View organization page for OpenAI

9,424,438 followers

Codex brings concepts into view. 🎥: Madeline Zhang

Here we enter in the era of vibe coding. But I use antigravity not codex.

We have codex in our devops pipeline deploying live in prod. We have tried across multiple vibe coding platforms but no one provide such migration friendly non bundled solution. One more thing have stack preference so if codex ain't you can jump in with version control. For Solution Architect and consultation feel free to reach me out🤝

Looks great. I’ve built a few similar prototypes with GenAI as well and can share if you're interested. I don’t want to hijack your thread for self-promo. I’m curious about your approach though. Could you share a link to the solution you showcased in the video?

Codex's visualization is brilliant, but currently confined to the four walls of an office. To make this run as true 'On-Device AI' on the edge, we need a radical shift in Hardware Architecture—specifically in extreme NPU optimization and intelligent power distribution. Without a robust hardware foundation, even the most advanced AI remains a digital mirage. It’s time to build the legs for this brain to walk outside.

This research on 'confessions' is a pivotal step for trust, but we must address the gap between mathematical uncertainty and communicated failure. I recently documented a high-stakes hallucination regarding Swedish law ('framtidsfullmakt'), where the model confidently erred on validity post-death. But the deeper paradox? When I asked ChatGPT to generate an image depicting itself making this mistake, the system refused/failed. 🚫🖼️ We are building models that can theoretically 'confess' in text, yet are visually aligned to hide their faults. True honesty requires the system to be transparent across all modalities. Full breakdown of the legal hallucination here: https://www.linkedin.com/posts/david-e-42b694289_chatgpt-framtidsfullmakt-och-d%C3%B6d-activity-7391445878630338560-2cOe?utm_source=share&utm_medium=member_desktop&rcm=ACoAAEYLzmMBjFYqXKkqZ-T97v45NjbOAjBhxmk #AI #LegalTech #Hallucinations #OpenAI #ChatGPT #Framtidsfullmakt 🤖⚖️"

  • No alternative text description for this image

Love how Codex brings concepts into view. Great visuals, Madeline Zhang! 👏🎥 #Visualization #Codex

Love how Codex turns abstract logic into something you can actually see. When understanding becomes visual, learning stops feeling “impossible” and starts feeling intuitive. Very cool demo.

To understand the severity and consequence of your incompetence ! We saved the engine that was capable of doing this ? but to know the severity of their failure and the consequences ? We were able to decode every single dead language. It wasn't physics for me the same way so there are things I did not save...

See more comments

To view or add a comment, sign in

Explore content categories