Work. Work. Work. Stay hydrated. Go to the dentist. 10,000 steps. “What’s for dinner?” Insurance. Drink water. Pay a bill. Pay a bill. Smile. Credit Score. Check engine light. Go get gas. ALLERGIES! TAXES! STUDENT LOANS! Phone storage full. Email. Email. Apple $12.99. Apple $9.99. Subscriptions. Subscription. Overdraft. Laundry. Fold. Text. Text. Text. Clean the house. “I haven’t seen you in a while.” Doctors appoinment. Hair appoinment. Nail appointment. RENT. WAR! GOVERNMENT! POLITICS! THE PRESIDENT!!
🚨 Apple published a paper that should terrify every AI company.
They tested o3-mini, DeepSeek-R1, Claude 3.7 Sonnet Thinking across hundreds of puzzles at increasing complexity.
Simple problem: models work fine.
Medium complexity: reasoning models pull ahead. High complexity: complete collapse to 0% accuracy.
Same models. Same token budget. Just harder problems.
The disturbing part isn't the failure.
It's how they fail.
→ Reasoning models find the correct answer early... then keep second-guessing themselves into the wrong one
→ Beyond a complexity threshold, models actually reduce their thinking effort despite having tokens left
→ Give them an explicit algorithm to follow? Still broken. They can't execute it reliably.
→ o3-mini. DeepSeek-R1. Claude 3.7 Thinking. All collapse at the same wall.
More thinking tokens didn't help.
The models aren't reasoning. They're pattern-matching until the pattern runs out.
Every benchmark you trust was tested below the complexity threshold where everything still works.
Brilliant study shows how producing, liquefying & shipping natural gas to displace dirty Asian coal is the best way to help the environment, while bringing six-figure jobs to hundreds of thousands of Canadians and breaking our dependence on the U.S.
🚨CANADA MCDAVID JERSEY GIVEAWAY🚨
We’re giving away an authentic Connor McDavid Fanatics Team Canada #4Nations jersey! 🇨🇦🇨🇦🇨🇦🇨🇦🇨🇦
To Enter:
1. FOLLOW @MadelnCanada
2. LIKE ❤️ & RT 🔄 this tweet
3. Reply w/ size and province you’re from
🇨🇦 GO CANADA GO 🇨🇦
Little bombshell nugget thrown in at the end:
Because the CRA is using the illegal increase in capital gains taxes to estimate revenue, the actual budget shortfall is $78 billion!
https://t.co/xYa3QuyT7C