Weekly Recap · Week 37/2026

AI Recap: the week AI's own CEOs pushed the brakes

An unusual tone out of Silicon Valley: Anthropic CEO Dario Amodei calls for a slower pace of model development — and gets backing from none other than Sam Altman and Elon Musk. Meanwhile OpenAI delays its own IPO, an AI math proof sparks a credit dispute, and Germany's Mittelstand shows it couldn't care less about the slowdown talk.

The week in one sentence

Three lab CEOs, one unusual tone: Dario Amodei calls for a slower pace on new AI models, Sam Altman and Elon Musk agree, and OpenAI delays its own IPO to 2027. In between, an AI math proof sparks a fight over credit. More relevant for day-to-day SMB work: Germany's Mittelstand keeps accelerating AI-agent adoption, no brakes applied.

Anthropic CEO Amodei calls for slower AI — Altman and Musk agree

Anthropic CEO Dario Amodei said publicly on September 12 that the pace of frontier AI development should slow down. His argument: model capabilities are growing faster than safety research, regulation and society can keep up with, and a deliberately slower pace is safer for everyone, even at some short-term competitive cost. What stands out is the reaction from rivals: Sam Altman (OpenAI) and Elon Musk (xAI) publicly agreed, a rare moment of alignment between three labs that otherwise out-ship each other on a near-weekly basis. Whether the words translate into an actually slower release cadence remains to be seen.

Source: Bloomberg Source: TechCrunch

OpenAI delays its IPO to 2027: Altman calls 2026 "ill-advised"

The same day, OpenAI CEO Sam Altman told Fortune that going public in 2026 would be "ill-advised" — the company is in the middle of safety-critical work that would be disrupted by the quarterly pressure of a public listing. An IPO is therefore expected no earlier than 2027; Altman gave no exact date. For a company last valued in the hundreds of billions, that is a deliberate brake, and it fits the tone of the week: first Amodei's call to slow down, now Altman passing on the obvious next growth step.

Source: Bloomberg Source: TechCrunch

10,000 AI agents claim to have cracked a century-old math problem — a mathematician disagrees, loudly

A story that started the week before kept making headlines: OpenAI says it ran 10,000 AI agents in parallel on the Navier-Stokes equations, one of the seven unsolved Millennium Prize Problems in mathematics, and reported a breakthrough after 88 hours of compute. NYU mathematician Tristan Buckmaster publicly disputed the claim, accusing OpenAI of taking credit for prior work by others and of handling the announcement poorly. OpenAI subsequently published a full, checkable proof — but the credit question stayed contested. A good reminder for us: any "AI solves century-old problem" headline deserves a second look.

Source: TechCrunch Source: Axios Source: CNBC

While the big labs talk about slowing down, Germany's Mittelstand keeps accelerating

A useful counterpoint comes from the Mittelstandsbund: while the frontier labs debate pace in public, German small and mid-sized companies are firmly in execution mode. 51.2% of Mittelstand companies now use or test AI, up 54% from 33.1% in 2024. The jump is even sharper for AI agents that act on tasks rather than just draft text: adoption has nearly doubled, from 8.7% to 16.6%. The survey itself dates back to March 2026, but it points at exactly the shift we see in our own client work: the question is no longer "whether", but "which process first".

16.6% of German Mittelstand companies already use AI agents, according to the Mittelstandsbund — nearly double the 8.7% in 2024.Mittelstandsbund, KI-Index Mittelstand 2026, March 2026
Source: Mittelstandsbund

Sources & further reading

As of September 13, 2026. Figures per company statements and the cited media, without guarantee. The Mittelstandsbund figures are from the "KI-Index Mittelstand 2026" survey published in March 2026 and do not reflect this week specifically, but illustrate the underlying trend.

FAQ: the AI week

Why is Anthropic CEO Dario Amodei calling for a slower pace of AI development?
On September 12, 2026, Amodei argued that AI capabilities are growing faster than safety research and regulation can keep up, and called for a deliberately slower development pace. Notably, Sam Altman (OpenAI) and Elon Musk (xAI) publicly agreed, even though their companies are direct competitors.
When will OpenAI go public?
Not before 2027. On September 12, 2026, Sam Altman told Fortune that going public in 2026 would be "ill-advised" given ongoing safety-critical work. He did not give an exact date for 2027.
Is the Navier-Stokes problem really solved now?
OpenAI says 10,000 AI agents working in parallel produced a complete proof for the Navier-Stokes equations. NYU mathematician Tristan Buckmaster disputes the originality of the work and accuses OpenAI of claiming credit for prior research by others. The credit question remains contested among mathematicians, even though the proof itself has been published.
How many German Mittelstand companies already use AI agents?
According to the Mittelstandsbund's KI-Index Mittelstand 2026 (as of March 2026), 16.6% of Mittelstand companies use AI agents, nearly double the 8.7% in 2024. Overall, 51.2% of companies use or test AI, up 54% from 2024.

While others debate pace, you can just get started

In the free 60-minute workshop we put your process on the board and find out which part of it can already be automated today, regardless of which model is making headlines this week.

Request a workshop