In partnership with

Welcome back!

OpenAI and Nvidia just unveiled a 10-gigawatt AI compute pact, a scale on par with the power use of a country. With millions of GPUs and $100 billion in investment, they reframe data centers as “AI factories” where energy turns into AI breakthroughs. Let’s unpack…

Today’s Summary:

• 🔥 OpenAI & Nvidia launch 10GW pact

• 👀 Qwen3-Omni & VL expand multimodal AI

• 📊 ChatGPT & Claude usage revealed

• 🚫 OpenAI restricts teens’ ChatGPT access

• ⚡ Grok 4 Fast launches cheap, fast model

• 💡 Why language models hallucinate

• 🛠️ 3 new tools

TOP STORY

OpenAI to deploy 10GW of Nvidia systems, matching a small country’s power use

The Summary: OpenAI and Nvidia announced plans to deploy 10 gigawatts of AI data centers, representing several millions of GPUs to power OpenAI’s next generation of AI models. Nvidia will invest up to $100 billion in OpenAI as infrastructure comes online, starting with the Vera Rubin platform in 2026.

Key details:

• A 10GW capacity equals the power draw of a small country with several million GPUs running nonstop

• First 1GW deployment on Nvidia Vera Rubin platform set for 2026

• Future chips will bring energy savings, meaning fixed 10GW of data centers will deliver more compute over time, while power remains the same

• Nvidia’s $100B investment is structured as non-voting equity, raising “circular financing” concerns since much of the cash will flow back again to Nvidia as GPU purchases

Why it matters: Scaling compute has become OpenAI’s growth engine. Sam Altman calls the 10 GW the “literal key” to revenue, framing data centers as AI factories that turn energy and chips into new AI breakthroughs and profit. This new Nvidia/OpenAI pact of unprecedented scale makes compute a strategic commodity, managed with the same importance once given to oil, steel, and semiconductors.

FROM OUR PARTNERS

Write 3x Faster Just by Speaking

Smart dictation that understands you

Typeless turns your raw, unfiltered voice into beautifully polished writing - in real time.

It works like magic, feels like cheating, and allows your thoughts to flow more freely than ever before.

With Typeless, you become more creative. More inspired. And more in-tune with your own ideas.

Your voice is your strength. Typeless turns it into a superpower.

Download for Mac today

OPEN SOURCE

Qwen3-Omni and Qwen3-VL expand open multimodal AI with speech and vision

The Summary: Several major releases dropped from the Qwen team. Qwen3-Omni 30B integrates text, audio, image, and video input with text and real-time speech output. Alongside it, Qwen3-VL 235B is an advanced vision model built for visual reasoning and action, including 2-hour video understanding and multi-language OCR. Both are released with open weights and a free license, positioning the Qwen3 series as the most ambitious open multimodal stack to date. Moreover, Qwen3-Max is a trillion-parameter closed-source text model that now sits above GPT-5 Chat on LM Arena.

Key details:

• Qwen3-Omni delivers low latency audio and video streaming and is runnable on consumer 24GB GPUs

• Includes a low-hallucination open audio captioner

• Bug: some voices play at slow speed in English

• Qwen3-VL supports 2-hour video analysis with timestamps

• OCR handles blur, tilt, and even rare scripts, with bounding boxes

• Both models released under an Apache 2.0 free license

• Max is closed-source and available via Qwen Chat and API

Why it matters: Qwen3-Omni and VL attack two different bottlenecks. Omni lowers the barrier on speech-first assistants and can run on local GPUs, making local multimodal assistants viable earlier than expected. VL enables reasoning and transcription on complex visual documents. On top, Max shows Qwen also competes at frontier scale, placing above GPT-5 Chat and rivaling Claude Opus and Grok Heavy.

Qwen3-Omni Demo | Qwen3-VL Demo

FROM OUR PARTNERS

Turn Feedback Chaos Into Clarity

How Canva, Perplexity and Notion turn feedback chaos into actionable customer intelligence

You’re sitting on a goldmine of feedback: tickets, surveys, reviews, but can’t mine it.

Manual tagging doesn’t scale, and insights fall through the cracks.

Enterpret’s AI unifies all feedback, auto‑tags themes, and ties them to revenue/CSAT, surfacing what matters to customers.

The result: faster decisions, clearer priorities, and stronger retention.

👉 See how top teams do it

RESEARCH

How people actually use ChatGPT and Claude

The Summary: New research details how ChatGPT and Claude are being used in practice. ChatGPT is now part of everyday life, with most conversations about writing help, practical advice, and decision support. Claude, by contrast, is more often used in coding and business automation. These patterns show AI splitting into two roles: a personal advisor, and an automation tool for work.

Key details:

• ChatGPT non-work use increased from 53% to 73% of total chats in one year, with newer users leaning more toward personal use

• Writing dominates ChatGPT work requests, with two-thirds involving rewriting or translating user-provided text rather than creating new material

• Educational demand is large with 36% of ChatGPT’s “practical guidance” for tutoring or teaching requests

• Claude automations rose from 27% to 39% of total tasks

• 44% of Claude API use is for coding

Why it matters: Usage patterns reveal the real story of generative AI. Decision support drives personal use and automation structures enterprise workflows. ChatGPT is evolving into an advisor for most people. Claude, in contrast, is used mainly for technical and enterprise workflows. Together, they show a split role for AI: partner in daily life and automator for enterprises.

Claude Economic Index report  |  ChatGPT usage research paper

QUICK NEWS

Quick news

• OpenAI will restrict ChatGPT access for teenagers

• Grok 4 Fast is a cheap fast model with 2M context window

• Why language models hallucinate

TOOLS

🥇 New tools

• Snapdeck - Create professional presentations with AI, fully editable

• You can now use OpenAI Codex inside Cursor or VS Code

• Notion 3.0 - Introducing Notion AI Agents