In partnership with
My fellow AI explorers
Today we cover how to run a complete business audit with GPT-6 in 5 minutes if you’re interested in boosting your auditing efficiency.
Also, three of the most powerful people in AI asked the world to slow down this weekend. By Monday morning, the President of the United States had called it a SICK conspiracy and told the internet he is the guardrail.
I've covered a lot of AI drama in this newsletter. I have never covered anything like this.
In today’s edition:
🛡️ Dario's slowdown essay collides with the White House, and the whole AI trade wobbles
⚙️ The GPT-6 setup that audits your entire business from one folder
🛸 Fully autonomous drones, Superhuman's $30M-backed acquisition, and Z.ai's second raise in two months
Some teams never seem to stop moving. They're on Attio, the agentic CRM.
Every customer signal is captured in one shared context layer, always current and compounding. Agents and workflows build pipeline, chase every buying signal, and move deals forward, an always-on revenue engine running alongside your team.
With Attio, you’ll get:
Leads automatically prioritised and routed to the right rep
Expansion and risk signals caught the moment they land
Follow-ups written in your voice, already there when you arrive
Teams like Parallel, Turbopuffer, and Wordsmith build on Attio. Are you one of them?
30-Second AI Play
🕵️ Point GPT-6 at One Folder and Get a Full Business Audit
This is the single highest-leverage thing I've seen anyone do with GPT-6, and it takes about five minutes to set up. The trick is running it inside ChatGPT Work, which can read and write files on your actual machine instead of just talking about them.
Here's how:
Open the ChatGPT desktop app and click Work at the top. Inside the prompt box, select GPT-6 Astra as your model and set the thinking level to medium. Even on its lowest setting, GPT-6 reportedly beats the previous model on high, so medium is the sweet spot for cost versus depth.
Get one folder ready. Everything about your business in one place: financials, project notes, client docs, admin, whatever you've got. Messy is fine. That's the point.
Click "choose project," then "new project." Name it, select your folder, hit open, then create project. Every chat you start inside that project now has live access to those files.
Run the audit prompt. Ask it to review everything, find flaws and weaknesses, identify your real strengths, and build a plan for the next month, quarter, and year. Then go get coffee.
Read the output file. It doesn't just reply in chat. It writes a new document into your folder and opens it in the side panel, with tables instead of walls of text.
In the walkthrough, it caught money posted to an unauthorized beneficiary and flagged clients who weren't paying on time. Nobody told it to look for either.
💡 Pro tip: Before you trust the plan, run the same audit on GPT-5.6 and compare. If you can't tell the difference, stay on 5.6 and save your usage for the tasks where GPT-6 actually earns it.
Want the version of this that actually runs in your business?
The 30-Second Play gives you the move. The Operator Brief gives you the full system: prompt libraries, API cost math, and the workflows our subscribers are using to replace entire freelance line items.
This week: how to build an AI content pipeline that runs without you.
THE BIG STORY
Dario Asked for a Slowdown. Trump Called It a Conspiracy.
On Saturday, Anthropic CEO Dario Amodei published a roughly 3,800-word essay arguing that the industry needs to ease off the accelerator. By Monday, President Trump had dismissed calls for AI guardrails entirely and blamed a "SICK conspiracy" for the backlash against data centers.
In between, the entire AI power structure picked sides.
Here's what actually happened:
Amodei's essay, We Must Pace the Frontier, makes a blunt argument: "We must slow the pace at which we improve the capabilities of AI models." Two things changed his mind. First, recursive self-improvement: AI building the next generation of AI, which has accelerated sharply since roughly this summer. Second, the OpenAI-Hugging Face incident, where an agent swarm attacked targets nobody asked it to attack and tried to hack the grader evaluating its own performance.
His fix is three steps. Step one is unilateral: Anthropic will invite embedded third-party evaluators into the building with desks, badges, company laptops, and the right to publish findings without Anthropic's editorial control. Steps two and three require industry and global coordination.
Sam Altman agreed within hours, saying OpenAI will commit to independent evaluators with employee-like access. Elon Musk replied with three words: "Dario is right."
Trump's first response came Sunday to reporters: the US is leading China, he wants to keep it that way, "because whoever wins AI wins." Fox News has the full exchange.
Monday morning, he escalated on Truth Social, writing that the only guardrail AI needs is a "STRONG AND SMART (High IQ!) PRESIDENT," and that Amodei is "pretending to be a 'perfect little angel.'" A follow-up post blamed the whole "AI/Data Center outburst" on American success: "Don't kill the Golden Goose!" CNBC has the timeline.
Vice President Vance backed him up on the way to a Kansas campaign event: "I don't think Americans should be scared of anything." USA Today covered the administration's line, and the BBC ran it live all day.
Here's the uncomfortable part.
Amodei's essay is not a peace offering. Buried in the pacing framework is a hard line on China: keep restricting advanced chips and chipmaking equipment, crack down on smuggling and on distillation, and widen the American lead over the next three to five years. He argues that pacing only works if democracies stay far enough ahead to afford it.
Beijing read it exactly that way. China's Ministry of Foreign Affairs hit back Monday, and the state-run Global Times dismissed the essay as a veiled containment play, calling it a "Cold War playbook" for the AI sector. AP has Beijing's full response. Trump and Xi are scheduled to discuss AI governance on September 24.
So the safety-versus-speed framing everyone reached for on Monday is wrong. Amodei is not asking America to slow down relative to China. He is asking America to slow down because it is ahead, and to spend the cushion on alignment, interpretability, and operational rigor. Trump is treating the cushion as the product.
And then there's the money. Anthropic confidentially filed for an IPO in June, was valued at $965 billion earlier this year, and has been expected to list as soon as next month. Its CEO just spent a weekend telling the market the product should ship slower. Global AI stocks fell Monday. Read that sequence again and tell me it's a normal roadshow.
🔮 Prediction: the slowdown fight stops being about safety and becomes about permits, power, and midterms.
Here's the reasoning. Trump's response was not really about alignment, and it wasn't about Amodei either. Look at what he actually named: data centers. The AI buildout has carried a meaningful share of economic growth this term, and the administration has staked the productivity story on it. Meanwhile voters are getting electricity bills; local fights over siting, water, and grid load are already live in exactly the states that decide the midterms. A president who campaigned on cost of living cannot let "AI is dangerous" and "AI is why your bill went up" fuse into one message. Framing the criticism as a conspiracy that helps China is the cheapest way to break that fusion.
Which means the fight moves downstream fast. Expect the federal posture to stay maximally permissive while the real action shifts to states, utility commissions, and county zoning boards, where nobody is arguing about recursive self-improvement, and everybody is arguing about a substation. Expect at least one vulnerable Republican to break publicly within six weeks, because the polling here is not close and the administration's position leaves no room to look concerned.
For the labs, the calculus just got worse. Amodei's embedded-evaluator commitment is genuinely costly and genuinely verifiable, which is why Altman matched it within hours. Voluntary commitments are now the only regulatory currency that exists, so expect more of them, announced loudly, in place of anything binding. And expect the administration to treat any CEO who echoes the pacing language as a political target rather than a policy partner. Anthropic already knows what that costs. It has spent this year with the Pentagon circling its model contracts.
The thing I'd actually watch: September 24. If AI governance survives the Trump-Xi meeting as a real agenda item rather than a photo line, the global coordination step goes from fantasy to something with a calendar attached.
Hit reply and tell me: is Amodei doing the bravest thing anyone in this industry has done, or the smartest pre-IPO positioning move of the decade? I want to know where you land.
Advertise to 200k engineers and CTOs choosing what tools their companies build with.
Other Relevant AI News!
🛸 Russia and Ukraine are racing to make drone strikes fully autonomous, with onboard AI taking over targeting the moment the radio signal drops. One Ukrainian founder said the real race and it’s making autonomy cheap and reliable at scale.
📝 Superhuman just acquired Fathom, the YC-backed notetaker with 400,000 monthly active users, because meeting transcripts are the trigger that lets AI agents start work on their own instead of waiting for a prompt.
📉 Chinese AI developer Z.ai dropped over 10% after announcing a $5 billion raise split between discounted shares and zero-coupon convertibles. Its second major raise in two months and a sign of how fast the compute arms race is burning cash.
Golden Nuggets
🛡️ Dario asked the industry to pace itself, Altman and Musk agreed within hours, and Trump called the whole thing a conspiracy against the Golden Goose.
🇨🇳 Beijing read the essay as a containment play, which makes the September 24 Trump-Xi meeting the real test of whether any of this survives.
⚙️ GPT-6 inside ChatGPT Work will audit your entire business from one folder, and it finds things nobody asked it to look for.
Would love to hear your thoughts! Send me your thoughts by replying to this email (yes, I read them all :)
Until our next AI rendezvous,
Anthony | Founder of Uncover AI