Support Us & Upgrade to the Operator Edition

Happy Halloween. The scariest thing I read this week wasn’t a ghost story. It was the guy who blew the whistle on Theranos, warning that AI is the perfect cover for the next big fraud. Meanwhile, Asana’s CEO quietly handed half his job to AI agents, and somebody built an AI that literally draws on your screen while you talk to it.

This week, we’ll deep dive into: how to tell real AI from AI theater, before it costs you.

Ready to transform artificial intelligence from a buzzword into your personal revenue generator?

HubSpot’s groundbreaking guide "200+ AI-Powered Income Ideas" is your gateway to financial innovation in the digital age.

Inside you'll discover:

Download your guide today and unlock a future where artificial intelligence powers your success. Your next income stream is waiting.

Get The Free Guide

Ever watched a YouTube explainer and still had no idea what the person was talking about? heyclicky fixes that. It’s an AI buddy that lives on your Mac or PC, sees whatever’s on your screen when you ask, and with its new real-time graphics feature, it generates custom diagrams and visuals on the fly while it explains. Not pre-rendered. Built live, for your exact question.

Here’s how to set it up:

🔍 Why it’s special:

💡 Pro Tip: Use it when onboarding to new software. Open the tool you’re learning (DaVinci Resolve, Figma, a new CRM) and ask, “Walk me through how to do X here.” It points the way directly in the interface, which beats pausing a tutorial every 10 seconds. If you love it, Pro is $20/month for unlimited talk.

Try heyclicky Free Here

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 train your AI platform with skills and then cut down the time you spend prompting and retyping with your other AI tools.

👉 Upgrade to The Operator Brief

🕸 OpenAI will soon honor a single robots.txt signal that lets websites opt out of AI training across multiple providers at once, and Anthropic, Google, and Microsoft are on board too. Opting out of AI training used to be a chore, and now it’s become a checkbox. Whether the companies honor it is the part worth watching.

👨‍💻 Proton’s CTO told engineers “your old job is gone” in an internal AI manifesto, claiming AI wrote 90% of the company’s October code, and the internet is split down the middle. It may be the future of the profession, but it may also be a security review waiting to happen. We’re in a weird moment in a number of industries for the time being; everyone, including CEOs, middle managers, and entry-level employees, is waiting to see how the dust settles on how AI will fit into companies in the long term.

⛪ Pope Leo XIV says AI must be “disarmed” and that decisions over human life can never be handed to machines, urging states to ban lethal autonomous weapons at next month’s Geneva talks. He’s asking for one rule above all: no machine decides who lives.

🎁 Anthropic’s co-founders have pledged 80% of their equity to charity, and the company has put 5% of its shares into a public benefit fund ahead of the IPO. Generosity ahead of a payday is hard to read without wondering about the timing. Either way, the pledge is on the record, and it’s far more than competitors have pledged.

📊 Anthropic’s revenue per employee just crossed $6.6 million, up from $1.4 million a year ago. That’s more than a fourfold jump in twelve months, and a number that makes most of Big Tech look sluggish.

Tyler Shultz, the former Theranos employee who helped expose Elizabeth Holmes, told TechCrunch that the AI boom is “a great opportunity to go deceive people.”

He would know. Theranos was his first job out of college, and the company’s blood tests were giving faulty results to real patients at Walgreens. Here’s why he sees the same conditions forming in AI:

The timing isn’t random. You Can See Everything, Nathan Fielder and Lance Oppenheim’s documentary about Holmes’ final weeks before prison, has become one of the highest-grossing documentaries ever. Theranos is back in the cultural conversation right as AI money hits its peak.

Here’s the uncomfortable part.

Look at how Theranos actually pulled it off, and then look at AI:

Now, swap “blood test” for “proprietary model.” Swap “trade secrets” for “we can’t disclose our architecture for competitive reasons.” Swap “prestigious board” for “backed by Microsoft.” It maps uncomfortably well.

And it’s already happened once. Builder.ai, a Microsoft-backed startup once valued around $1.5B, collapsed into insolvency in 2025 after its sales were found to be inflated by roughly 300%, amid reports that much of its “AI” work was done by human engineers.

🔮 Prediction: The next AI fraud won’t be a fake model. It’ll be fake numbers wrapped around a real one.

Here’s my reasoning. AI has one massive difference from Theranos: you can test most AI products yourself in about 30 seconds. Holmes could hide a broken machine in a lab. You can’t hide a bad chatbot from someone typing into it. So the “the tech doesn’t work at all” fraud is actually harder in AI than it was in biotech.

What’s easy to fake is everything around the tech. Revenue that’s really one-off pilots dressed up as annual contracts. “Proprietary AI” that’s an API call to OpenAI or Anthropic with a nice interface. Agent demos that are staged, scripted, or quietly backed by humans. Builder.ai is the template: the product looked real, the numbers weren’t.

The incentives point straight at this. When valuations are driven by revenue multiples and “AI-native” branding, every founder under pressure has a reason to stretch both. And with funding increasingly coming from crossover funds, sovereign wealth, and wealthy individuals who are terrified of missing out, the people writing the biggest checks are often the least equipped to audit a model.

My bet: within the next 12-18 months we see at least one high-profile AI startup unravel over revenue quality, not model quality. The fix is boring, but it works. Ask what the product does without the underlying model provider, ask for revenue broken out by contract length, and treat “it’s a trade secret” as a red flag, not a moat.

💬 Hit reply and tell me: which AI company do you think is the most overhyped right now? No judgment. I’ll share the most popular (anonymized) answers next week.

Read the Full Interview Here

Asana CEO Dan Rogers told Business Insider he now lets a squad of AI agents steer his workday, from his inbox to his competitors.

Rogers, 49, took over from co-founder Dustin Moskovitz in 2025 and is dogfooding the exact product Asana sells. His agents handle:

He also runs agents on competitive intelligence and customer signals, flagging delayed renewals and accounts that need a CEO’s attention.

The headline lesson? Agents are only as good as the instructions you give them. Rogers said the bots may need detailed direction to produce anything genuinely useful, and that governance and constant calibration are part of the job. Asana even restricts what data its business analyst agent can access and what actions it can take.

Here’s the asterisk.

This is a CEO demoing his own product to a national publication. Asana has been pushing “AI Teammates” since 2025, with Rogers famously arguing that “autonomy is the wrong goal.” So take the glowing parts with a grain of salt.

But look closely at what he actually replaced. Not engineers. Not salespeople. The people he used to email for a report. The analyst who built the dashboard. The chief of staff who tracked his commitments. The PM who matched engineers to backlog tasks.

🔮 Prediction: The first job AI agents genuinely eat isn’t the junior role. It’s the coordination layer in the middle.

Here’s why. Junior work is often hands-on and messy. Coordination work is structured: gather information from systems, summarize it, route it to the right person, follow up. That’s exactly the work agents are best at today, and it’s exactly what Rogers just automated. Every “can you pull the numbers on X” request that used to take a day and a Slack thread now takes a prompt.

The second-order effect is more interesting. Rogers’ biggest lesson was that agents need clear instructions, governance, and calibration. That’s a management skill. Which means the executives who win the next five years aren’t the ones with the biggest teams; they’re the ones who can write a precise brief, set guardrails, and check the output. “Manages a team of 40” on a resume starts looking less impressive than “runs 12 agents that do the work of 40.”

Expect this to show up in job descriptions within 18 months, and expect companies that own the “system of record” (where work and data actually live, like Asana, Salesforce, or ServiceNow) to have the edge, because agents are useless without context.

💬 Hit reply: what’s the first part of your job you’d hand to an agent tomorrow? I read every reply, and the best ones might show up in a future edition.

Read the Full Story Here

Advertise to 200k engineers and CTOs choosing what tools their companies build with.

👉️ Learn how to sponsor this newsletter

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