3 DAYS AGO • 5 MIN READ

☕️ DeepSeek's Stealth R1 Update, xAI's $300M Telegram Deal & The NYT's First AI Deal with Amazon

profile

AI Tangle

AI Tangle provides timely, relevant AI news and tools tailored to help business leaders stay ahead of the curve. Our concise, actionable updates ensure you’re equipped to make informed decisions in a rapidly evolving AI landscape. As part of The AIE Network, AI Tangle connects you with additional resources such as AI Marketing Advantage and The Artificially Intelligent Enterprise for complete AI-driven business transformation.

This week saw DeepSeek quietly release an update to R1, Elon Musk's xAI shake hands with Telegram for a Grok deal, and popular American newspaper NYT sign its first AI deal.

Here are our main key takeaways in more detail:

  • China's DeepSeek releases a stealth update to its R1 reasoning model, brings it closer to o4-mini and o3
  • xAI signs a deal with Telegram for $300 million to bring Elon Musk's Grok to the messaging platform
  • The New York Times pens its first AI deal with Amazon, will integrate its content into Alexa and AI training

Join us at AI Tangle as we untangle this week's happenings in AI!

First reported by Bloomberg on Wednesday, DeepSeek has quietly updated its R1 reasoning model to R1-0528 on Hugging Face without any official release notes or performance charts. According to the LiveCodeBench leaderboard by UC Berkeley, MIT, and Cornell researchers, DeepSeek R1-0528 now closely trails OpenAI's o4-mini and o3 models in code generation and outperforms xAI's Grok 3 mini and Alibaba's Qwen 3, though it does seem to be more censored. However, most of the AI industry's eagerness is held in anticipation for R2, a model that Reuters reported in February to be on the verge for some time now.

Elon Musk's AI startup, xAI, will pay Telegram $300 million in cash and stock to integrate its Grok chatbot into the messaging platform, bringing its AI to over one billion users. The one-year agreement will include a 50/50 split of any xAI subscription revenue generated through the app, as stated by Telegram's founder, Pavel Durov. On top of the revenue, the partnership could also yield valuable user data for xAI's future model training - X/Twitter's privacy policy says the company uses user data for training AI, yet it is unclear whether that may apply to its Telegram deal as well.

The New York Times has recently penned a multi‑year licensing agreement with Amazon to integrate its editorial content, meaning summaries and short excerpts, into the e-commerce giant's services like Alexa as well as to train its AI models. The partnership also covers The Times' sports site, The Athletic, and its recipe platform, NYT Cooking. After The Times' fierce lawsuits against Microsoft and OpenAI for alleged copyright infringement, The Times now joins the list of other publishers in striking paid AI content deals, such as The Atlantic, News Corp, and Vox Media.

Anthropic is rolling out an English-only beta version of a voice mode in its Claude mobile apps, powered by Claude Sonnet 4 by default. As is usual with voice modes, users will be able to speak to Claude and hear spoken replies, with key points displayed on‑screen together with seamless switching between text and voice. Anthropic currently offers five distinct voices, along with transcripts and summaries after each session. Anthropic's paid subscribers will also gain Google Workspace integration, such as Claude for Calendar and Gmail. Broader availability across all plans will come "over the next few weeks."

Meta is attempting to reorganize its AI teams once more into two focused, purpose-built teams. Led by Connor Hayes, Meta's new AI Products team will own Meta AI assistant, AI Studio, and in-app features across the company's social media apps like Facebook, Instagram, and WhatsApp. Meanwhile, The AGI Foundations unit will be co-led by Ahmad Al-Dahle and Amir Frenkel, which will develop core technologies like its main Llama line of AI models and enhance reasoning, multimedia, and voice capabilities. FAIR, Meta's AI research unit, will remain independent, with a multimedia subgroup moving under AGI Foundations.

Black Forest Labs, the company behind Grok's image generation capabilities, has recently unveiled FLUX.1 Kontext, a next‑gen image generation model designed for enterprise workflows that supports both text‑ and image‑based in‑context editing. Available in two flavors, [pro] and [max], Kontext preserves character consistency across scenes, enables high-precision local edits, adheres closely to the user's style references, and operates with minimal latency. Developers can experiment with all versions, including a private beta version called [dev], in the newly launched BFL Playground before integrating the full‑featured BFL API into their applications.

Popular writing assistant Grammarly has raised a staggering $1 billion in funding from General Catalyst's Customer Value Fund to speed up its transformation into a broader AI productivity platform. Grammarly says the funding will go towards fueling product innovation, expanding sales and marketing efforts, and strategic acquisitions, such as last year's buy-out of AI productivity tool maker Coda. With over 40 million users and $700 million in annual revenue, Grammarly plans to deepen its AI assistant and agent capabilities to become a fully-fledged work partner for brainstorming, composition, and tone optimization.

Supernormal - Spend less time writing, polishing, and sharing notes and more time on the work that matters with the help of Supernormal's AI-powered meeting transcriptions.

Mesh - Stay on top of expenses, runway, and revenue without the need for spreadsheets or guesswork by using Mesh, an AI analyst that integrates into and streamlines your finances.

Lindy - Automate your workflows and help your business grow with Lindy's all-in-one AI agent suite - build agents in minutes to automate lead generation, customer support, outreach, and more.

Gem - Unify your recruiting tech stack with Gem, a platform powered with AI, CRM, and analytics to provide an all-in-one solution to recruiters' problems.

Anthropic CEO's Bleak Future Vision for White Collar Jobs in 5 Years (5-min read)

Dario Amodei warns that AI's rapid advance could finger-snap away half of entry‑level white‑collar roles within five years if the world's governments and the industry's most advanced sit idly, spiking unemployment to levels reminiscent of the Covid times.

🔮 The Artificially Intelligent Enterprise - Don't Sleep on the ChatGPT Desktop App

💡 AI CIO - New NIST Metric Could Tackle AI Vulnerabilities

🎯 The AI Marketing Advantage - Google's AI Overviews Now Include Sponsored Content

📻 AI Confidential Podcast - On the Cutting Edge of Agentic AI with João Moura

Your AI Sherpa,

Mark R. Hinkle

Publisher, The Artificially Intelligent Enterprise (TheAIE) Network

Connect with me on LinkedIn

Follow me on X

AI Tangle

AI Tangle provides timely, relevant AI news and tools tailored to help business leaders stay ahead of the curve. Our concise, actionable updates ensure you’re equipped to make informed decisions in a rapidly evolving AI landscape. As part of The AIE Network, AI Tangle connects you with additional resources such as AI Marketing Advantage and The Artificially Intelligent Enterprise for complete AI-driven business transformation.