Good morning, AI enthusiasts. OpenAI just released o3 and o4-mini, two new reasoning models that president Greg Brockman called a “GPT-4 level qualitative step into the future.”

With o3 pushing SOTA across all benchmarks and supposedly capable of creating new scientific ideas, is this the leap that finally puts the AI world at AGI’s doorstep?

In today’s AI rundown:

• OpenAI’s o3 and o4-mini, new coding agent

• Copilot gets hands-on computer use

• How to run AI privately on your own computer

• Claude gains autonomous research powers

• 4 new AI tools & 4 job opportunities

LATEST DEVELOPMENTS

OPENAI

🤖 OpenAI releases o3 and o4-mini, new coding agent

Image source: OpenAI

The Rundown: OpenAI just released o3 and o4-mini, its smartest reasoning models yet that are now equipped with full agentic access to all ChatGPT tools and the ability to "think with images” — alongside the launch of a new open-source coding agent.

The details:

• OpenAI o3 is the new top-tier reasoner, pushing SOTA performance across coding, math, science, and multimodal benchmarks.

• o4-mini offers fast, cost-efficient reasoning, significantly outperforming previous mini models and even saturating benchmarks like AIME 2025 math.

• Both models can use and combine all tools within ChatGPT (web search, Python, image generation, etc.) as part of their problem-solving process.

• The models are also the first to be able to "think with images", integrating visual analysis and manipulation directly into their chain of thought.

• Also launching is Codex CLI, an open-source coding agent that runs in users’ terminals and links reasoning models with coding tasks.

• President Greg Brockman said the release is a “GPT-4 level qualitative step into the future,” with the models capable of producing novel scientific ideas.

Why it matters: Whatever the bar for AGI is, it feels like the latest SOTA models are getting close. While reasoners were already a massive leap, equipping them with access to tools and multimodal capabilities has led to a class of models that is creating new ideas — seemingly taking us to Step 4 of OpenAI’s ladder of AI intelligence.

TOGETHER WITH AUGMENT CODE

🤖 The AI agent for professional devs

The Rundown: Augment Code's powerful AI coding agent meets professional software developers exactly where they are, delivering unmatched productivity and top-ranking performance.

With Augment Agent, you’ll experience:

• The No.1 ranked open-source agent on SWE-Bench by combining Claude 3.7 and o1

• Easy integration into Vim, JetBrains, and VS Code environments

• Compatibility with 100+ native and MCP tools

Sign up and start building with Augment Agent.

MICROSOFT

🖥️ Copilot gets hands-on computer use

Image source: Microsoft

The Rundown: Microsoft just rolled out a new 'computer use' capability in Copilot Studio, enabling users and businesses to build AI agents that can directly operate websites and desktop applications.

The details:

• The new feature allows agents to interact with graphical user interfaces (GUIs) by clicking buttons, selecting menus, and typing into fields.

• The process unlocks automation for tasks on systems lacking dedicated APIs, allowing agents to use apps just like humans would.

• Computer Use also adapts in real-time to interface changes using built-in reasoning, automatically fixing issues to keep flows from breaking.

• All processing happens on Microsoft-hosted infrastructure, with enterprise data explicitly excluded from model training.

Why it matters: Copilot joins the likes of OpenAI and Anthropic’s Computer Use tools, marking another step in AI’s agentic shift from chat windows into everyday software. While it’s not the only UI automation tool, Microsoft users’ existing business workflows are a perfect use case to take advantage of this type of feature.

AI TRAINING

🤖 How to run AI privately on your own computer

The Rundown: In this tutorial, you will learn how to run powerful AI models directly on your own computer for complete privacy, zero cost, and offline use—without sending data to external servers.

Step-by-step:

1. Choose your platform by downloading Ollama or LM Studio based on your command-line or GUI interface preference.

2. Install the software and open it (both options are available for Windows, Mac, and Linux).

3. Download an AI model that's suitable for your computer

4. Start chatting with your AI using terminal commands in Ollama or the chat interface in LM Studio.

Pro tip: Match the model size to your computer's capabilities; newer computers might be able to handle larger models (12-14B), while older ones should stick with smaller models (7B or less).

SPONSORED BY SALESFORCE

⏰ The countdown is on for Dreamforce 2025

The Rundown: Dreamforce 2025 — the world’s largest, most trusted AI event — returns on Oct. 14 with three days of next-level innovation and real-world inspiration you can bring back to your business.

Here’s what you’ll experience:

• AI agent building alongside product experts with Salesforce’s Agentforce

• 50+ visionary and product keynotes

• 1,200+ breakout sessions across every product, role, and industry

• 150+ hands-on trainings and 240+ peer roundtables

Join the list and get early access.

ANTHROPIC

🔍 Claude gains autonomous research powers

Image source: Anthropic

The Rundown: Anthropic just unveiled major upgrades to Claude, introducing autonomous research capabilities and Google Workspace integration to allow the assistant to search both the web and user files for answers with better context.

The details:

• The new Research feature can autonomously perform searches across the web and users’ connected work data, providing comprehensive, cited answers.

• A new Google Workspace integration lets Claude securely access user emails, calendars, and docs for context-aware assistance without manual uploads.

• Enterprise customers also get access to enhanced document cataloging, using RAG to search entire document repositories and lengthy files.

• Research is launching in beta for Max, Team, and Enterprise plans across the US, Japan, and Brazil, with Workspace integration available to all paid users.

Why it matters: Anthropic continues to move at its own pace when it comes to feature rollouts, giving Claude a “Deep Research” type feature well after the other major labs. But as we’ve seen with other rivals, the combination of web search, user data integration, and SOTA models can lead to some extremely powerful results.

QUICK HITS

🛠️ Trending AI Tools

• 📽️ Veo 2 - Google’s SOTA video model, now available in Gemini App

• 🎥 KLING 2.0 Master - New video AI with improved prompt adherence

• ⚙️ Grok Studio - Canvas-like interface to collaborate with AI on docs and more

• 🔎 Embed 4 - Cohere’s new multimodal search model for enterprises

💼 AI Job Opportunities

• 🤖 xAI - AI Engineer & Researcher

• 🛠️ Runway - Technical Customer Support

• 🧪 Anthropic - Research Engineer

• 🎨 Parloa - Lead Product Designer

📰 Everything else in AI today

OpenAI is reportedly in talks to acquire coding platform Windsurf (formerly Codeium), in a deal worth as much as $3B.

Microsoft researchers unveiled BitNet b1.58 2B4T, a new 1-bit AI model that matches the performance of larger models while running efficiently on CPUs.

Tencent introduced FireEdit, a new AI image editing system that uses region-aware vision language models to enable more precise, instruction-based image modifications.

Anthropic is reportedly preparing to launch a new “voice mode” for Claude with three distinct AI voices named Airy, Mellow, and Buttery this month.

OpenAI’s testing partner Metr published its analysis of 3o and 4o-mini, noting an accelerated evaluation timeline that aligns with other reports of rushed safety testing.

Economist and author Tyler Cowen said he believes o3 qualifies as AGI, questioning if April 16 will be the day the technology officially crossed the barrier.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop on Tuesday, April 22nd, at 3 PM EST with Matt Waters from Superhuman. By the end of this workshop, you’ll have a fully optimized email system powered by AI, so you can move 4x faster and eliminate inbox chaos.

RSVP here. Not a member? Join The Rundown University on a 14-day free trial.

🤝 Share The Rundown, get rewards

We’ll always keep this newsletter 100% free. To support our work, consider sharing The Rundown with your friends, and we’ll send you more free goodies.

See you soon,

Rowan, Joey, Zach, Alvaro, and Jason—The Rundown’s editorial team