😸 Microsoft picks Claude over GPT

PLUS: OpenAI officially goes Hollywood

Welcome, humans.

For a fun sign of the times: OpenAI just dropped $30M backing “Critterz,” a feature-length animated movie set to debut at Cannes in May 2026.

We have SO many questions:

What’s up with that budget? Isn’t AI supposed to save companies money? Well, $30M sounds massive until you realize Disney's “Tangled” cost $200M and took three years to make. OpenAI claims they'll finish this in nine months.

So what’s the money go to? Compute, obviously, but not just that: human voice actors, writers, production crew, marketing, licensing, and distribution. Plus, they're hiring artists to draw sketches that get fed into GPT-5 and DALL-E, so it's not purely AI-generated. But TBH, this type of hybrid workflow makes the most sense for creative work like this.

So is this a genuine attempt to revolutionize animation, or just an incredibly expensive marketing campaign for OpenAI's tools? Chad Nelson, OpenAI's creative specialist behind the project, basically admitted it: “OpenAI can say what its tools do all day long, but it's much more impactful if someone does it.”

The real test? Whether Cannes audiences (notoriously tough on anything that smells like corporate tech posturing or anti-human art) will embrace woodland creatures made by algorithms. The original 2023 Critterz short has a brutal 3.2/10 on IMDB, so they're starting from... let's call it humble beginnings…

Here’s what happened in AI today:

Microsoft chose Claude over OpenAI for certain Office functions.
Claude launched direct file creation and editing in chat.
Mistral AI raised $2B at $13B valuation led by ASML.
NVIDIA announced specialized million-token AI chips for 2026.

Advertise in The Neuron here

P.S: Reminder that we are going LIVE tomorrow at 10am to compare text to speech, speech to text, and speech to speech models on Corey’s laptop! Click “Notify me” to get a heads up when we’re live.

The Claude Office Studio is here… in fact, it’s so good, Microsoft is letting Claude into Copilot

Preview: Claude can create and edit files

Is it just us, or were you fellow “knowledge workers” out there starting to feel a little bit abandoned lately, AI tools wise?

See, while the major labs have been favoring all the shiny new coding tools for devs, the rest of us have been stuck copying and pasting between AI chat windows and the actual files / services we need to use for our jobs.

Example: Want to tweak one word in a Claude Artifact? Too bad—you can only highlight text and click a button for it to “improve this.” Need to edit in ChatGPT's Canvas? Hit that tiny character limit. Want a simple in line hyperlink? Gotta write HTML like it's 1999.

Today, that changes (for Claude, anyway). Claude can now create and edit Excel spreadsheets, documents, PowerPoint slide decks, and PDFs directly inside Claude.ai (web and desktop).

What Claude can actually do:

Turn data into insights: Feed it raw data, get back cleaned spreadsheets with analysis, charts, and written insights.
Build spreadsheets: Financial models, project trackers with dashboards, budget templates with working formulas.
Cross-format work: Turn a PDF report into PowerPoint slides. Meeting notes > formatted doc.

Now, this file creation feature is only available as a preview for Max, Team, and Enterprise plan users atm, but Pro users ($20 tier) will get access in the coming weeks.

Simon Willison says Anthropic is “burying the lede”: they called this “Upgraded file creation and analysis,” when it's actually a Code Interpreter. What's that mean?

Claude now has its own computer where it can write and run actual programs to create real files.
Think of it like this: before, Claude could only tell you how to make a spreadsheet.
Now, it more or less has Excel installed on the backend and can make it for you.

How to turn it on: Head to Settings > Features > Experimental > Enable “Upgraded file creation and analysis.” Then describe what you need. Claude handles the rest.

Oh, remember ChatGPT's Agent? It also launched recently with similar capabilities, setting the stage to turn ChatGPT into a super app for work. But you still can’t edit files directly inside it (download required).

Our guess? There's probably a businessy reason for this (Microsoft's billions in investment? Don’t kill your Copilot?) rather than technical limitation, but it’s annoying for users who want the most fluid, useful tool possible where they work.

Well, surprise surprise: according to The Information, Microsoft will now use Anthropic for certain Office tools now because they reportedly “perform better than OpenAI's” at finance functions and presentations. So, if you’re one of the ~100M or so Copilot users, sometime in the next few weeks, you’ll be getting a big upgrade to your office suite! Thanks, Anthropic!

Our take: This solves the exact problem we've been complaining about lately: we’re two years deep into AI innovation, and we’re still stuck copy-pasting between tools. Direct editing docs within Claude would be the holy grail: no more digital gymnastics between AI and actual work tools.

We'll be putting Claude's file creation through its paces all day today, turning meeting notes into slide decks, cleaning messy spreadsheets, and seeing if it can handle our newsletter workflow. We’ll report back on Thursday during the live how we like it!

FROM OUR PARTNERS

The Future of Compliance: Faster, Smarter, Founder-First.

Getting compliant doesn’t have to take months. With Delve, it takes just 15 hours.

SOC 2, HIPAA, ISO, and more—automated with AI.

That’s why 500+ of the fastest-growing AI companies, including Lovable, Cluely, and Wispr Flow, already run compliance on Delve.

But compliance is only half the battle. The other half is enterprise reviews. We’ve sat through hundreds of hours of security questionnaires—every pointless checkbox, every SaaS-era stall.

So today, we’re publishing something different: a manifesto with Bland AI on how security questionnaires should work in the age of AI. 📖 Read the manifesto here.

⚡ Book a demo here with code NEURON1KOFF for $1,000 off + white-glove onboarding.

Prompt Tip of the Day.

Hate typing prompts? Jeff Su has a FANTASTIC tip to save you a ton of time “prompting.” Check out this 51 second clip where he breaks it down.

In the same way you can use projects, gems, or GPTs to save time typing, you can use what’s called “text expander apps” (think autocorrect, but you control what gets replaced) as shortcuts for prompt templates.

Here's the workflow: Jeff types “::copywriter”, which expands to “Assume the role of a senior copywriter with over 20 years of experience. I'm writing [TOPIC] and need 3 variations of [SCRIPT].” Then all you have to do is fill in the brackets, and you’re good.

Here’s the tools Jeff recommends: he uses Alfred (Mac, paid), but you can use Raycast + Snippets (Mac, free), or Beeftext (Windows, free).

Another alternative? You can use voice with voice mode, either via the native apps themselves or a tool like Wispr or Willow.

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.
AUI is really awesome; their model Apollo-1 actually enables “agentic shopping” to reliably book flights and process refunds; this is possible using neuro-symbolic reasoning, which combines neural networks with logic-based rules so you get predictable, traceable decisions instead of unpredictable responses like plain old language models.
Seedream 4.0 generates native 4K images and can create up to 9 coherent story images in one batch—perfect for making sequential comics or storyboards, which you can use on sites like Replicate.
Gemini Canvas now lets you edit any part of your web app by clicking and describing the changes you want to edit it instantly.
In Lovable, did you know you can now vibe code your own apps with your voice? Just click the voice mode and chat with your app; wild!

Around the Horn.

NVIDIA announced Rubin CPX, a specialized GPU (NVIDIA’s flagship AI chip) designed specifically for processing million-token context AI workloads (~850K words) like software development and video generation, slated to come out end of 2026—this is a big deal for agents using reasoning and tools in a loop, too.
Mistral AI raised ~$2B at a ~$13B post‑money valuation led by ASML, the dutch company that makes the high tech materials for making AI chips.
OpenAI’s restructure to a for profit faced specific pushback from California regulators, and apparently OpenAI execs even considered leaving CA over it.
Anthropic endorsed SB 53, backing California’s frontier‑model safety bill.
Sam Altman laid out his latest AI predictions and roadmap for OpenAI in an hour long talk with one of his investors, Vinod Khosla—we break it down here.