😺 Microsoft + Claude = Office-

PLUS: Wanna see a robot get chainsawed?

Welcome, humans.

Wanna see a robot dog get chewed up by a chainsaw? Here you go!

This might be the most “cracked” (and we mean that in the most Silicon Valley way possible) robot video we’ve ever seen…

First off: we had NO idea you could hack up a robot dog like that and make the equivalent of a robo-corgi.

Secondly: All the robots in that video (not only corgi-D2) have a shared brain, which is a little terrifying (but very cool). This is Skild AI demoing their “omni-bodied” controller that hops between humanoids, arms, and quadrupeds and “re-plans” its movements when hardware breaks.

If “any robot, any task, one brain” holds up outside the sizzle reel, expect fewer brittle demos and more resilient bots on factory floors, inspection routes, and perhaps even the battlefield…

Luckily, no robo-cats were harmed in the making of that video… this video, on the other hand… we’re not sure TBH…

Here’s what happened in AI today:

Microsoft added Claude models to Office 365 Copilot.
Alibaba invested $53B in AI infrastructure with NVIDIA.
Oracle planned $15B bond sale for AI infrastructure.
The #2 social app in the app store atm sells user phone call data to AI firms.

Advertise in The Neuron here

P.S: Join us LIVE today at 10am PST / 12pm CST! We're doing something wild: live voice interviews with ChatGPT, Gemini, Grok, and Claude like they're actual podcast guests—and maybe we'll surprise you with a fifth guest too

Click the image above to join LIVE (or watch after the fact). If you’re early, then click “notify me” on the YouTube vid to get an alert when we start!

In a way, you could think of this as the world's first AI talk show. Is this ridiculous? Yes. Are you going to learn a lot? Also yes.

Microsoft Just Let Claude Into Office 365

NEWS BRIEF: Read the full story on Microsoft and Anthropic’s partnership here.

Microsoft and Anthropic just announced that Claude models are now available in Microsoft 365 Copilot—marking what could be the first time hundreds of millions of Microsoft users can try Claude. Starting today, users can choose between OpenAI's models and Anthropic's Claude Sonnet 4 and Opus 4.1 when using the Researcher agent or building custom agents in Copilot Studio.

When you're in Microsoft's Researcher agent (which digs through your emails, meetings, and documents), you'll see a dropdown to switch between models. Need Claude's writing skills? Click. Want OpenAI's technical prowess? Click back.

The catch: Claude runs on Anthropic's servers, not Microsoft's, so your data leaves Microsoft's ecosystem. Companies must opt-in through their admin center.

This is a big deal—Microsoft's basically admitting no single AI model rules them all. As Microsoft AI CEO Mustafa Suleyman told us in our latest podcast interview, they need “maximum optionality.” Different models excel at different tasks. By letting users choose, Microsoft's effectively creating an AI model marketplace inside Office. Also: The model picker menu lives on!

Are we living in a new era or what? Today it's Claude in Copilot—tomorrow, who knows? Can Copilot get Nano Banana next?!

FROM OUR PARTNERS

50 LLM Prompts to Crush BFCM 2025

Master every click, cart, and customer journey. Triple Whale’s new guide equips you with 50 AI prompts that turn overwhelming BFCM data into clear, actionable insights for peak holiday profitability.

Download it now to learn:

Do’s and don’ts of BFCM LLM prompting
Step-by-step workflows, data preparation, and tips for deeper analysis
Automation hacks to scale faster

Triple Whale’s Moby Agents transform BFCM performance with AI built for ecommerce. Get 50 prompts that save you hours of manual analysis, help catch issues before they hurt revenue, and reveal insights no human could spot.

Steal the prompts

Prompt Tip of the Day.

Nate B. Jones is about to blow your mind with what you can do with Notion 3.0. Nate says most people are missing how powerful this really is. It's not just another AI add-on in a SaaS tool; it's basically a custom AI agent builder hiding in plain sight.

But here's the thing: it's super prompt-dependent. After testing it extensively, Nate discovered 8 prompting rules that make all the difference:

Scope precisely: Say “work only on this page and its subpages” to avoid unwanted changes elsewhere in your wiki.
Demand receipts: Have it add “DONE” or “BLOCKED + reason” at the bottom so you know exactly what happened.
Think tables-first: Use databases over raw text. Tables are easier to sort, operate against, review, and fix later.
Use quality checks: Be explicit about completion conditions (length limits, required data, specific elements to include).
Don't create duplicates: Make it update existing similar content instead of creating new copies that dirty your context.
Create run logs: Track each change as if it needs to be undone—good agent architecture requires undo capability.
Write in plain strict language: Say "create six questions" not "create a few questions"—avoid vague phrases that encourage hallucination.
Don't let it make things up: Tell it to use "check this" and not mark done if it can't find claims in your input data.

This video is freaking awesome and you should watch the full walkthrough where he shows you exactly how to do this, then try this out yourself.

Treats to Try.

*Asterisk = from our partners. Advertise in The Neuron here.

*Incogni erases your sensitive data—like addresses and phone numbers—from the web. The internet remembers, but you can delete. Get 55% off with code NEURON.
NVIDIA Nemotron gives you open-source AI models, training datasets, and development tools to build custom AI applications that can reason through complex problems, write code, and adapt to your specific business needs (try on NVIDIA or Openrouter).
Sprinto auto-completes your security questionnaires in under 48 hours using your pre-approved knowledge base, helping you close deals faster without sacrificing accuracy—free to try for up to 5 questionnaires.
Conduit AI connects your business data from 500+ sources, letting you automate workflows and analyze information with simple questions—no technical skills required (free plan available, premium starts at $10/month).
Memories lets you instantly find anything in your videos by asking natural language questions instead of scrubbing through footage (raised $8M).
Emergent turns your words into complete web apps with built-in payments, user accounts, and hosting—no coding required (raised $23M).
Huxe (from former NotebookLM ppl!) turns your emails, calendar events, and news topics into interactive audio briefings you can listen to while on the go (raised $4.6M).

Around the Horn.

IDK who made this but the Venn Diagram between them and listeners to the Dwarkesh Patel podcast (fantastic AI podcast that also dives into history more than I ever knew I wanted) might be a flat circle

Google launched Search Live, enabling real-time visual searches through smartphone cameras that combine Google Lens and Gemini AI models for contextual, conversational responses.
Cohere reached $7B valuation after raising $100M extension to its recent $500M round and announced a strategic partnership with AMD to run its AI models natively on AMD Instinct GPUs.
Oracle planned a $15B bond sale to fund AI infrastructure expansion supporting deals like its $300B OpenAI partnership, including a rare 40-year bond tranche.
Alibaba partnered with NVIDIA to add physical AI tools to its cloud platform, investing $53 billion in AI infrastructure over three years while expecting 10x data center power use by 2032.
Google expanded its $4.50-$6 monthly AI Plus plan to over 40 countries outside high-income markets, directly competing with OpenAI's similarly priced ChatGPT Go in international markets.
Neon, the No. 2 social app on the Apple App Store, reportedly pays users to record their phone calls and sells data to AI firms.