| | Welcome, humans. | Want to see the most gangster clip you’ll see today? It’s a video of Warren McCulloch, one of the O.G. creators of the neural network (which are the brains behind AI), trying to explain the hardest part of AI science while waving around a lit cig and straight chillin’ like a villain in an attempt to stay cool mid-interview. | As one commenter said, “no shirt, no shoes, just neuroscience.” | In all seriousness, the clip hinges on what McCulloch says is the one question that defines “the whole” of his “scientific life”: | “What is a number that a man may know it, and a man that he may know a number? I think I’m fairly clear on the first half of it, but as the second half… not yet." | | | | Warren McCulloch |
|
| Basically, he's saying: we understand math, but we don't understand how we understand math. How does a pile of neurons suddenly get abstract concepts like numbers? Here’s the full interview, in case you’re curious. | 70 years later, we still don’t know. And we’re making “AI brains” we don’t understand as a result. Despite AI companies releasing what they call “thinking” models, many researchers believe that until we solve how understanding itself works… we won’t be able to develop actually intelligent machines. | Here’s what you need to know about AI today: | Menlo Ventures revealed how real people are using AI. Meta poached four researchers from OpenAI. Uber launched robotaxis and robot delivery in Atlanta. Google released new AI models for music, robots, and your phone.
| | | Menlo Ventures Just Revealed How 2 Billion People Actually Use AI… And the Results Are Wild | Menlo Ventures just released the most comprehensive look yet at how everyday people are using AI, and the findings will make you rethink everything you thought you knew about AI adoption. | Here's the headline number: 61% of Americans have used AI in the past six months. To us, that is a shockingly high number. That's not just tech bros and Silicon Valley types… that's your neighbor, your mom, and probably your accountant. | Scaled globally, we're talking about 1.8B people who've tried AI, with 600M using it every single day. | Now here's where it gets interesting: | Despite this massive adoption, people are barely paying for AI. We're talking about a $12B market where only 3% of users actually subscribe to premium services. That's like having a stadium full of people where only the folks in the front row bought tickets.
| | The most surprising findings: | First, parents are the AI power users. They're nearly twice as likely to use AI daily compared to non-parents (29% vs 15%). | From planning scavenger hunts to creating packing lists for family trips, parents are turning AI into their personal assistant for managing family chaos. | Millennials, not Gen Z, are the daily users. While Gen Z leads overall adoption, 24% of Millennials use AI every single day—higher than any other generation. Boomers are quietly adopting AI. Nearly half (45%) of Baby Boomers have used AI, with 11% using it daily. So much for the "technology resistant" stereotype. Students are all-in. 85% of adult students use AI—the highest adoption rate of any group surveyed.
| The data also reveals a massive gap between what people do and where they use AI. While 82% of people pay bills, only 16% use AI to help. 71% research health questions, but just 20% turn to AI. Menlo thinks these are goldmines waiting for the right specialized tools. | Why this matters: We're witnessing the fastest technology adoption in history, but we're still in the “free trial” phase. Bay area residents remember the golden days of mobile growth when you could get every “Uber for X” service under the sun imaginable (laundry, taxi, personal chefs, even someone to park your car for you), with deep discounts to try the services. That’s where we are in AI right now. | So, the companies that figure out how to convert casual users into paying customers in these high-trust, high-frequency areas will capture the next wave of consumer AI revenue. | Right now, professional coding tool startups seem to be the most sticky. That’s because they’re probably the most useful. Once other assistants and agents get better at their jobs, we could see a similar trend. The report predicts AI will soon move from handling individual tasks to running entire workflows (think booking a complete vacation with one prompt, not just finding flights). As far as how long that will take? TBD, but… | | FROM OUR PARTNERS | Helium 10: The All-in-One E-commerce Seller Solution - Now with TikTok Shop + AI tools | | Helium 10 is the #1 all-in-one suite of e-commerce tools designed to help sellers start, grow, and scale their businesses across Amazon, Walmart, and TikTok Shop. Get the data, automation, and insights you need to succeed. | 4M+ e-commerce sellers use Helium 10 to scale smarter. Top tools include: | Black Box: Discover high-demand, low-competition products. Helium 10 Ads: Ai-powered ad optimization. Profitability Calculator: Track and manage real margins. Now with TikTok Shop Listing Converter: Easily expand to one of the fastest-growing marketplaces.
| Use Helium 10 alongside tools like ChatGPT to uncover product trends, and launch with confidence. Get Your Helium 10 Exclusive Discount Here | | Prompt Tip of the Day |  | AI Agent Insights From 72 Hours of Expert Panels |
|
| Tina Huang shared a TON of valuable insights in this video on the skills you need to learn to build AI agents, including why learning “prompt engineering” is still one of the highest leverage skills you can learn, what type of AI company to develop (vertical AI agents), her own personal prompt framework, and how to think about AI prompting for developing AI agents (she also suggests watching this video). | The other tip she recommends you learn besides prompting? Writing evals, or tests to assess if the prompts you use actually work. | She has a great breakdown on the common types of evals you might create, and why evals are actually the most important “IP” AI companies have… | | Treats To Try. | *Anything marked with asterisks is sponsored content. Advertise in The Neuron here. |  | PageAI Launch Video |
|
| *PageAI: The AI website builder that designs, writes copy and generates production code you can download or deploy with 1 click! Including a blog, SEO and 100s of features! Offerwall by Google lets publishers use micropayments to allow their web visitors choose how to access content: watch an ad, pay a fee, or complete a survey instead of hitting a paywall (this is low key genius NGL, but Google needs some competition here doing the same thing. Break up the ad duopoly!) NativeMind brings AI directly to your browser while keeping everything private and on-device through a Chrome extension. Higgsfield Soul is a “high aesthetic” image model (basically it makes high quality / artsy photos. Related: you can now run Flux1.Kontext’s AI image editor on your own computer (with an NVIDA graphics card with 16-20GB of VRAM on it).
Doppl is Google’s new virtual try on tool, where you upload a picture and see how outfits look on you. Middleware OpsAI automatically detects and fixes production bugs before they impact users by generating pull requests. Not to totally make this whole TTT about Google, but… it also launched: Magenta, a music AI model that lets you create and control live music in real-time (code). Gemini Robotics On-Device, an AI that runs locally on your robot without internet access and follows your spoken commands to fold clothes, zip bags, and handle objects. Gemma 3n, a phone AI model that understands your voice, images, and text locally on your phone without sending your data anywhere, which you can try for free on AI Studio or download here.
OpenAI now lets developers use Deep Research through the API (so you can make apps that use Deep Research in them) and also web hooks (read more).
| See our top 51 AI Tools for Business here! | | Around the Horn. | Meta finally hired four of OpenAI’s “best” engineers: Trapit Bansal, who was one of the main contributors to OpenAI’s original thinking model, “o1”…, as well as Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai (that’s in addition to WhatsApping 100+ others). CoreWeave, the newly public AI data center company, saw its stock boom ~300% since its IPO, making its CEO’s net-worth go boom too (~$1.5B to ~$10B)… and that’s despite some shaky business model fundamentals. Uber launched both robotaxis and robot food delivery in Atlanta, and may even help former CEO Travis Kalanick buy Chinese robotaxi startup Pony AI. Palantir invested $100M in The Nuclear Company to develop software for, you guessed it, nuclear companies.
| | Intelligent Insights | | Check out more insights we found this week (and more from this entire month) on the website here. | | |
| | | | | | | What'd you think of today's email? | |
| |
|
|
|
|
|
