Become An AI Expert In Just 5 Minutes

If you’re a decision maker at your company, you need to be on the bleeding edge of, well, everything. But before you go signing up for seminars, conferences, lunch ‘n learns, and all that jazz, just know there’s a far better (and simpler) way: Subscribing to The Deep View.

This daily newsletter condenses everything you need to know about the latest and greatest AI developments into a 5-minute read. Squeeze it into your morning coffee break and before you know it, you’ll be an expert too.

Subscribe right here. It’s totally free, wildly informative, and trusted by 600,000+ readers at Google, Meta, Microsoft, and beyond.

Today:

• $100B Power Play: OpenAI Taps NVIDIA for Next-Gen AI

• DeepSeek-V3.1 Launches with Smarter Tools and Cleaner Language

• Perplexity Launches AI Assistant for Gmail & Outlook

• Alibaba’s Qwen3-Omni Challenges U.S. AI Giants

• Oracle Names New Co-CEOs Amid Major AI Push

$100B Power Play: OpenAI Taps NVIDIA for Next-Gen AI

OpenAI and NVIDIA plan a huge buildout: at least 10 gigawatts of NVIDIA systems—millions of graphics chips—to train and run next-gen AI. NVIDIA may invest up to $100 billion as capacity comes online, starting with one gigawatt in late 2026 on the Vera Rubin platform.

KEY POINTS

• Scale: 10 gigawatts (a very large amount of power) of NVIDIA systems, representing millions of chips for training and running future models.
  

• Funding: NVIDIA intends to invest up to $100 billion, released in steps as each gigawatt is deployed.
  

• Timeline & roles: First 1-gigawatt site targets H2 2026 on Vera Rubin; NVIDIA is OpenAI’s preferred compute and networking partner, coordinating hardware and software plans together.
  

Why it matters
More compute means faster progress in AI. This deal aims to remove hardware bottlenecks so OpenAI can build stronger, more helpful tools. It also shows deep commitment—and big money—behind AI growth, which can shape new products, business uses, and the wider tech economy.

DeepSeek-V3.1 Launches with Smarter Tools and Cleaner Language

DeepSeek launched V3.1-Terminus, an open-source update that boosts tool-using skills (coding, web search), fixes English-Chinese mixing, and adds steadier chat and reasoner modes. Benchmarks rise on agent tasks; pure reasoning improves slightly. It’s available via Hugging Face and API, with low token-based pricing and self-hosting.

KEY POINTS

• Better performance on tool tasks (coding, browsing, search); small gains on tough reasoning; minor dip on one coding benchmark.
  

• Two modes: chat (with tools) and reasoner (deeper thinking); 128k context; higher output limits; fixes mixed-language glitches.
  

• Open MIT license; available on Hugging Face, apps, and API; inexpensive token pricing; can be self-hosted (enterprises should vet API use).
  

Why it matters

Terminus gives developers a faster, cheaper, and more reliable open model for real work. Stronger tool use means better coding and research help. The open license widens access, offering a practical alternative to closed systems while improving language quality for global users.

Perplexity Launches AI Assistant for Gmail & Outlook

Perplexity launched a $200 per month Email Assistant for Gmail and Outlook. It sorts mail, writes tone-matched replies, and auto-schedules meetings. Aimed at businesses, it promises inbox zero. Users question price and data access. Reviews show help on simple tasks but errors on complex ones.

KEY POINTS

• Features: manages inbox, drafts in your style, handles back-and-forth scheduling, daily summaries; works inside Gmail and Outlook.
  

• Pricing and positioning: Max plan only at $200/month; aimed at business users; backlash from existing Pro users.
  

• Trust and limits: needs broad account access; company says emails aren’t used for training; still makes mistakes on complex requests.
  

Why it matters

Email eats workdays. If reliable, this tool could save hours by sorting, replying, and booking meetings for you. The steep price tests whether businesses will pay for real time savings. It also pushes Google and Microsoft to speed up similar tools while keeping privacy promises.

🧠RESEARCH

RPG: A Repository Planning Graph for Unified and Scalable Codebase Generation

A new system called ZeroRepo uses a graph-based blueprint (RPG) to generate full software codebases from scratch. It avoids confusing language and plans software more clearly. On a tough benchmark, it built codebases four times bigger than the best rival and passed more tests, showing major gains in automation.

MANZANO: A Simple and Scalable Unified Multimodal Model with a Hybrid Vision Tokenizer

Manzano is a new multimodal AI model that can both understand and generate images using text. It combines a hybrid image tokenizer with a smart training setup, avoiding the usual trade-offs. With strong performance in both directions, it matches or beats specialized models, proving simple, scalable, and effective in real-world tasks.

Evolving Language Models without Labels: Majority Drives Selection, Novelty Promotes Variation

EVOL-RL is a new training method that improves language models without needing labels. It balances stability (by keeping the best answers) and variation (by rewarding new reasoning). This prevents models from becoming repetitive and brittle. It leads to better performance, longer responses, and stronger generalization across tasks—even outperforming label-free baselines.

🛠️TOP TOOLS

PicWish – AI Photo Editor & Image Generator - photo editor that bundles background removal, enhancement, object/watermark removal, AI background creation.

SciSummary – AI Research Assistant / AI Paper Summarizer - domain‑specific AI for academic papers.

Quick QR Art – AI Image Generator -turns links, text, and more into artistic, scannable QR codes.

📲SOCIAL MEDIA

🗞️MORE NEWS

• Alibaba’s Qwen3-Omni is an open-source, free-to-use-and-modify artificial intelligence that takes text, images, audio, and video, and replies with text or speech. It delivers quick responses and strong scores, challenging U.S. rivals including OpenAI and Google.

• Oracle named Clay Magouyrk and Mike Sicilia co-CEOs; Safra Catz becomes executive vice chair. The move strengthens its AI push, following compute deals with OpenAI and Meta and participation in the $500 billion Stargate center.

• Meta is adding an AI chatbot to Facebook Dating to refine profiles and surface tailored matches, plus a weekly “Meet Cute” surprise match—continuing the industry’s shift toward AI-powered matching across major apps.

• MIT mathematicians David Roe and Andrew Sutherland won AI for Math grants to link the LMFDB database with Lean’s mathlib, accelerating automated theorem proving. Four MIT alumni also received grants, broadening formal math tools and discovery.

• SchoolAI, built on OpenAI (GPT-4.1, GPT-4o, image generation, text-to-speech), powers 1M classrooms in 80+ countries. Teachers create lessons with “Dot,” students get coached by “Sidekick,” all safely teacher-observed—saving time, lowering costs, and boosting engagement and early interventions.