TWS #013: AI's Cognitive Debt Crisis; $1T Robots; Robotaxi Launch; Amazon's AI Jobs Announcement; Mental Models
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...
Coca-Cola's famous bottle was designed to be recognizable in the dark and when smashed to pieces
In the early 1900s, the company had issues with "copycat" brands using identical straight-sided bottles and also threatening their sales. Then in 1915, Coca-Cola challenged glass companies to create a bottle so unique it could be recognized by touch in the dark or even if shattered on the ground. The winning design from the Root Glass Company was famously inspired by the elongated, ribbed shape of a cocoa pod. The "contour" shape made Coca-Cola instantly recognizable, eliminating the problem of imitators and creating a powerful, ownable brand asset. The bottle became a global symbol of the brand, so successful that it was officially recognized as a trademark in its own right in 1961, a rare achievement for product packaging.
Here’s this week’s scoop:
MIT’s paper on AI's Cognitive Debt Crisis
Tesla launches its Robotaxi service
Softbank is looking to invest $1T to build robots
Amazon’s AI Job announcement (warning)
Namibia wants to build the world’ first hydrogen economy
Let’s dive in.
📖 Tactical Guide: Mental Models for Aspiring Leaders
As part of our Tactical Guide series, we’ve started to import some selected episodes from the podcast that were only previously available on YouTube, Spotify, and Apple Podcasts. A number of our subscribers noticed this was the case and suggested I take some action. I’m glad I did.
Inside each guide, you’ll find a link to the original video post, but also an uploaded audio version. It also includes a written in-depth summary of the conversation in Q&A format—helps if you’re a better reader than listener. This has been in high demand for a while.
//
This conversation is with Jennifer is the CEO of Cultivating Leadership. She blends deep theoretical knowledge with a driving quest for practical ways to make leaders’ lives better. She speaks about leadership and complexity, coaches executives and executive teams, and designs and teaches leadership programs.
In this episode, expect to learn the mental models of highly effective leaders, why imposter syndrome is a good thing, the importance of leadership in a complex world, why adult development theory is becoming more popular, how to deal with uncertainty, and much more.
Since its release, it’s been super helpful for founders I’ve shared this with and hope it can help you too.
🔥 Nuggets for the Road
Coatue’s 2025 East Meets West (EMW) Keynote Deck (thanks Sunil) — a comprehensive deep dive into the public and private markets and the current state of AI (will dig deeper into this next week) [LINK]
Software in the era of AI by Andrej Karpathy (Co-founder of OpenAI) — presentation given at YC’s AI Startup School. One key revelation here is that LLMs have similar properties to those of energy utilities. See image below. [YCombinator: YOUTUBE]
Inside an LLM — one of the better explanations of how LLMs work under the hood. [YT: 3BLUE1BROWN]
China’s Evolving Industrial Policy for AI — looks like Beijing is deploying industrial policy tools across the full AI stack, from chips to applications [LINK]
The “Product Onion” Framework Saved our Startup (When Everything Else Failed) by — great insights on how to grow your startup by looking at its core components as layers of an onion. [LINK]
Every marketing channel sucks right now (a16z Partner: ) — an adulterated look at the current state of marketing channels and what you can do to stand out [LINK]
📡 The Signal
MIT Study: Does AI lead to cognitive debt?
This one has been getting a lot of attention and rightly so. Basically, the paper explores the cognitive costs of using LLMs like ChatGPT for essay writing, particularly in educational settings. The study recruited 54 participants from several Boston-area universities and divided them into three groups: one using ChatGPT (LLM group), one using traditional web search (Search Engine group), and one using only their own knowledge (Brain-only group). Over multiple sessions, participants wrote essays while their brain activity was monitored using EEG, and their essays were analyzed and scored by both human teachers and an AI judge.
It’s a lengthy document, so here are just a few key takeaways on what I found interesting:
Search Engine users showed intermediate brain engagement and essay originality, better than LLM, but not as strong as Brain-only. Their essays reflected more external information integration (visual-executive brain networks).
Heavy reliance on LLMs may undermine deep learning, memory formation, and critical thinking, even if it increases short-term productivity and ease.
When LLM users were asked to write without AI, their brain connectivity and originality did not fully recover to Brain-only levels, suggesting lasting cognitive “debt” from prior AI reliance.
Human teachers rated LLM essays as less unique and “soulless,” despite high scores for grammar and structure. AI judges tended to rate all essays higher and less discriminately.
Participants who switched from LLMs to brain-only writing showed weaker neural connectivity and continued to struggle with memory and originality, indicating possible long-term effects of cognitive offloading to AI
Figure 6: LLM users struggled to recall or quote from their own essays (83% failed to quote), while Brain-only and Search Engine users performed much better.
Figure 8: Participants felt less ownership over essays written with LLMs. Brain-only writers almost always claimed full ownership, while LLM users often reported partial or no ownership.
SoftBank wants to invest $1T in Arizona to build robots
Masa Son is looking to pitch a $1T industrial complex in Arizona, called “Project Crystal Land”. The plan is to create a US-based hub for AI semiconductors and robotics manufacturing, rivaling China’s Shenzhen. Key partners would include TSMC and others like Samsung. The vision is to have a sprawling ecosystem with R&D labs, housing, and smart grid systems. This is a huge endeavor that will require federal and state support. Obviously, this is one of those moonshot projects, and some are calling it a pipe dream. But who knows, this could actually work. Masa Son’s reputation for making big, bold bets in Alibaba, WeWork, and Uber shouldn’t be underestimated. You should also read this piece on why Arizona is becoming the next tech hub. [BLOOMBERG]
Amazon’s AI announcement
Another week, and another company makes an announcement about AI and jobs. Amazon announced that GenAI will reduce the company’s workforce over the next few years as it becomes more efficient. Apparently, they have 1,000+ AI services in development aimed at streamlining tasks. They emphasized AI as a “once-in-a-lifetime” opportunity, urging employees to embrace it through training and experimentation to stay relevant. The fact that a company like Amazon is willing to put out something like this clearly shows there is a growing trend (and even warning) on what’s on the horizon. I don’t think this is a bad thing, but simply a shift and something we all have to adjust, at least for knowledge workers. You can read more on this in my essay: AI and the Great Adaptation. [AMAZON]
Tesla launches its Robotaxi service
Tesla finally launched its long-awaited robotaxis in Austin, Texas, on June 22. There were around 10-20 driverless 2025 Model Y cars equipped with an “unsupervised” version of its Full-Self Driving (FSD) software. Right now, the cars are geofenced around the South Austin area, and operate between 6a to midnight daily. It charges a flat $4.20 per ride but also includes a Tesla employee in the passenger seat for safety in the short term. This will be an interesting test for Tesla (camera only) as it seeks to compete directly with Waymo (camera + LiDAR). Right now, the robotaxis suspend operation in bad weather and avoid complex intersections. I’m an optimist, so seeing these milestones achieved is always promising. The ultimate test is whether, in the long term, self-driving will be better (not the same) as human drivers. [LINK]
Namibia wants to build the world’s first Hydrogen economy
Namibia has around 60% of the world’s best solar sites with minimal overcast skies, especially across their Namib Desert. This means a lot of clean tech companies are coming in to take stock. Apparently, it’s attractive for those looking to produce hydrogen-based products like ammonia and low-carbon jet fuel. Also, the country’s stable political landscape, huge land, and government initiatives make it very appealing. [LINK]
🧐 Poll of the Week
This week’s poll:
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.
Great issue, Barry, thanks!