😺Elon to buy OpenAI?!

PLUS: Two new AI models enter the arena...

Welcome, humans.

This just about sums up the results of yesterday’s poll on OpenAI’s Super Bowl ad:

*$14M, btw

About ~200 of you said yes, OpenAI had its Apple 1984 moment, and 1200+ said no… that’s almost as complete a shut out as the Super Bowl itself!

It’s interesting OpenAI made the choice to not use its own AI, like SORA, to make the ad, instead working with human artists to do it.

On that note, check out how wild the progress in AI animation has gotten over the last two years:

*Here’s the* *actual human dancer* *behind that video, btw.*

‍

Here’s what you need to know about AI today:

We explain Elon’s surprise bid to takeover OpenAI.
Anthropic released a study on the economic impact of AI.
France pledges $112B and nuclear power for AI.
Two new AI, “Chocolate” and “Kiwi”, are in the LM Arena…

Advertise in The Neuron here.

The day Elon Musk tried to buy OpenAI for just shy of $100B…

Things over at xAI headquarters must not be going so smooth, because Elon Musk just submitted a $97.6B bid to purchase their main rival, OpenAI. Grok 3 not living up to your wildest dreams, Elon? Overhauling the government with AI not keeping you busy enough?

Here’s the deets on the deal:

xAI (Elon’s AI company) and a bunch of investment funds made the offer.
The bid is for the non-profit that controls OpenAI.
If it happens, xAI could merge with OpenAI.

In case you forgot, OpenAI is still a non-profit—but it’s trying to become for-profit by 2026 (and raise a new round of $40B ASAP).

Now, OpenAI’s board must consider this offer as it attempts to value itself—and Elon’s crew is prepared to match or beat any counter offers. We didn’t go to business school, but we watched A LOT of Succession… is this the dreaded “bear hug”? Feels like it!

Quick history lesson—for those of you who never read The Founders, this move is ripped from the PayPal Mafia playbook:

When Elon’s original X.com company and Peter Thiel’s PayPal were fighting each other for online payment market share, they both lost tons of money. So instead of keep fighting, they merged.

That same exact thing is happening right now, but in AI:

We talked about this on Sunday, but AI is facing a price war driven largely by DeepSeek.
After the big models do all the heavy lifting, smaller AI models can pop up and bring the cost of intelligence waaay down.
Sam wrote about this yesterday, explaining how the cost of AI has dropped by 10x every 12 months (and 150x from 2023 to 2024, specifically).

What Sam didn’t write about is how this price war undercuts the economics of OpenAI’s infrastructure investments, which Sam says will require significant more scale (and matching spend) to train the next level of intelligence.

And for every DeepSeek that pops out and lowers the prices OpenAI can charge, the less ROI Sam (and really, Sam’s investors) get on all the NVIDIA chips they have to buy.

There’s one more thing, too: Not to spoil the ending of The Founders, but y’know what went down after that PayPal / X.com merger happened? Elon was out as CEO—and now, the same thing could happen to Sam. Call it Elon’s ultimate revenge?

Needless to say, Sam wasn’t happy. He (sarcastically) responded on X with a $9.74B offer to buy Twitter—which is about 8x Twitter’s current earnings.

This news comes as two new models (Kiwi and Chocolate) appeared in the LM Arena—where AI go head to head to answer prompts and users rate their favorite:

‍

Many on X are speculating these models could be Grok 3 and Grok 3 Mini—but when asked, the models keep saying they’re made by Claude. Opus 3.5 anybody??

For those of you don’t live and breathe AI news, here’s a quick refresher on Claude:

Besides Gemini 2.0 (Google’s new AI that tops the LM Arena leaderboard atm), Claude by Anthropic is ChatGPT’s main competition, and it comes in three flavors:

Haiku for speed and savings.
Opus for the heavy lifting.
Sonnet for the sweet spot.

Released last June, Claude 3.5 Sonnet quickly became a hit among developers for its enhanced coding and multi-step reasoning abilities—but nine months later, we’re still eagerly awaiting the powerhouse that was promised: Claude 3.5 Opus.

If one of these new models really IS Opus, it’s a huge deal. And it comes at the worst possible time for Grok 3, OpenAI’s o3, and Gemini 2.0.

If it’s not Opus, then it has to be either Grok 3 or a new OpenAI model (full o3?).

Why this all matters: If Elon’s negotiatrolling is successful, the board may have no choice but to accept his offer. Which means OpenAI might become xOpenAI.

Thankfully, OpenAI and ChatGPT are far from the only AI game in town… and between Gemini 2.0 and a potential Claude Opus, even a merged xAI will have its hands full.

FROM OUR PARTNERS

Here’s one AI tool we use every week at The Neuron.

‍

‍

A little well-kept secret for companies crushing it with AI is that only a handful of AI tools are actually worth using.

That’s why we—Noah & Grant—use ChatGPT and Claude, along with a killer product called Attention:

We tell Attention exactly what to pay attention to in our meeting (goals, budget, etc).
Attention listens to our call.
Attention outputs important insights, action items, and follow-up emails.

There are way, way more features, and it’s a must-have for all sales teams.

If you sell anything, literally drop what you’re doing and go book a demo with Attention ASAP.

Prompt Tip of the Day

In this video, Grace Leung shares five key use-cases for AI reasoners (GPT o3 and DeepSeek R1) to think deeper about your marketing data, showing how to:

Map user motivations from raw feedback.
Predict market shifts from product launches.
Spot friction in customer journeys.
Decode competitor messaging gaps.
Test UX against specific audience needs.

Here’s an example of how she structures her prompts:

"Analyze [specific content] and think deeply about:

What are the primary questions customers might have?

What are the trust signals that might be missing?

Where might users get confused?

What are the deeper motivations behind user behaviors?"

Grace Leung

Treats To Try.

Google made NotebookLM Plus available through Google One AI Premium ($20/month) with 5x more usage limits; fun fact—NotebookLM apparently got ~28M visits in the last three months alone.
DeepSeek-VL2 helps you understand and analyze images through text conversations, now accessible through a simple online demo (code).
Mistral, the French AI lab behind the Le Chat chatbot, is pretty good and lightning fast (1,000 tokens a second)—check it out vs Claude and OpenAI.
Furypage builds eye-catching 3D portfolio websites where visitors can view your work from any angle.
Reef analyzes your spreadsheets and explains the insights through charts and voice.
Pikadditions seamlessly blends your photos into any video - drop a picture of yourself into a safari clip in seconds.
Tough Tongue lets you rehearse tough conversations with AI feedback across 30+ languages.
Talo translates your video calls in real-time across 30+ languages so you can speak naturally in any language on Google Meet and MS Teams (demo).

See our top 51 AI Tools for Business here!

Around the Horn.

Anthropic analyzed millions of Claude convos to understand AI's impact on labor markets (paper) and found AI usage was highest in software development and technical writing, with 36% of occupations using AI for at least ~25% of their tasks—expect a deeper dive later this week!
France will invest $112B and one gigawatt of nuclear power to help train more European AI to try and make up lost ground.
Perplexity released a Super Bowl “tweet” and grew its daily downloads by 50% (45K vs the typical 30K).

FROM OUR PARTNERS

Want to build better AI agents?

‍

Galileo just dropped a massive (100-page!) guide all about agents.

Whether you're just starting out or already knee-deep in agent architecture, this guide will help you build more reliable agents that actually do what they're supposed to do (imagine that!)

Get the free ebook here.

Tuesday Ticker

‍

Here are the results from last week’s polls:

All 3 polls went this way.

S.H chose Gemini: “In order to get actual data on fuel efficiency etc., it would have been necessary to ask supplementary questions. Nevertheless, the answer was more succinct, actually did give "best model" suggestions, rather than an overwhelming and tedious mass of information”
C.N chose ChatGPT: “Deep Research seemed to juggle a lot more factors and math out averages, etc. It feels like a stronger answer and was a more pleasant/informative read.”
A.C. chose ChatGPT, but… “I don't have a favorite. I like the tables that Gemini provided. We as humans still need to be in the "driver's seat" on decision making if we outsource our critical and decision making then we might as well not even buy a car but sit on our couch all day on our phones!”