|
|
|
// YOUR AI EDGE
THE WHITE HOUSE IS NOW DECIDING WHO GETS TO USE THE NEXT CHATGPT
Plus: a chatbot becomes Exhibit A in an arson trial, Claude quietly poaches paying customers, Apple blames the robots for a $300 price hike, and a reminder that the hard part of the job is still yours.
|
|
// TODAY'S SIGNAL
|
THE NEXT BIG MODEL IS SHIPPING WITH A BOUNCER AT THE DOOR
OpenAI built the most anticipated model of the year. Then the government decided who gets to use it. The Trump administration, worried about security risks, asked OpenAI to stagger the rollout of GPT-5.6. Sam Altman told staff in a company Q&A that the model would ship as a limited preview to a small group of enterprise customers, with the government signing off on access customer by customer. Read that twice. A federal agency now has a say in which companies get to use a commercial chatbot first. It’s a softer version of what hit Anthropic, which got an outright export-control order this month suspending access to its top Mythos 5 and Fable 5 models. OpenAI gets a velvet rope. Anthropic got a locked door. The administration that promised a “hands-off” approach to AI is now standing in the doorway of the most powerful tools, picking who walks through. The cyber risk is real. The precedent is the part to watch.
|
// CONNECT THE DOTS
The AI boom isn’t just an abstraction on tech-stock charts. It’s reaching into your wallet through the price of ordinary hardware. If you’re planning a work-laptop or phone upgrade, the smart timing play is to buy what you need now rather than waiting for prices to “settle.” Memory pricing follows the data center buildout, and that buildout is accelerating, not cooling. Lock in the tool before the next round of hikes.
|
|
|
|
// SHORTCUT
Copy, paste, go:
|
"Act as a procurement analyst. I’m evaluating [TOOL NAME] for [SPECIFIC TASK, e.g. summarizing client contracts]. List the data-handling and access questions I should ask the vendor before we feed it real company data, ranked by how likely each is to be a dealbreaker. Keep it to a checklist I can read in a meeting."
|
|
|
THE CHATBOT JUST BECAME EXHIBIT A
Your chat history has always felt private. A Los Angeles courtroom just treated it like a diary read aloud to the jury. Jonathan Rinderknecht faced arson charges for a New Year’s Day 2025 fire that became one of the deadliest in LA history. Prosecutors built their case on iPhone location data, security footage, and something newer: his ChatGPT logs. They told the jury he’d asked the chatbot to generate images of fire, typed “Why am I so angry all the time?”, and ranted to it about the wealthy. The jury wasn’t buying it. They deadlocked 10-2 for the defense and the judge declared a mistrial. One juror told CBS LA she didn’t think the logs proved anything: “I talk to ChatGPT all the time.” She said it made her angry that anyone would treat using a chatbot as a character flaw. So the first big test of “AI conversations as evidence” failed because the jurors recognized themselves in the defendant. That recognition won’t last forever.
|
// STEP BACK
Every company rolling out an AI chatbot for customer service or marketing just inherited a new kind of risk. If your team is shipping AI-generated content with a brand name attached, someone needs to own accuracy, and that someone is increasingly a job. “AI quality control” is going from a nice-to-have to a line item. If you want to be useful in any content, comms, or legal-adjacent role, get fluent in where these tools break and how to catch it. That fluency is about to be very employable.
|
|
|
CLAUDE IS QUIETLY EATING CHATGPT’S LUNCH
Everyone assumes ChatGPT owns the consumer AI market. The credit card data says someone is sneaking up behind it. A firm called Indagari analyzed billions of anonymized transactions from about 28 million US consumers. Claude’s paying customers and revenue are up roughly 75% since January. The gains kept climbing even after a March surge, when Anthropic refused to let its models be used for government mass surveillance. The teaching data backs it up. On DataCamp, a learning platform with 20 million users, “Claude” is now the most-searched term on the whole site, beating even the word “AI.” Demand for Claude courses is outpacing ChatGPT three to one among self-directed learners. The lesson for anyone building an AI habit: the default tool isn’t always the best one. Claude tends to win on writing and long documents, ChatGPT on breadth and image generation. People who only know one are leaving the other half on the table.
|
// CAREER MATH
Fluency in one chatbot is a skill. Fluency in choosing between two is a better one. Spend an hour this week running the same real task, a client email, a messy spreadsheet, a tricky edit, through both Claude and ChatGPT. Notice which one nails it. Then you can walk into any meeting and say which tool fits which job, instead of defending the one you happened to start with. That judgment is the thing that’s hard to automate.
|
|
|
| |
|
APPLE SAYS THE ROBOTS MADE IT RAISE YOUR PRICES
Apple just made its gadgets more expensive and pointed at the AI boom as the culprit. Tim Cook called the increases “unavoidable” and the old pricing “unsustainable.” The 16-inch MacBook Pro went up $300. The 11-inch iPad Air jumped from $599 to $749. Even the little HomePod Mini got a $30 bump. The reason is memory. The whole industry is racing to build AI, AI runs on RAM, and RAM prices have skyrocketed as data centers buy up the supply. Tim Derdenger, a marketing and strategy professor at Carnegie Mellon, called it “basic economics.” Here’s the part that stings. Apple is posting record earnings. The AI gold rush is happening on data center balance sheets, and the bill is landing on the price tag of a laptop you actually need for work.
|
// THE MOVE
Treat AI product recommendations like a stranger’s hot tip, not a verdict. When a chatbot pushes a specific brand, supplement, or vendor, ask it for its sources and then go look at them yourself. For anything that touches your money or your body, that 30-second check is the whole game. The people who blindly buy what the AI suggests are the exact target this trick was built for.
|
|
|
CHATGPT GETS YOU 80% THERE. THE LAST 20% IS THE WHOLE JOB.
For all the panic about AI taking jobs, a quieter truth keeps showing up: the chatbot does the easy part and hands you the hard one. It can draft the email, summarize the report, and rough out the deck in seconds. What it can’t do is know which client hates jargon, which number your boss will question, or which slide to cut because the room only has ten minutes. PCWorld’s Ben Patterson made the case that the final stretch, the judgment, the taste, the context, is still entirely human work. Even the companies building AI are funding the cleanup. A new nonprofit called Raise US, led by former Commerce Secretary Gina Raimondo and backed by Amazon, Anthropic, Microsoft, and OpenAI, has raised $500 million for worker retraining and wage insurance. The people who thrive aren’t the ones who refuse AI or the ones who let it run unsupervised. They’re the ones who get great at the 20% the machine keeps getting wrong.
|
// THE REAL STORY
Your value is moving to the final 20%, so build there. Pick the part of your job that needs taste or judgment, the editing, the prioritizing, the reading-the-room, and get visibly excellent at it. Let AI eat the busywork and put your name on the calls only a human can make. “I use AI for the draft and I own the decisions” is the sentence that keeps you employed while everyone else argues about whether to use it at all.
|
|
|
| |
|
// THE TOOLBOX
GOBLIN TOOLS
Goblin Tools is a set of tiny, free AI utilities built for people whose brains don’t run on tidy linear lists. The star is the “Magic ToDo,” which takes a vague, overwhelming task like “plan the team offsite” and shatters it into bite-sized steps, with a spiciness dial that controls how much detail you get. There’s also a tone-checker that tells you whether your message reads as rude before you hit send.
The caveat: it’s deliberately small and a little janky, with no accounts and no polish. That’s the point, but don’t expect a slick dashboard. Paste in the task you’ve been avoiding and crank the spiciness to max.
|
|
// WAIT... DOES THIS ACTUALLY WORK?
THE UNHINGED PROMPT OF THE DAY
prompts nobody asked for. results nobody expected. try it anyway.
|
"You are my overly dramatic Italian grandmother. I will paste my weekly schedule. Roast my time management, guilt-trip me about not calling, and then tell me which one thing to cut so I can rest. End by telling me to eat something. [PASTE SCHEDULE]"
|
// OUR VERDICT
It opened by asking why I work like I’m being chased, correctly flagged that four back-to-back “syncs” on Tuesday were a single meeting in a trench coat, and told me to cancel the 8am. Then it guilted me for not calling. I cut the meeting. I also texted my actual grandmother.
ACTUALLY USEFUL: ★★★★½
|
|
// YOUR EDGE
|
01
Learn this: The phrase “staggered rollout.” It’s the new normal for frontier AI, where the most powerful models ship to a vetted few before everyone else, sometimes with government sign-off. Knowing that access is now a competitive advantage, not a given, changes how early you want to be.
|
|
02
Watch this: Whether Claude’s consumer surge holds. If a credible number two emerges in a market everyone assumed ChatGPT had locked up, every team that standardized on a single AI tool will have a reason to re-shop. Tool monogamy is about to look risky.
|
|
03
Say this: "Memory prices are climbing because of the AI buildout, and it’s already hitting hardware costs. If we’re budgeting for new machines or cloud, we should lock pricing now instead of assuming it settles." You’ll sound like the person who reads past the headline.
|
|
|
// GOT THIS FROM A FRIEND?
Your edge on AI, twice a week. Free forever.
Subscribe →
|
|
|
|
