Sign Up | Advertise | Podcast | AI University | | | Welcome, AI enthusiasts. | OpenAI has plenty of presents for consumers this holiday season, but the latest release is all about the builders. | With the o1 reasoning model now available via API and a toolbox of powerful new features, AI applications are about to level up in a big way. Let’s get into it… | | In today’s AI rundown: | OpenAI releases o1 for API, new developer tools Nvidia’s cheap, palm-sized AI supercomputer Run CLI commands from your prompts New DeepMind benchmark tests LLM factuality 5 new AI tools & 5 new AI jobs More AI & tech news
| Read time: 4 minutes |
|
| | | | OPENAI | |  | Image source: OpenAI on YouTube |
| The Rundown: OpenAI just rolled out a series of updates for developers on Day 9 of its live stream event, including API access to its advanced o1 reasoning model, major upgrades to the Realtime API, and a new preference fine-tuning method. | The details: | o1 comes out of preview with new API capabilities like function calling, structured outputs, vision, and reasoning effort to control thinking time. o1 API costs come in at $15 per ~750k words analyzed and $60 per ~750k words generated — roughly 3-4x more than GPT-4o. Realtime API costs drop 60% for GPT-4o audio, with a new 4o mini available at 1/10 the price and WebRTC integration for easier voice app development. New Preference Fine-Tuning enables customizing models using comparative examples vs fixed training data, improving tasks like writing and summarization. The company also launched beta SDKs for Go and Java programming languages, expanding development options.
| Why it matters: These releases mark a big day for AI builders. Access to the o1 reasoning model and new features opens up a new world of building and integrations. Developers just got a powerful new set of tools to create more customizable, capable, and sophisticated applications. |
|
| | TOGETHER WITH STAIRCASE AI | | | The Rundown: Staircase AI by Gainsight analyzes millions of customer interactions like emails, tickets, and chats to surface risks and growth opportunities you can actually take action on. | With Staircase, you can: | Automate data capture for insights without the hassle Uncover unbiased customer sentiment and health with AI Correlate effort to outcomes and optimize resources
| Learn how to stay one step ahead with Staircase AI. |
|
| | NVIDIA | |  | Image source: Nvidia |
| The Rundown: Nvidia just introduced the Jetson Orin Nano Super Developer Kit, a $249 compact generative AI supercomputer that delivers significant performance gains at half the previous model's price. | The details: | The palm-sized device delivers 1.7x the performance, 70% more processing power, and a 50% boost in memory compared to the previous model. The Nano can handle multiple AI tasks simultaneously, from powering chatbots to controlling robots and processing visual data from multiple cameras. The platform supports popular AI frameworks and tools through NVIDIA's software ecosystem, including Isaac for robotics and Metropolis for vision AI. Existing Jetson Orin Nano owners can access the same 1.7x generative AI performance gains through a free software update.
| Why it matters: Just as the Raspberry Pi revolutionized DIY computing projects, NVIDIA's affordable AI supercomputer could birth a new generation of developers building everything from smart robots to creative AI tools in their garages and dorm rooms. The barriers to advanced AI tools have never been lower. |
|
| | PRESENTED BY PROJECT IDX | | | The Rundown: Project IDX is your AI-enabled development environment in the cloud with code assistance from Gemini. | Step-by-step: | Log into Project IDX with your Google account. Create a project from the templates dashboard. Click the Gemini icon at the bottom of the workspace or press Cmd+Shift+Space (Ctrl+Shift+Space on ChromeOS, Windows, or Linux). Select “Interactive Chat with Gemini” and pass it a prompt to update your project configuration files and code.
| Pro tip: Project IDX offers a streamlined development experience by allowing you to execute terminal commands directly from the user interface. |
|
| | AI RESEARCH | |  | Image source: Google DeepMind |
| The Rundown: Google DeepMind just launched FACTS Grounding, a new benchmark designed to evaluate how well LLMs can generate factually accurate and comprehensive responses based on provided documents while avoiding hallucinations. | The details: | FACTS uses 1,719 examples, each with a document, a system instruction, and a user request, to test the ability to produce grounded long-form answers. Three AI models (Gemini 1.5 Pro, GPT-4o, and Claude 3.5 Sonnet) serve as judges, evaluating responses for accuracy and handling user requests. Scores are aggregated across all judges and examples, with results published on a public Kaggle leaderboard that will be updated as new models emerge. Google's Gemini models currently top the leaderboard, with Gemini 2.0 Flash Experimental achieving the highest score, 83.6%, for factual grounding.
| Why it matters: Hallucinations continue to plague even the most advanced LLMs, limiting reliability and real-world use cases. FACTS Grounding provides a more nuanced way to measure progress in an extremely important developmental area for AI by focusing on grounded responses and using a multi-LLM judging approach. |
|
| | | | | 📸 Google Imagen 3 - Google’s highest-quality text-to-image model, capable of generating images with even better detail, lighting, and fewer artifacts 🎨 Google Whisk - Generate images by using other images as prompts for a subject, scene, and style to create personalized visuals ☎️ NewOaks AI Phone Agent - AI phone agent that can listen, understand, and speak in real-time to automate inbound and outbound calls 🧠 Findr- Unlock infinite digital memory with your AI second brain 📫 MagicMail - AI email generator that turns text prompts into fully styled and ready-to-send HTML emails
| | |
|
| | | | Free event on Jan 7: Join Section for a discussion on how AI is redefining how we work, hire, and advance in the workplace. RSVP for free.* | Midjourney released Moodboards, a new feature that allows users to create personalized AI generation styles and profiles by uploading or adding images. | Google launched Gemini Code Assist tools, enabling developers to access external services and data directly within their IDE. | YouTube is partnering with talent agency giant CAA to develop AI detection tools to help celebrities and athletes identify and manage AI-generated content featuring their likenesses across the platform. | UAE’s Technology Innovation Institute released Falcon 3, a new open-source family of language models designed to run on lightweight hardware, with the 7B and 10B versions outperforming competitors like Llama and Qwen in key benchmarks. | OpenAI’s Romain Huet revealed during a community AMA that there are currently no plans to release an API for the company’s Sora video generation model. | Databricks secured a new $10B funding round at a $62B valuation, with the data analytics company planning AI product expansion and potential acquisitions. | *Sponsored listing |
|
| | | | SPONSOR US | Get your product in front of over 800k+ AI enthusiasts | Our newsletter is read by thousands of tech executives, investors, engineers, managers, and business owners around the world. Get in touch today. |
|
| | 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. | | See you soon, | Rowan, Joey, Zach, and Alvaro—aka The Rundown Team | |
|
|
|
