Codex Super App, OpenAI Chaos Drama, Gemini 3.2 Pro In Arena, GPT-Realtime-2, & NotebookLM Update!

· Source: WorldofAI · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Robotics & Autonomous Systems · Depth: Intermediate, long

Summary

This week in AI saw significant developments across major players, with OpenAI teasing a "Codex super app" that unifies ChatGPT, browsing, automations, and voice agents into a single desktop experience, alongside launching GPT Real-time 2, an intelligent voice model offering GPT-5 level reasoning for live interactions. Google's Gemini is undergoing aggressive experimentation in L Arena, with some users reporting "nerfed" performance in new variants, while also introducing Gemini Notebooks for AI-native workspace organization. Anthropic's Claude Code gained financial dataset connectors, allowing direct access to market data for advanced prompting. Baidu released Ernie 5.1, achieving frontier performance at 6% of typical pre-training costs, and xAI continues to evolve Grok into a super app with expanded connector support. Additionally, leaked private text messages surfaced regarding the 2023 OpenAI board coup, and discussions around an "anti-clanker movement" against embodied AI gained traction.

Key takeaway

For Machine Learning Engineers developing new AI applications, the trend towards integrated "super apps" and efficient, real-time models means focusing on building custom tools and integrations around core AI capabilities. You should prioritize exploring how to leverage new voice AI models for collaborative interfaces and consider the cost-efficiency of models like Ernie 5.1 for deployment, rather than solely chasing model size.

Key insights

Major AI labs are converging on "super app" strategies, integrating diverse functionalities into unified AI operating systems.

Principles

Method

OpenAI's "slash goal" feature allows Codex to iterate on tasks until a defined end state is reached, enabling long-running coding and reasoning tasks without constant human intervention.

In practice

Topics

Best for: Machine Learning Engineer, NLP Engineer, CTO, AI Engineer, AI Scientist, Director of AI/ML

Related on AIssential

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by WorldofAI.