¡OpenAI acaba de reventar la IA con Codex 2.0! Y esto es lo que puedes hacer 🔥
The video showcases OpenAI's Codex 2.0 desktop application through seven practical use cases, demonstrating its ability to organize files, generate Excel spreadsheets, create animations, control third-party tools like Blender, schedule automated tasks, build full applications, and deploy them live. The presenter argues Codex 2.0 offers superior value over Anthropic's Claude Code due to better credit limits, free plan availability, and broader capabilities. The workflow integrates multiple AI models, plugins, and parallel agents within a single interface.
Summary
The video opens by framing Codex 2.0 as OpenAI's response to competitive pressure from Anthropic's Claude Code, which the presenter criticizes for shrinking credit limits, frequent outages, and high pricing even on pro plans. The presenter notes they have been using Codex since its quiet release about three months prior and have largely abandoned Claude Code as a result.
The first use case demonstrates Codex's ability to take control of a local folder, analyze image and video content directly, and rename files based on their actual content rather than metadata alone — achieving accurate descriptive filenames in under a minute without manual instruction on each file.
The second use case shows Codex generating a detailed multi-tab Excel spreadsheet for a hypothetical bakery business in Cádiz, including market analysis, financial projections, break-even points, EBITDA scenarios, and cash flow forecasts — all grounded in real internet-sourced data and visualized with graphs, produced autonomously in a few minutes.
The third use case covers animation creation using the Remotion plugin, where Codex writes code to generate a professional comparison video between Claude Code and Codex. The presenter demonstrates running multiple parallel agent chats simultaneously — one generating logos and another improving animation quality — both working on the same project concurrently.
The fourth use case involves Blender, a complex 3D modeling tool the presenter has never used. Because no official Blender plugin exists in Codex, the presenter has Codex create its own MCP connection and skill from a GitHub repository, then uses it to generate and colorize a 3D car model directly inside Blender — all without the presenter understanding Blender's interface.
The fifth use case explores Codex's Automations feature, where the presenter schedules a daily 10 AM task to scan AI news, identify key stories for creators, companies, and professionals, write a report, and generate three LinkedIn-ready infographic images — all stored locally for privacy and running autonomously every day.
The sixth use case involves building a complex personal dashboard application called 'Live OS,' integrating finances, habits, calendar, energy levels, and household management into a single interface with 3D room models and AI agents. The presenter discusses the tradeoff between model power levels (low, medium, high, extremely high) and credit consumption.
The final use case demonstrates one-click deployment of the locally built application to a live public URL using the Vercel plugin, completing the full build-to-publish workflow within Codex without leaving the platform. The presenter concludes by noting that Codex's integration of tools, agents, and deployment significantly lowers technical barriers for completing complex, multi-step projects.
Key Insights
- The presenter argues that Claude Code's shrinking credit limits are so severe that even on a pro plan, just a couple of uses can cause the model to drop to medium power mode and exhaust all credits, making it practically unusable for intensive work.
- Codex can create its own MCP plugin and skill from a GitHub repository to connect with and control tools like Blender that have no official Codex integration, effectively teaching itself how to use an entirely new application.
- The presenter demonstrates that Codex supports parallel agent chats working on the same project simultaneously — one chat generating logos while another improves animation quality — a capability the presenter highlights as a key differentiator.
- Codex's Automations feature can schedule recurring AI tasks locally, storing all generated files on the user's machine rather than OpenAI's servers, which the presenter frames as a meaningful privacy advantage over the standard ChatGPT platform.
- After intensive use across all seven demonstrated use cases, the presenter reports consuming roughly 70% of their 5-hour credit allowance on the $20/month Plus plan, claiming this represents significantly better value than Claude Code's equivalent tier.
Topics
Full transcript available for MurmurCast members
Sign Up to Access