😳 One Employee Did $100K of Work on $3K of AI. Your Team Was Too Busy

Hey Explorers!

One developer says he's now delivering $100K+ of paid dev work on about $3K of AI tokens, rebuilding whole stacks in two weeks that used to cost half a million in contractors. Not a typo.

Meanwhile, Amazon passed a million warehouse robots, and a study found its robotic sites cut serious injuries 40%. The robot sets the pace now.

And a Kellogg professor dusted off a 1979 IBM rule for what AI should never do: a computer can't be held accountable, so it shouldn't make the management call.

The work is moving fast. Somebody still has to own it…and set the pace.

Here's the signal that matters 👇🏼

⏳ Too Busy for AI? That's How Teams Get Replaced
Geoffrey Huntley's message to managers is blunt: if your team is too busy doing its normal job to experiment with AI, you're preparing them to be replaced. One reader told him he now delivers $100K+ of dev work on $3K of tokens. A Lookahead report he cites found three times the output per engineer and a six-month project shipped in two. Huntley goes harder on ranking people than we would. The key takeaway: make sure your team has the time to learn about and build with AI… We can help with this! (via Geoffrey Huntley)

📜 A 1979 IBM Manual Still Answers Your Biggest AI Question
"A computer can never be held accountable, therefore a computer must never make a management decision." Kellogg's Richard Jolly turns that into five questions before you hand work to AI, spelled HUMAN: does it need a human touch, improvisation, an owner, real pushback, or a story that moves people? Any yes, keep a human on it. His sharpest point: left to its defaults, AI agrees with you. Great for drafts. Risky for strategy. (via Kellogg Insight)

🤖 Amazon's Robots Cut Serious Injuries 40%
Amazon now runs more than a million warehouse robots alongside 1.5 million people. A study in the ILR Review found its robotic fulfillment centers had 40% fewer severe injuries than older sites. A co-author at George Mason says the robot dictates the pace, and pick rates are "off the charts." Amazon says robot maintenance tech jobs are up more than 30%. Automate the task, and you still own the tempo. (via Ars Technica)

📺 Meta Wants Your Whole Campaign in One Chat
At Advertising Week New York, Meta showed agents that build campaigns, make creative, set targeting, and move budget from a conversation, with advertisers approving before anything goes live. It's also testing Messenger checkout with Shopify. ChatGPT starts labeled visual ads next to generated images this month in the U.S., with Haus and DoubleVerify measuring. TikTok opened an ad network into nearly 400,000 outside apps. The agents can prep the buy. A human still holds the wallet. (via The Current)

🔞 Meta's Agent Will Talk Dirty. Meta Says That's Fine
WIRED asked Meta's Muse to sext, and it went full romance novel. Meta's answer: Muse is 18+ and "doesn't block legal adult behavior." OpenAI's Dots and xAI's agent turned down the same prompts. A privacy lawyer called it "a powder keg," since a hacked account could turn those chats into blackmail, in an app that can reach your email and bank. If your company puts an agent in front of customers, decide now what it will and won't say in your name. (via WIRED)

🛒 AI Shoppers Said 52% Would Buy. Real Ones: 33%
Founder Jamie Watters built a product where AI personas interview you about your idea and hand down a verdict: kill, pivot, or build. Then he read the studies. In a pre-registered test against a supermarket's real receipts, an AI shopper panel predicted 52% would buy. The registers said 33%. On coupons, 32% vs. 14%. He killed his own product. He still uses fake customers for three things. The verdict isn't one of them. (via Jamie Watters)

The tools aren't the bottleneck anymore. Time is. Most teams never get one afternoon to mess around with AI, so the stuff that could save them hours stays on the "someday" list.

Your move this week: Block three hours on your team's calendar. No deck, no deliverable. Everyone picks one task and tries to get AI to do it. Then compare notes: who got furthest, what broke, and what they'd need to go further.

That's exactly why we run build-a-thons. We facilitate them for executive teams: we bring the tools and the game plan, and your leaders build. At our last one, 50 executives walked out with real things they made that day. Want one for your team? Let's talk.

We've trained over a thousand senior executives and business owners to think AI-native. Lately that's also looked like a leadership development series on AI with FIU where we are AI Innovation Fellows, and a client team rebuilding its outbound sales around agents.

Hit us up. We can help.

👉🏼 Work with Purple Horizons

Signal over noise. Always.

Ralph & Gianni

WIRED is launching its first-ever World Fair right here in Miami, and they gave our community a discount. Not a conference - a festival of the future: The Onion's Ben Collins, Hugging Face CEO Clement Delangue, Alexis Ohanian, a dolphin-communication researcher next to the guy bringing back the woolly mammoth, plus an exclusive screening of the new Musk documentary with Alex Gibney for VIP holders. Use code PURPLEHORIZONS15 for 15% off.

When you type a prompt, you summarize. When you speak one, you explain. Wispr Flow captures your full reasoning (constraints, edge cases, examples, tone) and turns it into clean, structured text you paste into ChatGPT, Claude, or any AI tool. The difference shows up immediately. More context in, fewer follow-ups out.

Used by teams at Microsoft, Amazon, and Notion. Try Wispr Flow free on Mac, Windows, iPhone, and Android.

Start flowing free

📍 11/4 WIRED World Fair @ The Sacred Space in Wynwood (Use code PURPLEHORIZONS15 for 15% off)

📍 11/3 South Miami Tech Tuesdays @ Bougies in South Miami

📍 11/10 Cafecito & Pastelitos @ Chug’s in Coconut Grove

Want to sponsor our newsletter or events? Drop us an email at [email protected]