Using Codex vs Gemini for development

Episode 600 opened with Beth hosting solo before Andy and then Carl joined. They reflected on the show’s long run and joked about the chaotic start due to technical issues and multiple versions of the studio running at once. Beth highlighted how Gemini 3’s image creation, especially “Nano Banana Pro,” is producing highly accurate layouts with readable text. The group discussed how far multimodal models have evolved and how different tools now specialize in different strengths. The rest of the episode covered AI agents, Codex Max, Gemini prompting, SEO disruption, group chats in ChatGPT, and how users are shifting their habits across platforms.

Key Points Discussed

Gemini 3’s “Nano Banana Pro” creates accurate layouts and readable text

Problem solving around Talk Studio bugs during the live show

Gen Spark hits a $1.25B valuation and expands workplace agent automation

TikTok adds controls for AI generated content and labels deepfake material

Users increasingly search how to delete or deactivate social platforms

Adobe buys SEMrush, triggering worries about the future of SEO tools

SEOs struggle because AI search results are personalized, inconsistent, and agent driven

Codex Max improves complex backend builds, Gemini excels at front end and multimodal work

New ChatGPT group chats allow shared sessions across teams and free users

PRDs become essential for scoping apps before coding with agents

Advice on using Cursor, Codex, Gemini CLI, Cloud Code, and avoiding multi tool conflicts

OpenAI launches free ChatGPT access for verified K 12 educators

Timestamps and Topics

00:00:00 🎉 Opening, episode 600, first solo start

00:02:44 👋 Andy joins, discussion on Nano Banana Pro image accuracy

00:04:50 🖼️ Gemini layout and multimodal strengths

00:07:00 💻 Gemini Pro for image generation and model selection

00:09:34 🤖 Gen Spark’s $275M round and workplace agent capabilities

00:12:03 🛠️ How Gen Spark automates complex workplace tasks

00:14:24 🧩 Agent platforms vs built in agents in big model ecosystems

00:15:21 🧭 How users may lean on ChatGPT for end to end work

00:16:58 🔀 Technical chaos navigating multiple Talk Studio instances

00:19:40 🗞️ TikTok labeling AI content and user decline across platforms

00:21:58 👥 New ChatGPT group chats demo and quirks

00:28:36 📝 OpenAI gives teachers free ChatGPT with integrations

00:32:32 🔍 SEO disruption as AI search becomes personalized and inconsistent

00:34:26 📉 Adobe buys SEMrush, concerns about tool decline

00:38:40 🤖 AI agents change how users perform search and comparison

00:40:58 🎯 Codex Max vs Gemini 3 for coding, strengths differ by task

00:45:49 🧪 Why building simple test apps matters before real projects

00:50:16 🔧 Using Gemini for front end and Codex for complex backend logic

00:52:41 🧠 Avoiding tool conflicts when coding across multiple IDEs

00:56:03 🛠️ Cursor recommended as the unified working environment

01:02:44 📂 Importance of GitHub when switching across platforms

01:06:59 🏁 Closing, weekend content reminders