Sponsored by
Hello readers,
Welcome to the AI For All newsletter! Today, we’re talking about what may be the next frontier in AI models, how AI is rewriting performance management, and more!
AI Decoder: World Models
If you've ever watched an AI video generator make a dog change breed as it runs behind a couch, you've witnessed the core limitation that world models are trying to solve. Most of today's AI systems — including the large language models powering ChatGPT — don't actually maintain a coherent, continuously updated picture of reality. They predict what's statistically most plausible to come next, whether that's the next word in a sentence or the next frame in a video. A world model is different: it's an internal representation of how reality works — tracking space, time, and cause-and-effect — that an AI can update in real time and use to plan ahead.
The practical implications go well beyond cleaner video generation. In robotics, a world model lets a machine predict what will happen in its environment before it acts. In augmented reality, it keeps virtual objects stable and spatially consistent as the user moves through the world. In autonomous vehicles, it enables better prediction of what other drivers and pedestrians are about to do. Researchers are also increasingly convinced that this capability would be a prerequisite for a hypothetical artificial general intelligence — because an AI that can't update its understanding of the world from streaming, real-time input will always hit a ceiling.
The idea is gaining serious commercial momentum. Fei-Fei Li's World Labs recently launched software to generate full 3D worlds from images and text. And Yann LeCun — the Turing Award-winning AI scientist who spent years arguing that LLMs alone can never reach human-level intelligence — just left Meta to co-found AMI Labs, which raised $1.03 billion in seed funding at a $3.5 billion valuation to pursue exactly this approach, backed by NVIDIA, Samsung, Toyota Ventures, and Bezos Expeditions, among others.
This is still early-stage research, and AMI Labs' own CEO acknowledges it could take years before world models move from theory to commercial products. But with that level of capital, that caliber of researchers, and that breadth of industrial backers, it's a space worth watching closely.
🔥 Rapid Fire
Commentary: why are we still doing this?
Block did not lay off workers because of AI — this is AI washing
Block overhired during the pandemic
Headcount grew 2.5x from 4,000 to 10,000
Block layed off 40% of staff to please Wall Street
Stock jumped 20% on news of the layoffs
CEO cited “intelligence tools” in vague shareholder letter
No use of the word AI — is Jack Dorsey just pretentious? yes
Meta signed a $27B compute capacity deal with Nebius
$12B of said capacity is intended to be available in 2027
$12B worth of capacity is around 960MW of IT load
As of Feb 2026, Nebius has only 170MW of active power
Nebius used the contract to raise $3.75B in debt
AI debt is increasingly drawing on retirement and insurance funds
Billions in such funds went into CDOs preceding the 2008 crisis
Meta plans layoffs that could affect 20% or more of the company
Meta needs to free up cash as AI infrastructure costs mount
Meta has burned $100B+ on generative AI
Meta claims it will spend $600B on data centers by 2028
Meta does not have this money
Attributing the layoffs to “AI efficiency gains” is AI washing
CEO alluded to efficiency gains but provided no examples
Amazon laying off 16,000 workers had nothing to do with AI
Analysis: The Hater’s Guide to the SaaSpocalypse
PE firms are misstating the value of their software holdings
Amazon convenes internal meeting following AI-related outages
Amazon now requires two-person reviews of AI-assisted code
AI code creates 70% more issues than human code per CodeRabbit
The Iran war disrupts global helium supply that AI depends on
How LLMs and other kinds of AI are used by the military
Much of it confirms what was listed in this newsletter’s Rapid Fire
A team of Claudes built a C compiler. Here’s why that’s not impressive.
Simple scaling “laws” may be insufficient to achieve reliable autonomy
Tesla’s robotaxis are 4x worse at driving than humans per Tesla
Key findings about how Americans view generative AI
Here’s how I use Attio to run my day.
Attio is the AI CRM with conversational AI built directly into your workspace. Every morning, Ask Attio handles my prep:
Surfaces insights from calls and conversations across my entire CRM
Update records and create tasks without manual entry
Answers questions about deals, accounts, and customer signals that used to take hours to find
All in seconds. No searching, no switching tabs, no manual updates.
Ready to scale faster?
📖 What We’re Reading
A recent State of Performance Enablement report found that a whopping two-thirds of HR managers need more support to effectively manage employee performance.
However, as artificial intelligence continues to disrupt the HR ecosystem we’ve come to know, could GenAI quietly rewrite the future of performance management?
