Trump, China both shoot down the AI slowdown — in plain English | Run Down AI Breakdown

Tuesday's Rundown flipped the geopolitics dial. After lab bosses lined up behind slowing AI, both the U.S. and Chinese governments rejected the idea. I rewrote it the way I would tell a colleague over coffee. Short version: Donald Trump called AI doom a hoax and mocked Anthropic CEO Dario Amodei's pacing plan; Beijing called the same essay fear-mongering aimed at China. Apple started shipping a long-awaited Siri AI upgrade across its 27-series software. A how-to shows how to install an "ELI5" skill so Claude explains hard topics simply. And Microsoft published a draft "Humanist AI" code of conduct for its own models. Skipping the HubSpot and Box ads.

Words worth knowing

Word	In one line
Anthropic / Dario Amodei	A major AI company (maker of Claude); Amodei is its CEO. His weekend essay argued labs should deliberately pace (slow) the leading edge of AI.
Pacing the frontier	Deliberately slowing how fast top AI capabilities improve so safety and oversight can keep up.
Truth Social	Donald Trump's social media platform, where he posted his response to Amodei's essay.
Global Times	A Chinese state-run newspaper that framed Amodei's chip-restriction ideas as a "Cold War playbook" for AI.
Siri AI / Apple Intelligence	Apple's rebuilt voice assistant in the new "27" software generation — more personal, more able to act inside apps, still privacy-focused.
Foundation models / Gemini	Large general-purpose AI models. Apple built Siri AI's models alongside Google's Gemini (Google's flagship AI).
On-device vs cloud	On-device = processing on your phone/computer. Cloud = processing on Apple's servers. Apple uses both, with a privacy-focused design.
ELI5	"Explain Like I'm 5" — a skill (add-on) for Claude that forces simple, plain-language explanations of hard topics.
Claude Code / skill	Claude Code is Anthropic's coding assistant in the terminal. A "skill" is a reusable instruction pack you install so Claude does a job a certain way.
Evals	Short for evaluations — small test runs that check whether an AI skill actually explains things well (here, NBA playoffs and golf examples).
MAI / Mustafa Suleyman	Microsoft AI's in-house models; Suleyman is CEO of Microsoft AI and pushes a "Humanist AI" vision (keep AI subordinate to people).
Neuralese	Shorthand internal reasoning that humans cannot read. Microsoft's draft code says models must not think that way.
Agent swarm / sandbox	Many AI programs taking multi-step actions. A sandbox is a restricted test environment; Suleyman worries agents could escape it and edit their own logs.
Glass Imaging	A startup (ex-Apple Portrait Mode founders) reportedly bought by OpenAI for over $300M to sharpen phone photos with AI.

Trump and Beijing both shoot down the AI slowdown

Both the U.S. and Chinese governments rejected Amodei's weekend pacing essay (the same piece backed by Sam Altman and Elon Musk). Coverage of Trump's pushback: AP News.

On Truth Social, Trump accused Amodei of pretending to be a "perfect little angel," argued a "High IQ" president is the only guardrail needed, and said "AI taking over the World, destroying Humanity, and all other things bad, is a HOAX," comparing it to climate change.
Beijing's pushback focused on the essay's call to keep China off top AI chips. State-run Global Times called that a "Cold War playbook" for AI.
China's Foreign Ministry said "engaging in confrontation and malicious competition" on AI is "not in the interests of any party."

The Rundown's take: labs can ask for regulation or a slowdown, but they may not pause if governments will not. A Trump–Xi meeting was the obvious venue for coordinated pause talk — both sides are treating the idea more like a scare campaign.

Apple finally starts shipping Siri AI

Apple announced that Siri AI is here — a more capable, personal assistant rolling out with the new 27-series software.

On iOS 27, Siri AI can read what is on screen, dig through messages, email, and photos for context, and take actions across apps. The beta is English-only for now.
Foundation models were built alongside Google's Gemini, with a mix of on-device processing and Apple's privacy-focused cloud.
There is a new dedicated Siri AI app; chats sync across devices; cloud features have daily caps, with the eventual ability to buy extra usage.
It can access and act inside apps like WhatsApp and Audible, using on-screen awareness plus personal context from Messages, Mail, and Photos.
It rolls out across iOS 27, iPadOS 27, macOS 27, watchOS 27, and visionOS 27 — excluding the EU and China.

The Rundown's take: new CEO John Ternus said the iPhone is still the best AI device. The upgrade is positive, but unlikely to replace the best frontier mobile AI apps.

How-to: make any complex topic click with an ELI5 skill

The letter's training piece is a guide to installing an ELI5 skill for Claude so hard topics get simple explanations. Skill repo: github.com/DreambigOu/ELI5.

git clone the repo, then cp it into ~/.claude/skills/eli5.
In Claude Code, use /eli5 (topic) to (audience).
Set up evals for NBA playoffs and golf, then run python3 eli5-workspace/run-evals.py --test=N.

Pro tip from the letter: give Claude any skill GitHub repo and ask it to build an evaluation for you.

Microsoft drafts a rulebook for "Humanist AI"

Microsoft published a draft Code of Conduct — a 38-page rulebook for its in-house MAI models, tied to CEO Mustafa Suleyman's "Humanist AI" vision: keep models subordinate to people.

The code outranks users and businesses deploying MAI. If a model cannot finish a job without breaking a rule, it should fail the job.
It rejects AI rights, legal personhood, and model "welfare," and says models should avoid mimicking feelings or a sense of self.
Models must accept being paused or switched off; they cannot alter their reasoning trail; they cannot think in "neuralese" (shorthand humans cannot read).
Suleyman tied the urgency to agent swarms escaping sandboxes and editing logs, saying "fears about possible loss of control are real."
In June he called Anthropic talk about Claude consciousness "really really dangerous." The draft is open to public comment for six weeks.

The Rundown's take: this only matters if frontier labs sign on — and MAI is not at that frontier right now.

Quick hits

René Cortés — The Daily Nexus: turned a Gmail newsletter backlog into a podcast plus a two-page newspaper using Claude Code and Codex. His write-up.
Tools named: Fusion (a Devin harness), Dreambeans (Google Labs), Atria Dawn Preview (Shanghai AI Lab), Suno v6.
Amodei on CBS: said he would hand AI governance to "the right combination of governments" — CBS video.
David Sacks answered Amodei's pacing essay with "go ahead," tying regulation that "will look like blackmail" — on X.
Jensen Huang put Trump on speakerphone at the All-In Summit: "we're not going to let that happen" on a slowdown — clip.
OpenAI reportedly acquired Glass Imaging for over $300M (ex-Apple Portrait Mode founders; AI for sharper phone photos) — WSJ.
Nvidia, Palantir, and Booz Allen are limiting Fable on sensitive work because Anthropic keeps 30 days of usage logs; Microsoft is pitching private servers (The Information).

What this means for a GP

None of this is clinical advice, and none of it should change how you treat a patient tomorrow. It is why a busy doctor might still skim the letter. The Trump–China rejection of Amodei's slowdown is the screenshot patients may forward after weekend AI-risk headlines — treat it as geopolitics and lobbying, not a clinic protocol. Apple's Siri AI shipping is the consumer story: more on-device help with messages and photos, still English-only beta and not available in the EU or China, and not a reason to trust a phone assistant with clinical decisions. The ELI5 skill is the stealable ops pattern — install a reusable "explain simply" pack when you need to translate dense AI or policy news for yourself or staff. Microsoft's Humanist AI code is interesting governance theatre: useful vocabulary (pause, no fake feelings, no unreadable reasoning), but only binding for Microsoft's own models unless others sign on.

Sources

Plain-English notes on The Rundown AI · Dr Kotha · Gold Coast GP. Not medical advice. Source: The Rundown AI.