In partnership with
Sign Up | Advertise | Dispatches | Consulting
Hey Explorers!
AI has gotten very good at getting the work done. The edge now is judgment… What you choose to hand over to the AI, and what you keep on your side of the table.
Here's the signal that matters 👇🏼
🗞️ TECH & TRENDS | The Judgement Layer
📂 She Asked an AI to Grade Her Own Job Performance
Writer Katie Parrott built a career-coach agent inside a single desktop folder, then told it to audit her quarterly goals. In ten minutes it pulled receipts from her Slack, her Drive, and the open web, and handed back a verdict that flattery and panic could not touch. The move worth copying: she set the sources and the boundaries, the agent did the gathering, and the judgment about what actually counts stayed hers. (🔗 Link)
🛠️ HubSpot Wrote Down Exactly How It Went AI-First. Steal It.
CEO Yamini Rangan published the actual playbook: 94% of employees on AI weekly, 3,900 agents built in-house, and one rule above the rest, start with the work and not the tool. The receipts are specific, including AI video spots at $300 to $3,000 instead of $300K and 80% of interview scheduling fully automated. Her sharpest line is to chase outcomes, not token counts, which is the whole difference between a transformation and an expensive habit. (🔗 Link)
🎯 AI Is Sending You Customers and Crediting Someone Else
When ChatGPT or Gemini recommends a brand to someone new, that person is about 182% more likely to go search it within the week. But your dashboard files that visit under "organic" or "direct," so the channel that actually moved them never shows up. Scrunch dug through millions of search events and found a recommendation does roughly twice the work of a passing mention, and brands named first or called "best" get searched far more, which means the AI answer is quietly feeding your funnel while you thank something else. (🔗 Link)
🩻 The Company Behind Your AI Art Wants to Scan Your Whole Body
Midjourney just opened a medical division building something it calls Ultrasonic CT, a full-body scan that takes 60 seconds with no radiation and no magnets, just sound and water. The first one lands inside a flagship "Midjourney Spa" in San Francisco at the end of 2027, complete with saunas, cold plunges, and ten scanners. The plan is a billion body scans a month, which is a strange and enormous bet from a company that made its name generating pictures. (🔗 Link)
🎵 The Millions of Songs Mashed Into AI-Generated Music
The Atlantic published searchable databases naming more than 21 million copyrighted tracks circulating among AI developers, from Taylor Swift to the Beatles, and SZA found 238 of her own songs in the pile within hours. For two years these companies hid behind "nobody can prove what trained the model," and that cover is now gone. With a fair-use ruling expected this summer, the fight shifts from how a song sounds to how the data was acquired, which makes provenance the new liability for anyone building on these tools. (🔗 Link)
👗 Luxury Shoppers Buy 10x More When They See It on Themselves First (Thanks AI!)
DressX says shoppers who use its AI try-on are 50% more likely to buy, and for luxury labels the conversion jump runs as high as ten times. The pitch is a digital twin built from one selfie, so a customer watches the coat move on their own body before paying a cent. The decision is sliding off the product page and into the mirror, and the brands without a try-on layer are asking buyers to gamble on faith. (🔗 Link)
🧠 Choosing to Stay Human
Ethan Mollick's argument is that the habits forming right now, frictionless by design, are quietly deciding what you stop doing for yourself. He points to Wharton research on "cognitive surrender," where sharp professionals take a confident AI answer even when it's wrong, and to consultants who got measurably worse on the one task the AI couldn't handle. Staying human isn't refusing AI. It's deciding on purpose what to keep, before the default decides for you. (🔗 Link)
🛑 Most Companies Can't Switch Off an AI Agent Gone Rogue
A Meta agent recently posted wrong technical information in public with no human approval and left sensitive data exposed for hours, and that was not even the month's first failure. Most enterprises running agents still cannot enforce what those agents are allowed to touch, and most cannot shut one down mid-task, which is why JPMorgan and Goldman built the approval layer before they built the agents. In this era, the human sign-off is the product. (🔗 Link)
Somewhere next quarter, an agent on your team ships something you never approved. A discount to the wrong list. A customer answer legal would have killed on sight. It happens in seconds, and it happens with your name on it.
Nobody has answered the obvious question yet. When that goes wrong, who owns it? Not the model. Not the vendor. Someone in your building, on a regular Tuesday, with their name on an approval that was never there.
Everyone is going to have the agents. Almost nobody will have decided where the humans still belong.
That decision, where AI runs loose and where a person still has to say yes, is the line we help draw with leadership teams. Training, advisory, and live buildathons that turn it into muscle memory instead of a policy doc nobody opens.
Signal over noise. Always.
Ralph & Gianni
Your prompts are leaving out 80% of what you're thinking.
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.
89% of messages sent with zero edits. Used by teams at OpenAI, Vercel, and Clay. Try Wispr Flow free — works on Mac, Windows, and iPhone.
📅 LET’S HANG OUT | Upcoming Events
📍 7/7 South Miami Tech Tuesdays @ Bougies
📍 7/14 Cafecito & Pastelitos @ maman West Palm Beach