|
|
|
// YOUR AI EDGE
TEN MINUTES OF AI MADE PEOPLE MEASURABLY WORSE AT THINKING
Plus: The résumé black hole explained, Apple owes iPhone owners money, Google puts Reddit in search, and California’s plan to AI-proof jobs.
|
|
// TODAY'S SIGNAL
|
Ten Minutes of ChatGPT and the Wheels Come Off
Researchers at Carnegie Mellon, MIT, Oxford, and UCLA sat people down with GPT-5 and timed what happened next. After roughly 10 minutes of AI-assisted problem solving, participants had their AI taken away. Their performance cratered by 20 percent. They skipped questions at nearly double the rate. Math and reading comprehension both tanked. The effect was cognitive, not task-specific. The kicker: participants who used AI as a direct answer machine suffered the most. Those who used it for hints and clarification? No measurable impairment. The researchers called it "a heavy cognitive cost" and warned that current AI systems, optimized for short-term helpfulness, "risk eroding the very human capabilities they are meant to support." A 2025 MIT study on ChatGPT and critical thinking found the same thing. So did a separate Microsoft and Carnegie Mellon paper on AI-driven cognitive decline. Three research groups, same conclusion. The tool works. The person using it gets dumber.
|
// THE REAL STORY
"Write this for me" and "help me think through this" produce completely different cognitive outcomes. The study proved it. Default to hints mode. Make the AI argue with you, not work for you.
|
|
|
|
// SHORTCUT
Copy, paste, go:
|
"I need to [write a quarterly business review / draft a client proposal / prepare a budget summary]. Before you write anything, ask me 5 questions about the context, audience, and goals. Then outline the structure and let me approve it before you draft. I want to stay in the driver’s seat."
|
|
|
The Résumé Black Hole Finally Has a Name
A medical student applied to hundreds of jobs. Got almost nothing back. So he taught himself Python and investigated the systems screening him out. He’s got company. A Greenhouse report from May 2026 found that 63% of job seekers have encountered an AI interview. Of those, 51% never heard back. Not a rejection. Not a next step. Nothing. And 70% were never told AI was evaluating them. The screening side is uglier. Over 75% of résumés get filtered by applicant tracking systems before a human ever sees them. One study analyzing three major screening models found that résumés with white-associated names were preferred 85% of the time. Male names beat female names in 52% of tests versus 11%. Shorter résumés and less common names made the bias worse. Forbes reported in March that 98% of Fortune 500 companies now use an ATS. The system trained to find the best candidates is statistically screening out qualified people based on how a name reads to a pattern-matching algorithm.
|
// CAREER MATH
A machine decides your fate before any human knows you exist. The highest-ROI job search move right now: stop optimizing résumés for robots and start having conversations with people at the company. Employee referrals skip the ATS entirely. One 15-minute coffee chat is worth more than 50 cold applications.
|
|
|
Apple Owes iPhone Owners Up to $95
Apple agreed to pay $250 million to settle a class-action lawsuit over Siri’s AI features that never showed up. The company marketed Apple Intelligence as a headline feature of the iPhone 16 launch in September 2024. Buyers got writing tools and notification summaries. The Siri upgrade that was supposed to tie it all together? Still pending, two years later. The settlement covers roughly 37 million devices bought between June 2024 and March 2025: all iPhone 16 models plus iPhone 15 Pro and 15 Pro Max. Eligible owners can claim $25 to $95 per device, depending on claim volume. Apple pulled a Bella Ramsey ad showing the upgraded Siri in action. The National Advertising Division told them to stop saying features were "available now." Apple denied wrongdoing in a statement that read like it was written by the same Siri that still can’t set a complex timer.
|
// TRANSLATION
Apple sold a promise, collected the money, and delivered a progress bar. The per-person payout is small ($25 to $95), but the signal matters: companies can’t market AI features that don’t exist yet and call it a product. Check eligibility. File a claim. Two minutes, free money from a $3 trillion company.
|
|
|
|
🔴🟡🟢 RED LIGHT / GREEN LIGHT
|
🔴 // RED LIGHT
AI as Answer Machine
The Carnegie Mellon/MIT study found a 20% performance drop after just 10 minutes of using AI for direct answers. Outsourcing your thinking makes you measurably worse at it.
|
|
🟢 // GREEN LIGHT
AI as Sparring Partner
Same study, different result: people who used AI for hints and clarification showed zero cognitive decline. The tool is the same. How you use it changes everything.
|
|
🔴 // RED LIGHT
Cold-Applying Through ATS
75% of résumés get filtered before a human sees them, with documented racial and gender bias in screening models. Optimizing for the algorithm is a losing game.
|
|
🟢 // GREEN LIGHT
Employee Referrals
Referrals skip the ATS entirely. One conversation with someone at the company is worth more than dozens of online applications into the void.
|
|
| |
|
Google’s AI Search Now Quotes Reddit Users by Name
Google is pulling Reddit comments, forum posts, and social media quotes directly into AI Overviews, complete with creator names and handles. The logic: people already append "Reddit" to their searches when they want real answers. Google figured it should just do that part for them. AI Overviews are wrong about 10% of the time, according to a New York Times analysis. Google processes trillions of queries a year. That’s hundreds of thousands of inaccurate results per minute. And now Reddit comments (famously a coin flip between genuine expertise and confident nonsense) are feeding into it. Google says the update adds "a preview of perspectives from public online discussions" and will show "a creator’s name, handle, or community name." Useful for niche questions like "best budget headphones." Less reassuring when someone searches "is this medication safe" and gets a quote from u/DrFeelgood420.
|
// STEP BACK
Google is building a curated comments section into search. Fine for "best budget headphones." Genuinely scary for "is this medication safe." Click through to the actual source. Every time.
|
|
|
California Wants to Tax AI Companies to Create the Jobs AI Eliminated
California gubernatorial candidate Tom Steyer proposed a "token tax" on AI companies, charging a fraction of a cent per unit of data processed. The revenue would fund what he calls the Golden State Sovereign Wealth Fund, earmarked for healthcare, housing construction, and energy infrastructure jobs for people displaced by automation. The timing stings. For the second consecutive month, AI was the leading reason companies cited for layoffs, per outplacement firm Challenger, Gray, and Christmas. AI-related cuts accounted for 26% of 88,387 recorded layoffs in May, with 21,490 people losing jobs. A Bernie Sanders analysis from October projected AI could eliminate nearly 100 million jobs in a decade. Steyer’s plan also includes expanded unemployment insurance and a new AI Worker Protection Administration. Will it pass? Probably not in its current form. But a serious gubernatorial candidate running on AI job displacement as a platform issue is new. That part matters more than the policy details.
|
// CONNECT THE DOTS
AI job displacement went from think-tank white paper to campaign trail in under a year. The policy will lag the layoffs. It always does. Worth getting good at the stuff that’s hard to automate: reading a room, making calls on incomplete information, building relationships that don’t fit in a training dataset.
|
|
|
| |
|
// TOOLBOX
NAPKIN AI
What it does: Paste any text (meeting notes, a blog draft, a strategy doc) and Napkin AI generates polished infographics, diagrams, and visual summaries in seconds. Like having a graphic designer on call who reads faster than you write.
Why it’s worth 5 minutes: The free tier gives you 10 generations per month. Paste a paragraph from a quarterly review and get a clean visual you can drop into a presentation or Slack message. Flowcharts, comparison tables, stat callouts, no Canva required.
The honest caveat: Complex data with lots of numbers can produce visuals that look clean but misrepresent relationships. Double-check hierarchy and labels before sharing externally.
|
|
// WAIT... DOES THIS ACTUALLY WORK?
THE OBSCURE PROMPT OF THE DAY
prompts nobody asked for. results nobody expected. try it anyway.
|
"Act as a brutally honest efficiency consultant from the 1920s who has been transported to 2026. I’ll describe my typical workday and you tell me, in character, which parts would horrify a person from 100 years ago and why. Be specific and opinionated."
|
// OUR VERDICT
Claude went full Henry Ford-era industrialist. "You spend forty-seven minutes each morning reading messages from people who sit fifteen feet from you. In my day, we called that 'being afraid of eye contact.'" It flagged three meetings as "committees formed to discuss the existence of other committees." Best part: it calculated that my actual productive output occupied roughly 3.5 hours of an 8-hour day and said, "In 1926, we would have paid you for 3.5 hours." Cold. Accurate. Forwarded it to my entire team.
SURPRISINGLY PRACTICAL: ★★★★☆
|
|
// YOUR EDGE
|
01
Learn this: AI-assisted thinking vs. AI-dependent thinking. The Carnegie Mellon/MIT/Oxford study found zero cognitive decline when people used AI for hints versus answers. Completely different brain outcomes from the same tool.
|
|
02
Watch this: AI hiring lawsuits are coming. 75% of résumés never reach a human. Documented racial and gender bias in screening algorithms. The Greenhouse report and Forbes data are worth bookmarking now.
|
|
03
Say this: "Did you know AI was the number-one cited reason for layoffs two months in a row? A California governor candidate is running on a platform to tax AI companies and fund a jobs guarantee." (Drops well in any meeting about workforce planning.)
|
|
|
// GOT THIS FROM A FRIEND?
Your edge on AI, twice a week. Free forever.
Subscribe â
|
|
|
|
