Sunday, September 13, 2026
TL;DR
Musk, Rishi Sunak and Google DeepMind's Hassabis all backed Anthropic's call to slow AI development within a day of the essay; OpenAI delayed its IPO to 2027 over safety concerns. None of the four companies has cut spending to match the rhetoric.
Signals
Telling a coding agent to follow standard testing methods often makes it worse, not better #
Dan Luu's September 7 test of 26 prompting conditions on a real Rust Zstd implementation found agents given no testing instructions scored above average, while explicit TDD produced twice as many tests that enforced incorrect behavior and formal-verification tools like Verus and Lean 4 mostly proved properties unrelated to the actual bugs. Property-based testing was the one structured technique that helped. Luu
The fight over whether one resignation marks a real safety turning point is intensifying, not settling #
Zvi Mowshowitz's September 11 post treats Jacob Coxon's resignation as a genuine tipping point, citing Anthropic's Evan Hubinger publicly backing odds above 10% that AI kills everyone within a decade — a direct rebuttal to Nathan Lambert's September 10 piece calling the reaction a 'wildfire effect' that polarized rather than clarified. Zvi
A practitioner is calling the week's lab-backed pacing proposals self-serving, not safety-driven #
Armin Ronacher's September 12 post argues Amodei's pacing plan hands oversight to METR, which he says has 'strong ties to both OpenAI and Anthropic,' and predicts the real damage is economic concentration, not extinction — arguing open-weight models already provide 'built-in pacing' through proliferation. Ronacher
News
WATCHOpenAI delays its IPO to 2027, citing AI safety concerns #
Altman told Fortune an IPO 'this year would be an ill-advised moment' given safety concerns, confirming 'not 2026' — the first concrete business move tied to the week's pacing debate. Fortune
WATCHNo lab has cut spending to match the slowdown talk #
Despite four public figures now endorsing Amodei's pacing call, Microsoft guides toward $175 billion in fiscal 2027 capex, Google Cloud grew 82% last quarter, and Nvidia guides Q3 revenue to $108 billion. Yahoo Finance
WATCHSanders says pacing is 'not enough,' renews his push for an outright AI pause #
Sanders called Amodei, Musk and Altman's pacing stance 'a start, but it's not enough': 'when you're racing towards a cliff, you don't ease off the gas, you hit the brakes.' He repeated his call for a legal pause on advanced AI. Soap Central
The long view
A year ago, no sitting CEO of a frontier AI lab had publicly backed slowing capability development — 2023's pause letter was signed by outside researchers, not lab leadership, and every major lab kept shipping through it. This week, Anthropic's and OpenAI's CEOs, xAI's Musk and Google DeepMind's Hassabis all endorsed Amodei's pacing call within about a day, and OpenAI made the first concrete move: delaying its IPO to 2027 over safety concerns. If this holds, the next test is whether spending bends too — Microsoft still guides toward $175 billion in 2027 capex, unchanged so far.
Also noted
- Claude Code v2.1.269 added claude plugin eval, scoring a plugin's eval suite with a reproducible JSON and HTML report. Claude Code changelog #
- Google's threat-intel team found a separate criminal group harvested thousands of cloud credentials using autonomous AI agents in under six hours. The Hacker News #
- Ant International, Visa and Mastercard proposed a shared Know-Your-Agent framework to interoperate their competing AI shopping-agent identity protocols. PYMNTS #