Skip to content

How this OpenAI engineer uses Codex + ChatGPT Work to automate everything



Nick Baumann is on the Developer Experience team at OpenAI, where he spends his days building with, testing, and …

Tags:

26 thoughts on “How this OpenAI engineer uses Codex + ChatGPT Work to automate everything”

  1. I cannot make it through the constant upspeak. How has this awful habit infested mainstream society? Every statement sounds like a question, the way a child would ask, with their voice trailing upward at the end of each sentence. Unbelievable 🤦🏻

  2. All kinds of awesome sauce all over this. Nice tips, latest ways to use the tech and super nifty dynamic between you both. More like this please. I just got AI to go and order some coffee beans for me – although I really don't want this to just all be about easy shopping – a real digi-assistant is brilliant though and also great for thinking through problems and testing concepts.

  3. One thing is that I'm pretty sad that real time takes up your usage. It would be nice if they would just give it free to the pro users. I think that's a good idea Like you guys are already giving us a free computer that we can use with you know work. Why not just throw in the real time to not take your usage

  4. ✅Shared this informative video in my
    🅰️I upskilling groups ( 3175+ members ) for wider reach

    🎯💎🏆 Great insightful & fruitful video 🏆💎🎯

    लोकः समस्ताः सुखिनो भवन्तु
    ( May all beings lead prosperous life across Globe 🌍 )

  5. Thank you for this episode in particular! This really helped clear up something in the back of my mind.

    On my drive to work this morning, I was starting to wonder whether the $100+ subscription tiers included capabilities I didn’t have, or whether people talking about these more advanced voice workflows were actually using Dictation on mobile rather than the iOS ChatGPT Voice.

    I’ve loved the Voice experience on iOS for years, but I’d been struggling to reconcile what I could do there with talk of the new capabilities. The demos here illustrated the differences between the macOS and iOS experiences finally making it click for me.

    It’s a little unfortunate that the "pure" Voice experience is currently so different between macOS and iOS, but I’m excited to see where it goes. Hopefully we eventually get to the point where Voice on mobile can reach into remotely connected devices (or Work on the web) and offer feature parity.

    Thanks again to you both for such a helpful vid!

  6. Puntos destacados by TubeLens:
    03:49 La nueva función de voz de Codex/ChatGPT permite activar un orbe mediante hotkey desde cualquier parte del sistema operativo. El asistente puede ver la pantalla, tomar 'appshots' (capturas contextuales con Command+Command) y gestionar simultáneamente cinco o seis hilos de tareas en paralelo.
    07:01 Nick demuestra en vivo cómo delegar por voz tareas logísticas: revisar el calendario, buscar vuelos desde SFO para un offsite en París, y gestionar un reporte de gastos buscando recibos de Amazon, todo sin tocar el teclado y usando el navegador Chrome existente.
    05:52 ChatGPT Work en móvil funciona como un agente siempre activo que monitorea Gmail, finanzas y paquetes, enviando notificaciones proactivas. Nick lo describe como una experiencia mágica comparable a tener un asistente personal disponible todo el día.
    17:29 Se construye en vivo un sitio web en chatgpt.site para el canal 'How I AI', con branding personalizado (morado y negro), categorización por herramienta y función, base de datos SQL, almacenamiento S3 y control de acceso por email. El sitio queda funcional en minutos.
    21:33 Nick usa Codex con un plugin personalizado (UGC Video) para editar videos de forma automatizada: sube decenas de clips, Codex transcribe, analiza frames, selecciona las mejores tomas según indicaciones verbales del creador y exporta en formatos 9×16 y 4×5. Se recomienda usar Claude Ultra por la capacidad de procesamiento paralelo.
    13:29 En el debate sobre latencia vs. inteligencia en experiencias de voz, Nick concluye que prefiere latencia, especialmente cuando el sistema tiene buenas herramientas para delegar. Claire predice que la latencia será el principal diferenciador competitivo en la segunda mitad del año.
    36:55 Nick reflexiona sobre el momento 'AGI' que supone ver a un agente de codificación leer archivos, crear archivos y ejecutar comandos de forma autónoma. Considera que ChatGPT Work lleva estas capacidades al gran público a través del móvil, con plugins de fácil instalación.

  7. For enterprise rollouts of OpenRouter vs direct Anthropic API: when is the markup worth it?, start with permissions and data boundaries, not feature demos. I shared a checklist on my profile.

Comments are closed.