Monday, September 14, 2026 · 6 min read
Together with
Over the weekend, the people building the most powerful AI on earth started asking each other to slow down. Meerkat had to read that twice.
Anthropic's Dario Amodei made the case, Sam Altman and Elon Musk backed him within hours, and the argument over whether they mean it began just as fast.
Let Meerkat explain.
Why the AI bosses suddenly want a speed limit
The robots already taking desks, and the AI excuse doing a lot of work
How big the AI money machine has actually got
A near-certain Fed hike on Wednesday
SCOREBOARD
NVIDIA | MICROSOFT | ALPHABET |
META | AMAZON | ORACLE |
S&P 500 | NASDAQ | VIX |
FED, Wed 16 Sep: about 86% a quarter-point hike, 14% a hold. CME FedWatch, as of 12 Sep. August inflation at 3.4% has all but settled it.
Friday close. S&P up 0.9%, a broad rally into the weekend, though all three big indexes still finished the week lower. On the week: Nvidia −4.45%, S&P −1.17%, Nasdaq −0.94%.
ON WATCH
The builders hit the brakes
On Saturday, Anthropic's chief executive Dario Amodei published a four-thousand-word essay called "We Must Pace the Frontier", arguing the industry should deliberately slow how fast its models get smarter. His message was blunt: the labs are improving AI faster than anyone can properly understand or control it.
The surprising part was who agreed with him. Sam Altman of OpenAI posted that "we need to pace the frontier," Elon Musk replied "Dario is right," and Google DeepMind's Demis Hassabis called it the right path forward.
The plan has some teeth on paper. It asks for independent evaluators inside every lab with employee-level access, federal rules behind them, and eventually global limits on the one thing researchers fear most, AI that rewrites itself.
A call to slow down from the firms racing hardest is rare, and it moved money fast. SoftBank, OpenAI's biggest backer, fell more than 11% on Monday after Altman said now would be an "ill-advised" time to go public.
None of this is binding, though. Amodei named no speed limit and no deadline, and a pledge to go slower is cheap for firms that still need better models to justify their prices.
MEERKAT'S TAKE: when the firms flooring the accelerator start asking for a speed limit, it's worth a look. Whether they lift off the gas is the only test that counts, and nothing here forces them to.
SPONSORED
Free email without sacrificing your privacy
Gmail is free, but you pay with your data. Proton Mail is different.
We don’t scan your messages. We don’t sell your behavior. We don’t follow you across the internet.
Proton Mail gives you full-featured, private email without surveillance or creepy profiling. It’s email that respects your time, your attention, and your boundaries.
Email doesn’t have to cost your privacy.
THE MACHINE
The robot took the desk
While the bosses argue about the future, AI is already turning up in the job numbers. More than half of the layoff events tracked in 2026 now name AI or automation as a factor, up from under 8% a year ago, according to figures compiled from Layoffs.fyi.
The count is real people. That same tracker puts the AI-attributed total above 170,000 workers this year, with Amazon cutting 16,000 corporate roles in January and the payments firm Block shedding about 4,000, the largest AI-blamed layoff on record.
Say it back simply. When a company wants to cut costs and look modern at once, "AI" is a convenient word to reach for, so some of these losses are ordinary belt-tightening in a smarter jacket.
QUICK PAUSE. WHY SHOULD YOU CARE? If your work sits anywhere near a screen, watch this more closely than the philosophy debate. The safest read is that AI is trimming the edges off a lot of jobs rather than swallowing them whole, and enough small trims add up to a payroll that quietly shrinks.
SOMEONE SMART SAID
Anthropic is built on distilling the world's ideas into their models; moaning about how others are distilling their distillation is a textbook pull-up-the-ladders move.
Marcus is neither a doomer nor a booster, which is why his read on the essay is the one worth having. He gave Amodei "two cheers out of three": the goal is right, but the plan hands the safety checks to an auditor, METR, that he calls "creatures of the Valley," and it uses the threat of China to excuse moving fast anyway.
Strip out the mission language and the worry is simple. A pledge to slow down is only as good as who enforces it, and this one lets the labs pick their own referees and set their own pace.
THE CHART
How big the AI money got
NVIDIA data-center segment revenue by fiscal year. Source: company filings (10-K/8-K).
Forget the essays for a second and look at the size of the prize. This is Nvidia's data-centre revenue, the chips behind almost every big AI model, and it went from $6.7 billion in 2021 to $193.7 billion in 2026.
That's close to a thirty-fold jump in five years. One company now sells nearly $200 billion of chips a year to feed the boom, which tells you how much money is riding on it carrying on.
These are booked sales to paying customers, and demand still outruns supply. The open question is how much of it leans on a handful of AI firms that are still losing money.
WHAT TO WATCH: if Nvidia's next data-centre number keeps climbing, the slowdown talk stays talk; the quarter it flattens is the quarter the boom is actually cooling.
HEADS UP
Nvidia weighs up to $10 billion to anchor Anthropic's record IPO · The chipmaker may help bankroll its own customer's roughly $2 trillion listing.
Anthropic's own report flags Claude misuse in surveillance and weapons research · The firm says it caught and blocked the attempts, but it wanted them on the record.
Musk and DeepMind's Hassabis also back the slowdown · "Dario is right," Musk wrote, a rare moment of four rival labs agreeing on anything.
Amodei admits China is the "toughest dilemma" for slowing AI · A US brake means little, he concedes, if Beijing's open-weight labs keep sprinting.
The layoff trackers everyone in tech is quietly refreshing · A running tally of who is cutting jobs in 2026 and how often AI gets named as the reason.
ON THE HORIZON
When (ET) | What |
|---|---|
Tue 15 | FOMC meeting begins, and SoftBank's $25.9bn bridge loan falls due |
Wed 16, 2:00pm | The Fed rate decision and fresh projections, and what they do to the AI trade |
Wed 1 Oct | SoftBank's final $10bn OpenAI payment is due, listing or not |
Before Nov | Anthropic's record roughly $2tn IPO is targeted to price, with Nvidia weighing a $10bn anchor |
MEERKAT ON YOUTUBE
AI Is Making Everything More Expensive. Meerkat's newest, on the Finance channel, does exactly what the title says and walks through where the AI build is pushing up everyday prices. Watch on Meerkat Explains Finance.
THE BURROW
Quick vote. The people building AI are now asking to slow it down. What do you make of it?
The people building AI are now asking to slow it down. What do you make of it?
Trivia. Dario and Daniela Amodei founded Anthropic in 2021 after walking out of a bigger rival. Which company did they leave? Answer's at the bottom of the email. No peeking.
ALSO WORTH YOUR INBOX
Meerkat also reads Crypto Nutshell, a plain-English take on digital money and the machines it runs on.
HOW DID WE DO?
One tap, and Meerkat listens. How did today's issue land? Hit a button, and add a line if you want to tell Meerkat why.
How did Meerkat do today?
Two of the three biggest names in AI just asked to slow down. Meerkat's back at 6:30 tomorrow, once the dust settles.
Someone in your life still thinks the AI crowd never doubts itself? Forward them this.
Meerkat is not telling you to buy anything, sell anything, or short anything. Meerkat is not a financial adviser. Meerkat is a small animal with a whiteboard and a brake pedal the robots keep ignoring. Please do not sue Meerkat.
TRIVIA ANSWER
OpenAI. The Amodei siblings and a group of researchers left OpenAI in 2021 over how fast to push the technology, then built Anthropic as the safety-first alternative, which is why this weekend's essay carries so much weight.