OpenAI, Anthropic reveal how users use AI

PLUS: OpenAI's GPT-5 Codex for upgraded autonomous coding

Good morning, AI enthusiasts. While there is plenty of speculation on how AI is being adopted across the globe, two of AI’s biggest players just provided receipts on how the technology is actually being put to work.

With insights like personal use overtaking professional tasks, massive adoption gaps between rich and poor nations, and users increasingly treating AI like a search engine, the real story of AI adoption is unfolding in ways nobody predicted.

Speaking of adoption… We love hearing how our readers are using AI in their work or daily life. Share your story for a chance to be featured in the newsletter!

In today’s AI rundown:

  • OpenAI and Anthropic reveal how millions use AI

  • OpenAI’s GPT-5 Codex for upgraded autonomous coding

  • Use ChatGPT to improve your thinking patterns

  • Reve revamps creative platform with advanced editing

  • 4 new AI tools, community workflows, and more

LATEST DEVELOPMENTS

OPENAI & ANTHROPIC

Image source: OpenAI

The Rundown: OpenAI and Anthropic both published new data on AI usage patterns across their platforms, revealing demographic shifts, geographic divides, and a growing split between personal and business applications.

The details:

  • Claude users focus heavily on coding, while ChatGPT sees more writing and decision support, with users seeking advice over content creation.

  • Personal use of ChatGPT surged from 53% of messages in June 2024 to 73% by 2025, with non-work conversations growing faster than professional ones.

  • AI adoption in low and middle-income countries is growing 4x faster for ChatGPT, while Claude usage is largely concentrated in wealthy regions.

  • Both platforms show users delegating tasks more frequently over time, with an increase in “information seeking” and search rather than output generation.

Why it matters: AI usage across the globe is rising fast, but these reports offer a fascinating look at how different both adoption and use cases can be across both demographic and regional divides. The personal use surge is particularly interesting, showing that it’s not just work habits being completely altered in the AI age.

TOGETHER WITH AUGMENT CODE

The Rundown: Augment Code's powerful AI coding agent and industry-leading context engine meet professional software developers exactly where they are, delivering production-grade features and deep context into even the largest and gnarliest codebases.

With Augment Code, you can:

  • Index and navigate millions of lines of code

  • Get instant answers about any part of your codebase

  • Automate processes across your entire development stack

Build with the AI agent that gets you, your team, and your codebase.

OPENAI

Image source: OpenAI

The Rundown: OpenAI just introduced GPT-5 Codex, an upgraded, specialized coding model that can dynamically adjust its compute effort based on task complexity — spending seconds on simple fixes or several hours on more complex issues.

The details:

  • The model outperforms GPT-5 on SWE-bench Verified for real-world software use cases, with even larger gains on refactoring tasks at 51.3% versus 33.9%.

  • GPT-5 Codex cuts token usage by 94% for simple tasks, while dedicating 2x the reasoning time to complex problems, with autonomous runs of over 7 hours.

  • Built-in code review capabilities navigate entire codebases, execute tests, and validate dependencies to catch critical software bugs.

  • The update also includes revamped CLI tools, IDE extensions for VS Code and Cursor, and handoffs between local and cloud environments.

Why it matters: Agentic coding tools have completely changed development workflows in just a year — and although Anthropic’s models and Claude Code tool initially dominated the scene, OpenAI’s Codex and models have made for a powerful alternative and competitor despite the company’s broader user base.

AI TRAINING

The Rundown: In this tutorial, you will learn how to break out of circular thinking and generate creative solutions for any business problem by using GPT-5's structured questioning approach to surface blind spots and unlock fresh perspectives.

Step-by-step:

  1. Go to ChatGPT, toggle on GPT-5 in Auto mode, and use this prompt: "I'm trying to [describe your goal or problem], but I'm continuously stuck on the same ideas. Ask enough questions about the problem to find a new approach."

  2. Add context about your current approaches so GPT-5 knows what you're already trying and can push you beyond those solutions

  3. Answer GPT-5's questions across categories like user understanding, product experience, engagement, and analytics to reveal new angles

  4. Review the fresh solutions that emerge from this process and iterate on promising ideas

Pro Tip: Use this workflow for any problem, not just churn. The magic is in how GPT-5 questions you until you hit new thinking pathways for better brainstorming.

PRESENTED BY FIDDLER

The Rundown: The Fiddler Agentic Observability Guide breaks down what’s needed to monitor and control complex, multi-agent systems in production.

Learn how to:

  • Pinpoint issues across the agentic hierarchy

  • Visualize aggregated metrics and granular insights

  • Gain context on agent decision paths and chains of logic

REVE

Image source: Reve

The Rundown: Reve just unveiled a newly revamped image platform that combines AI image generation, natural language editing, and drag-and-drop controls into a single freely available interface.

The details:

  • The platform uses a "layout representation" system that converts images into code-like structures, enabling precise edits while preserving original images.

  • The ‘new Reve’ features a drag-and-drop editor that allows for granular changes to elements within both uploaded and generated images.

  • Reve also adds a chat box to create, blend, and edit images via natural language commands, with the ability to search the web for inspiration.

  • The company released API access in beta, allowing developers to integrate Reve's image creation and editing into third-party applications and workflows.

Why it matters: It’s been just weeks since Google’s Nano Banana changed the image editing game, and now we’ve already had ByteDance’s Seedream 4.0 and Reve launch similar capabilities. Image models have already reached insane quality levels, but advanced editing is the next frontier that opens up completely new use cases.

QUICK HITS

  • ⚙️ GPT-5 Codex - OpenAI’s new upgraded model for agentic coding

  • 🛡️ VaultGemma - Google’s open model to safeguard user data

  • 🎨 Reve - Free AI image generator with new advanced editing features

  • 💻 Holo 1.5 - H Company’s open foundation models for computer use agents

Innovate with AI at AWS Symposium Ottawa, Oct 1. Join public sector leaders, explore emerging tech, and shape the future of AI in Canada. Register to attend for free.*

Google released VaultGemma, the largest publicly available AI model trained with "differential privacy" to protect user data.

OpenAI is aggressively hiring robotics researchers with humanoid expertise, according to a new report from WIRED, signaling a push into physical AI systems.

H Company introduced Holo 1.5, a new family of open weights Computer Use models that achieve SOTA performance across a series of agentic benchmarks.

Microsoft announced the addition of Copilot Chat and agents across its 365 apps (Word, Excel, etc.), incorporating the AI into a sidebar for quick and seamless access.

OpenAI’s chairman, Bret Taylor, said that there’s an AI bubble and “a lot of people will lose a lot of money”, but the tech will still “create huge amounts of economic value.”

*Sponsored Listing

COMMUNITY

Every newsletter, we showcase how a reader is using AI to work smarter, save time, or make life easier.

Today’s workflow comes from reader Mark D. in the United Kingdom:

"Instead of handing my two sports-mad boys their school reports as dry PDFs, I turned them into something they’d actually enjoy. Using ChatGPT, I rewrote the reports as soccer post-match analysis scripts — complete with commentary on their “stats,” highlights, and “areas for training.” Then I ran the scripts through ElevenLabs, giving them the voice of an excitable sports broadcaster.

The result? School feedback delivered like a Match of the Day highlight reel. Strengths became “goals scored,” targets became “training drills,” and report day suddenly felt like match day. For the first time, they asked to hear their reports more than once."

How do you use AI? Tell us here.

That's it for today!

Before you go we’d love to know what you thought of today's newsletter to help us improve The Rundown experience for you.

Login or Subscribe to participate in polls.

See you soon,

Rowan, Joey, Zach, Shubham, and Jennifer — the humans behind The Rundown

Reply

or to participate.