TWS #011: Apple Bashes AI Models, Megawatt EV Charging; Circle's IPO, Practical AI Podcast
and much more...
Join 2,300+ curious minds who, every week, get the latest curated high signals on news, insights, and ideas around technology, science, and business to help you become a better builder and thinker. Please fill out the survey so we can deliver the best content to your inbox. Thanks for reading!
Did you know...
In 1996, Google’s founders; Larry Page and Sergey Brin built their first server rack using LEGO bricks. It was a low-cost frame to house 10x 4GB hard drives to power their early web crawler, called project “Backrub”, a precursor to Google. The goal was to analyze the web’s link strcuture to rank the imporatnce of web pages which eventually became what we famously know as Page Rank.
Here’s this week’s scoop:
EV’s new era of Megawatt Charging has arrived
Apple’s “Illusion of Thinking” research paper
Stablecoin Giant: Circle’s IPO
Global AI Supercomputer Data (pretty charts)
Bain & Co. releases their insurgent D2C brand report
Let’s dive in.
🔥 Nuggets for the Road
The Gentle Singularity — Sam Altman’s essay [LINK]
The 2-week Vacation Test for Engineers and Managers by
—some good advice on how to delegate and share knowledge with your team [LINK]Podcast: Conflict to Connection: How to Have Crucial Conversations that Count [MATT ABRAHAMS]
Going Inside a 3D Printer Factory (eufyMake E1 UV Printer) [YOUTUBE]
How to Think Like Rick Rubin: 20 min crash course in creativity — deck by
[X]
🎧 Podcast: Practical AI
This is the first video in a series of episodes designed to help non-technical individuals become more familiar with and proficient in AI.
In a world where technology is changing so fast, AI is becoming commoditized and transcending the technical community. This reminds me of when access to Cloud infrastructure became democratized thanks to the likes of AWS, Google Cloud, and Microsoft Azure. Anyone could come in and build anything on top of these platforms. But even then, you still required technical skills to be able to build and deploy your app. There was a big technical barrier. However, this time it’s different. The barrier has collapsed, which has opened the floodgates to a larger non-technical audience that now have access — this is monumental.
Everyone from product managers and operations to social media marketers is now using AI in some shape or form, and it’s becoming commonplace for knowledge workers to engage with a variety of AI tools.
But it’s still early, and these videos are an attempt to further broaden the scope of learning for non-engineers and non-developers.
Enjoy.
Listen / Watch the full episode here 👇
📡 The Signal
The Era of “Megawatt Charging” Has Arrived
Chinese EV top firm BYD has revealed a new 1000-kilowatt (kW) charger. It can add 250 miles (400 km) of range in just 5 minutes, which in some cases is on par with typical gas-powered vehicles. BYD’s big growth aims to establish 15,000 chargers in China, with Europe being next. This marks a global move to kill charge fears and usher in a new era of Megawatt chargers. The barriers to buying and owning an EV are slowly collapsing. Previously, it was all about range (that’s been solved), and now it’s about charging times. This is what happens when a company like BYD is allowed to vertically integrate to own and control its entire supply chain — you can move more efficiently, but also make executive decisions on how to build your battery from the ground up! [IEEE]
“If I can add 250 miles in five minutes, what does this mean for oil and gas?” says Tu Le, managing director of Sino Auto Insights. “And if China is the only country that’s going to support this, are the rest of us going to live in an analog world?…BYD reduced the battery’s internal resistance by 50 percent, redesigning electrolytes, separators, and electrodes for “ultrafast ion channels.” Its latest refrigerant cooling system helps deliver a 35 percent gain in high-temperature lifespan, ensuring that megawatt charging won’t degrade the battery.”
Apple’s “Illusion of Thinking” Exposes AI’s Reasoning Limits
Apple just released a research paper — "The Illusion of Thinking”. It gives a look at Large Reasoning Models (LRMs) like OpenAI’s o1, DeepSeek-R1, and Claude 3.7 Sonnet Thinking, showing there is some fragility in their complex solving ability when pushed to their limits. They did this using controlled puzzles (Tower of Hanoi and others). It’s a hefty doc, but the TLDR on this is that LRMs do outperform standard LLMs on medium-complexity tasks but collapse entirely beyond the “critical complexity threshold”, dropping to 0 (see graph below). Basically, their reasoning peaks at moderate complexity but drops on harder problems, suggesting “overthinking” could be taking place. This is all still pretty early, but there might be some validity to this. You can look at this two ways: Apple is so far behind the AI game, they’re making excuses and coming up with ways to justify their lagging position, OR they’re genuinely cautious not to invest heavily in AI (burning resources) because they’re not yet convinced about the ROI to be gained from potentially building their own LRM. [APPLE]
Circle’s IPO: The Stablecoin Giant Goes Public
The creator of the popular USDC stablecoin, Circle, has finally gone public, which was oversubscribed. USDC is the second-largest stablecoin (behind USDT) and is becoming more widely accepted, and pending US legislation could legitimize stablecoins even further. You should read this primer on stablecoins to give you a sense of how it works. In a previous TWS edition, I mentioned that Stripe is doubling down in the stablecoin space with their purchase of Bridge to help with global payments. There are a lot of benefits to using stablecoins over fiat currencies (like USD, Euro, Pound). Visa is also getting into the game. [LINK]
Meta goes Nuclear
Meta just signed a 20-year deal with Constellation Energy to secure nuclear power starting in 2027. It marks a huge move to fuel its AI-driven data centers with clean and reliable energy. Although Meta says it’s doing this to meet their net-zero goals, I think it’s a broader move to diversify their reliance on traditional sources of energy (coal) and find cheaper and more abundant ways to power their infrastructure. It’s probably why they’re hedging themselves against nuclear by also investing in natural gas-powered data centers. The paradox here is that there’s a race to harness nuclear’s steady, low-carbon power to sustain AI’s growth, but the slow pace of nuclear development, often a decade or more, clashes with AI’s fast growth rate. [LINK]
Global AI Supercomputer Data
This is a neat dataset from EpochAI on over 500 AI global supercomputers. They’re often called clusters, with xAI taking the lead (so far) with 200,000 chips in their Memphis datacenter. Here’s just a quick summary of the findings, but feel free to look at the visuals. [EPOCH AI]
2.5x annual growth in computational power since 2019, driven by a 1.6x yearly increase in chip quantity and performance per chip
Private companies now dominate, owning 80% of AI supercomputing power
U.S. hosting 75% of global capacity and China at 15% (this could be higher!)
Power demands are soaring, with top systems requiring hundreds of megawatts, equivalent to mid-sized cities, raising concerns about energy constraints
Bain & Co: The Explosion of Insurgent Brands
The branding game is changing for D2C. There has been an explosion of insurgent brands disrupting the incumbents. Small, agile brands are now outpacing legacy giants by building strong customer intimacy and digital channels, proving that there is public demand to explore and experiment with lesser-known companies. Buyers are now looking towards addressing their unmet needs with authentic founder-led stories. This means that founders are putting more of themselves out there on the socials, showcasing their personality rather than letting the brand speak for itself. It’s a powerful differentiating factor. [BAIN]
A proven playbook
Regardless of category dynamics, winning insurgent brands achieve long-term sustainable growth by applying a unique growth playbook focused on a few key levers:
Unlocking incremental growth by addressing unmet consumer needs in a new, authentic, and often founder-led way, thereby bringing new consumers into the category or increasing category attractiveness through premium offerings.
Sustaining velocity growth while expanding distribution by engaging deeply with core audiences and partnering closely with the trade to create awareness and conversion while also ensuring that distribution does not get ahead of velocity.
Keeping complexity out by focusing on hero SKUs and core assortments that deliver the greatest value for consumers and avoiding unnecessary proliferation.
Maintaining a Founder’s Mentality® by embedding an insurgent mindset throughout the organization, testing and learning along the way, and remaining ruthlessly consumer centric.
🧐 Question of the Week
Last week’s question: What starts at 20 and turns into 32?
Thanks to Andrew S. for being the first to answer:
“I think it’s teeth. 20 baby teeth by the age of about 2-3, which are then eventually replaced by 32 adult teeth.”
This week’s riddle:
Around 30 minutes before bedtime, Naomi spreads a light film of cooking oil onto a plate. Why?
You can reply to this email with your answer. The first one to answer correctly will be mentioned (first name only) in next week’s newsletter. Good luck!
If you liked this, here are a few more popular posts:
🙋🏻♂️ Survey
To build the best newsletter for you all, I’d love to get some feedback. If you have some time, please fill out the following survey.
🤩 Deals, Discounts & Affiliates
You can find carefully selected offers and deals here.
🙏🏼 Connect with Me
Are you new to the newsletter? Subscribe Here
Check out my YouTube channel (and subscribe!)
If you’re a founder, apply here (Metagrove Ventures) for startup funding or contact me directly at barry@metagrove.vc
If you think this could be helpful and informative to others, please share it :)
Thanks for reading, and see you next week!
Barry.