In partnership with

Good morning. It’s Friday, October 25th.

Did you know: On this day in 2001, Windows XP was released?

In today’s email:

• OpenAI’s Orion

• Apple Intelligence Update

• Elevenlabs Prompt-to-Voice

• Anthropic’s Code Writer

• 4 New AI Tools

• Latest AI Research Papers

You read. We listen. Let us know what you think by replying to this email.

Writer RAG tool: build production-ready RAG apps in minutes

• Writer RAG Tool: build production-ready RAG apps in minutes with simple API calls.

• Knowledge Graph integration for intelligent data retrieval and AI-powered interactions.

• Streamlined full-stack platform eliminates complex setups for scalable, accurate AI workflows.

Learn more about our production ready RAG tooling here.

OpenAI plans to release its next model ‘Orion’ by December

OpenAI is preparing to launch its next model, Orion, by December, according to sources who spoke with The Verge.

Unlike its previous models, GPT-4o and o1, Orion won't be widely available on ChatGPT at first. Instead, OpenAI plans to offer early access to select partner companies to integrate it into their own products and features. Sources also indicate that Microsoft engineers — OpenAI's key deployment partner — are gearing up to host Orion on Azure as early as November.

While Orion is internally considered the follow-up to GPT-4, it’s not confirmed whether it will be branded as GPT-5 for public release. As with past launches, the timeline could shift. OpenAI and Microsoft did not provide comments on this information. Read more.

Apple Intelligence expands in iOS 18.2 developer beta, adding Genmoji, Visual Intelligence and ChatGPT

Apple's iOS 18.2 developer beta expands its Apple Intelligence features, introducing Genmoji, Visual Intelligence, and ChatGPT integration. Genmoji allows users to create custom emojis by entering descriptions. Image Playground generates images based on conversations or user inputs, while Image Wand enhances sketches made with the Apple Pencil.

ChatGPT integration with Siri enables users to request complex tasks, such as itinerary planning, with suggestions to utilize OpenAI's capabilities. Visual Intelligence offers iPhone 16 users real-time information by pointing their camera at objects. Read more.

ElevenLabs now lets you create a custom voice from a text prompt

ElevenLabs has launched a Voice Design engine that allows users to create custom synthetic voices from text prompts, enhancing generative AI's dialogue capabilities. This tool enables the rapid design of unique voices, making it ideal for applications in gaming and multimedia where real-time voice changes are needed.

While trained voice actors are still important for character portrayal, this technology offers indie developers the chance to craft engaging narratives and interactive experiences independently. The effectiveness of the Voice Design engine depends on the specificity of the prompts, enabling tailored outputs like “an old British male with a raspy, deep voice—professional, relaxed, and assertive.” ​​Read more.

Anthropic introduces a new feature that lets it write data analysis with JavaScript

Anthropic has launched a JavaScript-powered analysis tool for its Claude AI chatbot, allowing users to perform in-depth data analysis directly in the Claude.ai interface. This feature, currently in preview, enables Claude to write, execute, and adjust code to process datasets, generate dynamic visualizations, and perform complex calculations. Users can upload files such as CSVs and PDFs, enabling Claude to create various interactive charts like trend lines, pie charts, and scatter plots.

Tailored for both sales and engineering applications, the tool provides insights from resource logs to global performance trends, with interactive elements that reveal additional data points. Built on Claude 3.5 Sonnet’s enhanced coding capabilities, the tool also suggests follow-up queries and modifications, making AI-driven data exploration accessible for users without coding expertise. The new feature is positioned as an alternative to OpenAI’s Advanced Data Analysis tool. Read more.

• Microsoft Silently Released OmniParser: A tool that converts UI screenshots into structured formats

• Raspberry Pi releases more AI-focused add-ons

• Meta debuts slimmed-down Llama models for low-powered devices

• Cohere announces Aya Expanse multilingual AI model family for researchers 

• OpenAI disbands another team focused on advanced AGI safety readiness 

• Perplexity launches a native Mac app

• Nvidia CEO touts India's progress with sovereign AI and over 100K AI developers trained

• Microsoft's "DIFF Transformer" promises more efficient LLMs with fewer hallucinations

• Nvidia's design flaw with Blackwell AI chips now fixed, CEO says

• OpenAI's new AI model generates images 50 times faster than before

• Hugging Face's new dollar-per-hour service helps companies break free from third-party AI

• Apple will pay security researchers up to $1 million to hack its private AI cloud

• Ray Kurzweil’s 2025 Vision: AGI, Robots in the Workforce, and the Nanotech Frontier

• China Unveils New Approach to Road Construction. It Involves 98 Miles, 10 Machines, and No Humans at All

• Disney Poised to Announce Major AI Initiative 

• OpenAI scientist Noam Brown stuns TED AI Conference: '20 seconds of thinking worth 100,000x more data'

• AI chatbot company Character.ai adds safety measures after suicide of teen user

Escaping AI POC purgatory: Techniques for enterprise AI engineers

Many companies struggle to move generative AI from experimentation to production.

Join us Oct. 29 at 9am PT. Sam Julien, Writer's Director of Developer Relations, will share practical strategies to overcome enterprise AI engineering challenges using a full-stack approach.

Topics include:

• Managing project scope

• Improving accuracy with domain-specific models & graph-based RAG

• Navigating AI application development

Can’t make it live? Register anyway and we’ll send you the recording.

Register now

📄 PyramidDrop: Accelerating Your Large Vision-Language Models via Pyramid Visual Redundancy Reduction

📄 WorldSimBench: Towards Video Generation Models as World Simulators  ​​

📄 MIA-DPO: Multi-Image Augmented Direct Preference Optimization For Large Vision-Language Models

📄 Breaking the Memory Barrier: Near Infinite Batch Size Scaling for Contrastive Loss  

📄 Unleashing Reasoning Capability of LLMs via Scalable Question Synthesis from Scratch

Thank you for reading today’s edition.

Your feedback is valuable. Respond to this email and tell us how you think we could add more value to this newsletter.

Interested in reaching smart readers like you? To become an AI Breakfast sponsor, reply to this email or DM us on X!