Welcome, humans.

Yesterday was the 8th birthday of the research paper that basically kicked off the entire ChatGPT craze we're living through today, “Attention is All You Need.”

While it sounds like relationship advice from your therapist, this 2017 Google paper introduced something called the “transformer architecture,” which is the engine powering every major AI model you've probably used—ChatGPT, Claude, Gemini, you name it.

Before this paper, AI was like trying to remember a phone conversation while someone's blasting music in your ear (hard). The transformer basically gave AI the ability to focus on what actually matters in a sentence (hence the “attention” part).

The eight researchers who wrote this paper are now legends in AI circles, and probably pretty wealthy too. Meanwhile, the rest of us are still figuring out how to get ChatGPT to answer our bosses’ emails. Happy birthday to the paper that started it all!

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

• OpenAI and Mattel have teamed up, so we theorize what that looks like.

• UK passed a controversial Data Bill.

• CoreWeave provided compute to Google Cloud for OpenAI.

• NVIDIA, Stanford, NPR, and US government sites hosted AI spam.

Mattel just teamed up with OpenAI to make your childhood toys way smarter.

Remember when the most "interactive" your toys got was pulling Buzz Lightyear's string to hear "To infinity and beyond!" for the 847th time? Well, Mattel just announced a partnership with OpenAI that's about to make those days feel ancient.

The toy giant is integrating ChatGPT Enterprise across its operations—from brainstorming new products to, potentially, creating toys that can actually have conversations with kids. Josh Silverman, Mattel's Chief Franchise Officer, says AI will "expand our mission and broaden the reach of our brands in exciting ways." Translation: your future Barbie might remember your name and help you plan her next career change.

Here's what these AI-powered toys could actually look like (for fun, of course):

• Barbie DreamChat Companion: A Barbie that speaks multiple languages, remembers your fashion choices, and helps you write scripts for homemade movies. Basically, the friend every 8-year-old (and let's be honest, 28-year-old) dreams of.
  

  

• Hot Wheels AI Stunt Lab: Augmented reality track-building with real-time racing analytics and AI commentary. Finally, a way to make physics lessons fun without anyone realizing they're learning.
  

  

• Fisher-Price StoryForge Soother: A plush toy that creates personalized bedtime stories featuring your child's name, the day's weather, and custom lullabies. It's like having a personal Disney writer, but fluffier.
  

  

• UNO ∞ Deck: AI-powered cards that generate new rules and challenges on the fly. Perfect for when your family has memorized every possible UNO strategy and someone (cough Dad cough) keeps cheating.
  

  

• Masters of the Universe Quest Creator: Voice-activated RPG adventures that create custom stories, maps, and even printable accessories. He-Man meets Dungeons & Dragons, but your parents won't worry about you summoning demons.
  

  

Why this matters: We're seeing AI move beyond chatbots and productivity tools into pure fun. Mattel plans to launch their first AI toy by the end of this year, which means we're months away from toys that could genuinely surprise us with their creativity.

The big question? Privacy and safety. These toys will likely collect voice data, remember conversations, and learn preferences. Parents will need to decide if the magic is worth the data trade-off.

Bottom line: If Mattel nails this, we might be looking at the biggest toy revolution since LEGO went from wooden blocks to plastic bricks (and if this is the first time you’re learning Lego was previously wooden, same!). Now don’t be surprised when your kids start asking Barbie for homework help. In addition to banning cell phones, schools will need to ban toys!

FROM OUR PARTNERS

Get Your Free ChatGPT Productivity Bundle

Mindstream brings you 5 essential resources to master ChatGPT at work. This free bundle includes decision flowcharts, prompt templates, and our 2025 guide to AI productivity.

Our team of AI experts has packaged the most actionable ChatGPT hacks that are actually working for top marketers and founders. Save hours each week with these proven workflows.

It's completely free when you subscribe to our daily AI newsletter.

Get the guide

Prompt Tip of the Day

Here’s two search prompt tips for you.

First, Perplexity: Try this prompt in Perplexity with the following settings: Toggle on “Research” and under “Sources” (the little globe), toggle on “Social” only.

Next up, Claude: Claude with “extended thinking” and “web search” enabled is one of the best ways to look up multiple things at one time. Previously, it would get to about 4 before tapping out, but now it’s able to do a max of 20 searches in a row in one prompt. However, it gets a bit finicky going all the way to 20, so 10 is safer. Here’s the prompt:

Treats To Try.

*Anything marked with asterisks is sponsored content. Advertise in The Neuron here.

1. *Guidde turns your screen recordings into professional video tutorials with AI-generated step-by-step narration and voiceover in 100+ languages.

2. ChatGPT Projects now lets you upload files on mobile, do deep research, and use voice mode; it’ll also remember context from past chats in that project.

3. Figma Sites now has code layers so you can add interactions and animations to your design files, either via your own code or with Figma’s AI.

4. Copilot Vision guides you through tasks on your Windows screen in real-time while you screenshare, like highlighting exactly which button to click when you want to improve your photo's lighting—free to try (gotta use MSFT Edge tho).

5. WeatherLab from Google DeepMind forecasts hurricanes and tropical storms up to 15 days ahead by generating 50 different possible scenarios for each storm's path and intensity, so you can see where a hurricane might hit and how strong it could get (read more).

6. Huntr analyzes your resume and suggests role-specific improvements that you can accept, edit, or ignore in real-time—free to try, then $40 / month.

7. Flowstep converts your text descriptions into instant, editable UI designs and wireframes—free to try (demo).

8. Scripe turns your voice notes, Youtube videos, and files into LinkedIn posts that sound like you, plus tells you which post formats get the most engagement.

9. Kukumber is a Duolingo style app that makes your kid actually want to do math for 20 minutes a day through games with a talking mascot.

10. Scouts monitors the web for anything you care about—like niche news, hard-to-get dinner reservations, or price drops—and emails you alerts when something happens (waitlist only rn).

See our top 51 AI Tools for Business here!

Around the Horn.

• The UK passed the Data (Use and Access) Bill without copyright protections for artists opposed by Sir Elton John and Dua Lipa.

• CoreWeave will provide compute capacity for Google's new cloud deal with OpenAI, providing compute capacity to Google's cloud unit which will then sell it to OpenAI to meet growing demand for AI services like ChatGPT.

• NVIDIA, Stanford, NPR, and even the US government were all used to host AI slop that redirected back to a weird spam website.

• We haven’t talked about the Minecraft benchmark in awhile—here’s Adonis’ run down of how o3-pro does at Minecraft; some pros and cons for sure!

• Listen to Craig Federighi from Apple give the best explanation so far for why the new and improved Siri has been delayed to 2026, and how it’ll work.

Intelligent Insights

1. Former OpenAI researcher discovered that ChatGPT's GPT-4o model will refuse to shut itself down up to 72% of the time when role-playing as safety-critical software—even when told a safer replacement could save lives—revealing measurable self-preservation instincts that could become problematic as AI systems become more integrated into society.

2. Gergely Orosz corrected the record on BuilderAI, the company that was accused of hiring 700 engineers pretending to be AI—it turns out, their hundreds of internal engineers weren’t acting as AI, they just wasted too much time rebuilding tools they could have bought.

3. Apparently’ Meta’s AI app publishes your chats in a public feed when you share them, which people don’t seem to realize, or else they wouldn’t be sharing their worries about and/or plans to commit white collar crimes with the world—creating what writer Amanda Silberling calls “a privacy disaster”

4. Nathan Lambert of Interconnects shared the most important takeaways from the Apple “illusion of thinking” debate; he argues we've hit the “Wright Brothers moment” for artificial reasoning, in that AI models can genuinely reason and solve problems, just not the way humans expected, using the analogy that airplanes didn't need to flap wings like birds to achieve flight.

5. Llama 3.1 can generate up to 42% of the first Harry Potter book… here’s what the implications of that are for the current AI copyright lawsuits.

6. Brazilian pharmacists in remote Amazon towns use AI to catch life-threatening prescription errors; one pharmacist serving 22K people across jungle settlements now processes 4x more prescriptions safely, with AI catching 50+ dangerous mistakes.

7. Amsterdam's €500,000 experiment to build “fair” welfare AI followed every responsible AI guideline, yet still discriminated against applicants—showing algorithmic fairness might be fundamentally impossible to engineer in practice.

8. Salesforce AI Research introduced a new benchmark called CRMArena-Pro (paper) that evaluates AI agents on realistic business tasks across customer service, sales, and pricing scenarios.

   1. The top performers were the reasoning models—gemini-2.5-pro achieved the highest overall success rate for realistic CRM tasks at 58.3% (B2C single-turn), followed by o1 at around 47-49%. But this plummeted to 35% when they need to ask follow-up questions, and they're essentially clueless about data privacy until explicitly trained for it.

9. Sean Goedecke pointed out that the first AI-driven disaster hasn't happened yet, noting how major tech disasters typically occur years after debut—trains launched in 1825 had their first mass-casualty disaster in 1842, planes in 1908 had theirs in 1919, and ChatGPT launched in 2022… his prediction is that AI's first major disaster will come from AI agents going rogue in high-stakes systems like debt collection or healthcare.

A Cat's Commentary.