Nvidia's 6-in-1 chip

ALSO: How to dictate anywhere on your computer with AI

Author

Zain Kahn
January 06, 2026

Welcome back, Superhuman. Nvidia CEO Jensen Huang delivered an action-packed CES keynote, featuring everything from self-driving car tech to on-stage droids. But it was the company's next-generation chip architecture that took center stage. Literally…it took up the entire stage.

Today: Alexa becomes a chatbot, learn how to dictate anywhere on your computer with AI, and get the latest prompts and trending social posts.

TODAY IN AI

Click to watch Nvidia’s CES 2026 Keynote in 9 minutes. Source: CNET.

1. Nvidia launches next-generation AI chip architecture: At CES 2026, CEO Jensen Huang unveiled the Vera Rubin platform, calling it the new "state of the art" in AI hardware. The Rubin platform — Blackwell’s successor — uses six chips in one supercomputer to deliver 5x faster inference and 3.5x faster training than Blackwell, all while using 10x less cost per token. Huang confirmed Vera Rubin is "in full production" and will ramp up in the second half of 2026, with major cloud providers already slated to deploy Rubin systems.

2. Amazon brings Alexa to the web, challenging ChatGPT and Gemini: Also at CES, Amazon unveiled Alexa+, a web-based chatbot for its Alexa assistant. The chatbot will reportedly focus on families, tackling tasks like controlling smart devices, updating shared calendars, and handling requests across Amazon’s other offerings. Alexa+ is free during Early Access, then free for Prime members or $19.99/month without. Notably, Amazon has also invested $8 billion in Claude owner, Anthropic.

3. Turbulence brews at Meta as it rushes to catch up in AI race: The social media giant is reportedly planning to release two new AI models in early 2026: an image/video generator codenamed "Mango" and a text model called "Avocado” — but friction is mounting at Meta’s AI division. Yann LeCun, Meta's former chief AI scientist, departed in November and has since criticized the company’s AI strategy. At least three researchers have left Meta's AI team since its launch, with some returning to OpenAI after brief stints.

Tech moves fast, but you're still playing catch-up?

That's exactly why 100K+ engineers working at Google, Meta, and Apple read The Code twice a week.

Here's what you get:

  • Curated tech news that shapes your career - Filtered from thousands of sources so you know what's coming 6 months early.

  • Practical resources you can use immediately - Real tutorials and tools that solve actual engineering problems.

  • Research papers and insights decoded - We break down complex tech so you understand what matters.

All delivered twice a week in just 2 short emails. Join 100K+ engineers.

  1. Download Wispr Flow for Mac or Windows and install.

  2. Open the app and sign in.

  3. Click into any text field, hold your hotkey, talk, and release to insert text.

    • Mac default: Fn

    • Windows default: Ctrl + Win

  4. To match tone with your writing style, open Flow Hub → Style, pick a context tab (Personal, Work, Email, or Other), and choose your vibe (Formal, Casual, Very Casual, or Excited).

We recommend Casual for friendly, clean phrasing with lighter punctuation. Your choice applies everywhere you dictate on the desktop.

Download Wispr Flow and stop letting the keyboard bottleneck your brain.

FROM THE FRONTIER

DeepSeek just published new research — the last time it did, a model followed that shook the market

Made with Midjourney

DeepSeek’s newest paper. DeepSeek’s researchers just introduced a new training framework: Manifold-Constrained Hyper-Connections (mHC). Wei Sun, an AI analyst at Counterpoint Research, called the technique a "striking breakthrough," while Omdia's chief analyst said it could have a "ripple effect across the entire AI industry." But what makes this framework so special?

Solving a scaling issue. When AI models get bigger, they often crash mid-training because signals between layers can spiral out of control. These mid-training crashes waste valuable time and resources, but mHC offers a solution. MHC lets researchers train bigger AI models safely by constraining how information moves between layers — so signals stay stable at any scale.

Does this actually work? The mHC framework has been tested on models up to 27 billion parameters, but not the frontier-scale models where training failures are most costly. Some analysts believe DeepSeek might deploy a new model featuring the research in the coming months.

The implications. While US labs pour billions into GPUs and data centers, DeepSeek is betting on efficiency. Its last model, R1, matched OpenAI's performance at a fraction of the cost — a disruption that erased nearly $600 billion from Nvidia's market capitalization. If the scenario repeats itself, 2026 could bring another shock to the market.

Get audit-ready fast and stay continuously compliant — and turn compliance into a deal accelerator.

Join Vanta’s live demo on Jan 15 to learn how to accelerate compliance with deep automation and agentic workflows across frameworks like SOC 2, ISO 27001, HIPAA, HITRUST, and more.

IN THE KNOW

What’s trending on socials and headlines today

Meme of the day

📝 Prompt Tips: Learn 10 prompting best practices to supercharge Claude Code’s output — directly from the team at Anthropic.

⏳️ Work Consolidation: A Principal Engineer at Google says Claude Code consolidated a year’s worth of work into just one hour (this post has 7 million views).

🤖 Claude Skills: A viral post resurfaced instructions on how to set up Claude skills — claiming that it’s one of the best upskills you can learn this year.

🤯 Neural Networks: This video breaks down how neural networks really work using a visual graph.

😺 Cat Electricians: What happens when you let cats work on your electrical wiring? Click here to find out.

PRODUCTIVITY

5 New & Trending AI Tools

  • 💼 The AI Consultancy Project*: Turn your AI expertise into a 6-figure consultancy as the market explodes. We’ve trained 1,200+ consultants–apply now for early access.

  • 🎲 Imagen3D: Convert images to high-quality 3D models.

  • 🧑‍🏭 Gobii: Create your own digital worker that works for you 24/7.

  • 🖌️ Sketchflow: Generate designs, prototypes, and front-end code for your app or website.

  • 📃 Kael: Understand complex documents, connect ideas across sources, and produce clear outputs with AI.

* indicates a promoted tool, if any

PROMPT STATION

High-Clarity Rewrite

Prompt: Take the text below and refine it for clarity, tone, and flow while preserving the original meaning, intent, and key details.

Remove awkward phrasing, redundancy, and filler

Improve sentence structure and transitions

Make it sound natural, confident, and human-written (not AI-ish)

Do not add new ideas, facts, or opinions

Maintain the original voice and level of formality

Constraint: Keep the final version under [X] words.

Output: Return only the improved text—no explanations or commentary.

Landmark Infographics

Source: TechieBySA

NanoBanana Prompt: “Create an infographic image of [LANDMARK], combining a real photograph of the landmark with blueprint-style technical annotations and diagrams overlaid on the image. Include the title “[LANDMARK]” in a hand-drawn box in the corner. Add white chalk-style sketches showing key structural data, important measurements, material quantities, internal diagrams, load-flow arrows, cross-sections, floor plans, and notable architectural or engineering features. Style: blueprint aesthetic with white line drawings on the photograph, technical/architectural annotation style, educational infographic feel, with the real environment visible behind the annotations.”

Pro Tip: Use the same prompt in ChatGPT to get similar results.

Whenever you’re ready to take the next step

Grow customers & revenue: Join companies like Amazon, HubSpot, and Salesforce. Showcase your product to our 1M+ readers and 2M+ followers on socials. Get in touch.

What did you think of today's email?

Your feedback helps me create better emails for you!

Login or Subscribe to participate in polls.

Until next time — Zain & the Superhuman AI team