-
22 maggio, di Team — Artificial Decisions
An AI Adds Typos to Our Emails to Make Us Look Human
We spent years teaching machines to write well. Now we pay another machine to mess up what they wrote.
It's called Sinceerly. And no, it's not misspelled by accident. It's a Chrome plugin that takes your AI-polished email and puts back the typos, the contractions, the broken sentences. So it looks like you wrote it yourself, in a rush, from your phone.
Three modes. The first is Subtle, it strips filler words and drops in a contraction here and there. One typo in the first line, just because. The second is Human, it makes the tone more conversational. Another typo in the first line, obviously. The third is called CEO: all lowercase, very short sentences. If you don't have a signature, it sometimes throws in a "sent from my iPhone." So you sound real. Someone with no time to write properly.
For centuries typos meant "you're careless." Now they mean "I'm real, I wrote it myself." We flipped the meaning of words.
What do you think?
#ArtificialDecisions #MCC #AI
-
30 aprile, di Team — Artificial Decisions
CFOs Are Discovering AI Costs More Than the People They Fired
CFOs are discovering an embarrassing problem. AI tokens cost more than the salaries of the people they fired to put AI in their place. Here in the States a well-known investor said it out loud: 300 dollars a day for a single AI agent, that's 100,000 dollars a year. And that agent was running at 10 or 20% of what a human would do.
When does the cost of tokens overtake the salary of the employee? Boardrooms, meetings and work dinners here in New York keep asking the same question. We're already past that line. Other founders I talk to say the same thing: an AI agent has to be at least twice as productive as a human to pay for itself, otherwise the cash burns out.
Per-token prices have collapsed, total bills have exploded. 73% of companies blow through their AI budget. Average budgets jumped from 1.2 million to 7 million dollars in two years. Some Fortune 500 firms pay tens of millions a month just in inference.
In 2026 around 500,000 layoffs in the US are attributed to AI, nine times last year. But over 80% of those same companies admit AI hasn't yet produced measurable productivity gains. Many firms are blaming AI for layoffs that have nothing to do with AI. They call it AI washing. Do you remember greenwashing?
CFOs are starting to see the numbers on the screen. And they're starting to understand. What do you think?
#ArtificialDecisions #MCC
-
28 aprile, di Team — Artificial Decisions
Meta Helps Parents Talk to Their Kids About AI
Almost no parent I know has any idea what their kids actually do with AI. Homework? Small talk? Stuff they wouldn't tell anyone else? Meta figured it out. Or maybe they figured out it was getting expensive in court (last month they lost a child safety case in the US, 375 million dollars). Either way, they shipped something useful, and we must say it.
Parents supervising teen accounts on Facebook, Instagram and Messenger now see a new "Insights" tab. They don't read the messages. They see the topics their kids discussed with Meta AI over the last seven days. School, travel, entertainment, health, wellbeing. Tap on "health" and you get fitness, physical health, mental health. If your teenage daughter is talking to a chatbot about mental health, it's probably time you talked too.
Knowing isn't enough. You need a way to open the conversation without the kid shutting down. Meta worked with the Cyberbullying Research Center on eleven questions. They're well done. No lecturing, no interrogation.
"What's the most useful thing AI has helped you with?" You start from the positive, because kids assume by default that the adult wants to scold them.
"Have you ever asked AI something because it was easier than asking a real person?" This is the most important one. You're telling your kid they can come to you even for the awkward stuff.
The "Insights" have a big hole though. They say "mental health" but don't tell you if your kid is looking up breathing exercises or something much worse. The real alerts, the ones on self-harm, Meta is still developing.
What do you think?
#ArtificialDecisions #MCC
-
19 aprile, di Team — Artificial Decisions
Newspapers and TV News Won't Tell You This!
Sounds like clickbait, but you know me. If I'm writing something like this, there's a serious problem. They should be talking about it every single day, TV news and newspapers. There's a revolution happening and nobody's covering it. And it's coming like a tsunami.
See these people? We're inside an Indian clothing factory and every single worker has a camera on their head. They're training the Artificial Intelligence that will power humanoid robots. Robots that will learn to do exactly what these workers do. Intelligent robots with AI can learn any manual job the same way ChatGPT learned to write anything. ChatGPT and the others copied what we humans wrote for thousands of years. Now robots are copying what we've done for thousands of years. And they'll be just as good at it as AI is already good at writing text and sounding intelligent.
Elon Musk has paused production of some car models to go all in on intelligent humanoid robots.
They'll be in our homes, our restaurants, our bars, our factories. They don't get tired and they never sleep. They go recharge their own batteries. If one breaks, other robots fix it.
A tsunami we need to face. I'm sure there will be a way through, just like there was with previous revolutions. But you see how it's not possible that I'm the only one talking about this constantly, while newspapers and TV news cover society gossip that means absolutely nothing?
The clickbait title. Yes, it grabs clicks. But for a good reason. Share this video with everyone you know.
#ArtificialDecisions #MCC #Robot
-
15 aprile, di Team — Artificial Decisions
Here's an Example of AI Agents in Action That I'm Sure Will Inspire You
A guy who sells pools in Florida took OpenClaw, an open-source AI agent, and programmed it to find homes without a pool and convince homeowners to build one. He wrote the instructions, hit enter and went to sleep.
The agent starts from satellite imagery. It scans lots one by one in the Tampa area, measures the land, checks zoning constraints and decides if there's room for a pool. If the lot qualifies, it keeps going on its own. It generates a rendering: the homeowner's backyard seen from above, with a pool already in it. It calculates the construction cost and how much the home value goes up.
Then it searches public records for the homeowner. Finds the listing agent tied to the property. Prepares a postcard with the aerial rendering and a QR code. Ships it through Lob, an automated print and mail service. Three days later it's in the mailbox. On top of that the AI agent automatically creates a personalized website just for that homeowner, with all the details, reachable from the QR code.
What used to take months of work gets done in one night while the seller sleeps. The agent runs, scans new homes, produces new postcards, mails them out. Every homeowner gets a different message, built on their actual house, with numbers calculated on their lot. One single AI agent doing the job of an entire team: real estate analyst, graphic designer, copywriter, mailing service and salesperson. For a few dozen dollars in API costs.
And it doesn't just work for pools. Same logic applies to solar panels, roofing, fencing, landscaping. Anything visible from above and sellable door to door, without knocking. As I always say, the first to be replaced by AI are those who don't use AI. What do you think?
#ArtificialDecisions #MCC