an in-depth video course detailing 28 features and workflows possible using OpenAI’s Codex interface paired with the GPT-6 Astra reasoning model. Drawing on his experience running his startup Agent Native, Brown demonstrates how to build web, desktop, and mobile applications, automate communications, manipulate 3D graphics in Blender, run scheduled automations, and manage tasks from both desktop and mobile environments.

---

What is shown

• Setup & Subscription Tiers [01:15–03:20]: Accessing Codex via the ChatGPT desktop application and comparing available tiers ($0 Free, $8 Go, $20 Plus, and the recommended $100/mo Pro tier for full Astra usage).
• Architecture & Layout [03:28]: Demonstrating Codex as the three-panel desktop application interface executing against the GPT-6 Astra model.
• Web App Generation & Production Deployment [06:10]: Live editing of the Agent Native codebase, testing changes locally on localhost:3000, and deploying updates via Git to Vercel.
• GPT Sites [10:37]: Using @Sites to generate and deploy an authenticated Arizona real estate investment calculator site with built-in database storage and company login restrictions.
• Desktop & iOS Applications [18:15]: Prompting Codex to build a native macOS spotlight search tool and creating an iOS ride-sharing clone (“Ruber”) in Swift running inside an integrated iOS simulator.
• Integrated Browser & Automated Testing [24:10]: Using the embedded browser (Command + T) to browse social networks, run automated UI testing across desktop and responsive mobile viewports, and detect rendering bugs.
• Document Creation & Annotation Mode [32:28]: Generating interactive slide decks, spreadsheet models, and reports; using the visual annotation tool [35:08] to pin feedback directly onto slide elements and UI components.
• Full-Stack Site Stack Setup [38:42]: Instructing Codex to configure a persistent technology stack consisting of GitHub, Vercel hosting, and Convex database storage.
• Projects vs. Chats & Multitasking [44:39]: Setting up project folders linked directly to local directories and running background tasks alongside /side scratchpad chats [50:10].
• Blender 3D Modeling [54:28, 63:40]: Instructing Codex to control Blender headlessly, generating custom 3D computer-mouse car models and updating their geometry and materials via image feedback.
• Playable Video Game Development [68:25]: Assembling the Blender mouse models into an interactive, physics-enabled soccer video game (“Click Kick”) hosted on Vercel.
• Automated Scheduling [73:13]: Setting up recurring tasks run at designated days and times (e.g., Monday, Wednesday, Friday at 10:00 AM) to review communication channels and compile briefings.
• Email and Messaging Control [76:21]: Reviewing incoming Gmail threads, generating summarized updates, and sending approval-gated responses via Gmail and Apple Messages.
• Custom Skills & Screen Recording [78:57, 82:56]: Generating reusable skills (e.g., Golden Longform) and using @Record & Replay to record manual UI workflows in Paper to convert them into automated agent skills.
• External API & Video Analysis Integration [84:46, 90:01]: Connecting external API keys (ScrapeCreators, Gemini API) to delegate long-form video frame analysis back to Codex.
• Inline Diagrams [95:52]: Outputting inline Mermaid flowchart and sequence diagrams directly in chat streams.
• Remote Mobile Control [98:04]: Pairing an iPhone with Codex via QR code to monitor projects, launch sub-agents, and trigger voice-directed workflows remotely.
• Image Editing & Sub-Agents [102:49, 105:19]: Editing generated YouTube thumbnail graphics using comment pins and dispatching parallel sub-agents to gather research data simultaneously.
• Direct Computer Control [108:10]: Demonstrating background OS mouse and keyboard automation drawing digital artwork in Canva.

---

Claims & numbers

• The presenter states he has used Codex continuously for the past 6 months to run his startup (Agent Native) and scale social media accounts to over 2 million followers [00:51].
• The presenter notes that OpenAI’s desktop subscription options include Free ($0), Go ($8/mo), Plus ($20/mo), and Pro ($100/mo) [02:20].
• The presenter claims he conducts approximately 90% of his daily work directly inside Codex’s in-app browser [00:27].
• The presenter states that Codex’s GPT Sites feature has an account ceiling of 15 hosted sites [18:02].
• The presenter claims that pre-release Gemini models take approximately 20 seconds to review and output a complete timestamped frame breakdown of a video [91:24].
• The presenter predicts that AI browser automation speeds will surpass human operator speeds within three months [30:22].

---

Notable quotes

• “If I could only use one AI tool and I had to throw out all of the rest, I would choose Codex.” [00:06]
• “Astra is the model, Codex is the app.” [03:28]
• “Over the next six months, this is going to get 10 times better and 10 times faster... You must try this for yourself.” [92:04]

---

Assessment

This video is a comprehensive tutorial and practical feature walkthrough demonstrating the capabilities of OpenAI Codex with GPT-6 Astra. The demonstrations are conducted in real time within working desktop environments, mobile simulators, external web applications, and developer tools, showing genuine user prompts, execution latency, and live generated outputs.

Described by gemini-3.8-flash on 2026-10-09 from the video’s audio and frames.