AI researchers just admitted they messed up. Big time.
Leading scientists demand a complete reboot of how we build AI systems. Today's models excel at parlor tricks but fail at basic logic. They're like that overconfident friend who never lets facts get in the way of a good story. 😅
The problems run deep. AI systems devour energy like bitcoin miners on steroids. 🔥 They also can't tell truth from fiction.
Stay curious,
Marcus Schuler
PS: Today we gave our newsletter pics the Renoir-Monet treatment. Because nothing says "AI crisis" like an Impressionist makeover. 🎨
AI's Identity Crisis: Scientists Call for Complete Research Overhaul
Credit: GPT 4o
The world's leading AI researchers just dropped a bombshell report. Their message? We need to completely rethink how we build artificial intelligence.
The AAAI 2025 Presidential Panel didn't mince words. Current AI systems excel at party tricks but fumble basic reasoning. They spew facts with confidence, yet struggle to tell truth from fiction. Like a student who memorized the textbook but missed the point entirely.
The report tackles 17 core challenges. Top of the list? Getting AI to think logically. Today's language models prefer plausible-sounding nonsense over proven facts. They're that friend who always has a great story - even if it's not quite true.
Hardware poses another headache. Modern AI breakthroughs drain power like a teenager's gaming PC. Researchers dream of running complex models on smartphones, but current systems demand industrial-grade cooling just to stay alive.
The experts saved their sharpest criticism for AI evaluation. Current testing resembles grading a student solely on multiple choice tests - it looks good on paper but misses real-world skills entirely. They demand more rigorous standards.
Why this matters:
The AI community admits current systems are built on shaky foundations - like constructing skyscrapers without checking the soil
Academic researchers fear big tech companies are hoarding AI advances, turning scientific progress into corporate secrets
Sam Altman & The Messiest Tech Coup That Never Was
Credit: GPT 4o
In November 2023, OpenAI's board fired CEO Sam Altman without warning. The tech world lost its mind. Yet this wasn't just another Silicon Valley drama - it was a masterclass in corporate chaos.
The spark came from an unlikely source: OpenAI's own Chief Technology Officer, Mira Murati. She had grown tired of Altman's management style and his habit of pitting senior employees against each other. When she voiced concerns, Altman brought HR to their meetings. Not exactly a trust-building exercise.
Chief Scientist Ilya Sutskever joined the resistance. He compiled a detailed dossier on Altman's alleged deceptions, complete with screenshots from Slack. The evidence landed in board members' inboxes via self-destructing emails - because nothing says "we need better governance" quite like Mission Impossible-style communication.
The board struck on November 16th. They fired Altman and asked Murati to step in as interim CEO. Plot twist: Murati agreed, then promptly led a revolt against the very board that promoted her. Nearly every employee threatened to quit unless Altman returned.
By Monday, both Murati and Sutskever had signed a letter demanding Altman's reinstatement. The coup against the king had failed. The king's enemies now pledged their loyalty. Shakespeare couldn't have written it better.
Why this matters:
The world's most important AI company almost imploded because its board tried to enforce actual oversight
In tech's version of Game of Thrones, even the rebels end up backing the leader they tried to overthrow
Google just flipped the AI world on its head. Their most powerful model, Gemini 2.5 Pro, now comes free for everyone. No credit card required.
This move shocked industry watchers. Just days ago, accessing 2.5 Pro demanded a $20 monthly subscription. Now Google hands it out like candy at a parade. The model tops performance charts and excels at everything from math to science.
The free release brings serious muscle: file uploads, app integration, and Canvas support. Users can test these features on Google's website now. Mobile apps will follow soon.
Credit: Google
But don't tear up your subscription just yet. Google stays mysteriously quiet about the future. This "experimental" phase might be a clever taste test - get users hooked on premium features, then yank them back behind the paywall.
Meanwhile, paying customers still enjoy some exclusive perks. Their AI can process texts longer than "War and Peace." They also dodge usage limits that might frustrate free users.
Why this matters:
Google weaponizes its best AI against competitors by making it free - imagine bringing a tank to a knife fight
The "experimental" label suggests this generosity comes with an expiration date - enjoy the caviar while it lasts
Cover diverse regions (not just Western countries)
Add a brief explanation after each answer
Provide difficulty rating for each question (easy/medium/hard)
Include final score interpretation
Make questions engaging and visual where possible. Aim for real learning, not just memorization.
AI & Tech News
Dr. iPhone Will See You Now
Apple wants to turn your phone into an AI physician. Their new health coach promises personalized medical advice without the waiting room. Just don't expect this digital doc to write prescriptions - or have a better bedside manner than Siri.
China's New AI Star Wants Your Cash (and Lots of It)
China's hottest new AI assistant just hit users with a cold reality: pay up or scale down. Manus slapped a $39 monthly fee on its service faster than you can say "beta testing." Their premium tier costs $199 - matching OpenAI's prices while still working out the bugs.
Chinese Startup Zhipu Crashes the AI Party with $0 Price Tag
Zhipu AI just crashed Manus's premium party. While rivals charge up to $199 per month, Zhipu's new AI agent costs exactly zero dollars. Nothing says market disruption quite like the price of a cup of air.
Silicon Valley's Top Salesman Now Sells AI Panic to DC
A venture capitalist's protégé just became America's AI strategist. Michael Kratsios spent years in Silicon Valley's hype machine - now he peddles digital doomsday scenarios to Congress. China's DeepSeek triggered his perfect pitch: cheap AI plus stolen chips equals national emergency.
China's Robot Army: More PR Than R2-D2
China wants to dominate humanoid robots. But their dancing droids can barely last a TikTok video without recharging. While Chinese robots struggle through marathon photo-ops, real factory work moves at human speed - if you're patient enough to wait 12 seconds for a robot to load a truck.
DEI Goes Undercover at Big Tech
Tech firms changed their diversity programs' names but kept the work. Google turned "Chief Diversity Officer" into "VP of Googler Engagement." Same job, safer title. Welcome to corporate America's latest stealth mission.
Space Race Gets Personal: Apple vs Musk
Apple and SpaceX are fighting over satellite rights like billionaires squabbling over beach houses. After SpaceX launched hundreds of satellites, Apple responded by dumping $1 billion into rival Globalstar - proving that even the galaxy isn't big enough for two tech egos.
Google's AI Drug Lab Snags $600M to "Solve All Disease"
Google's AI drug division just raised $600 million to turn biology into code. Their CEO's modest goal? Eliminating all human disease - though they might want to cure hubris first.
French Regulators Whack Apple with €150M Privacy Fine
Apple claims to protect user privacy. France just slapped them with a €150 million fine for doing it wrong. The tech giant forced apps to bombard users with consent pop-ups, then acted shocked when regulators popped their bubble.
🚀 AI Profiles: The Companies Defining Tomorrow
Cursor: AI Coding Assistant Reshaping How We Build Software
Cursor transforms coding with an AI-powered editor that slashes development time and amplifies creativity. Built on VS Code, it gives devs superhuman abilities to navigate, generate, and refactor code at unprecedented speeds.
THE FOUNDERS 👨💻👨💻👨💻👨💻
Michael Truell, Sualeh Asif, Arvid Lunnemark, and Aman Sanger launched Anysphere Inc. in 2022. All MIT grads with elite tech backgrounds, they built Cursor to eliminate the "molasses" in software development. Now based in San Francisco with a rapidly growing team, they're making their "magical tool" vision reality.
THE PRODUCT 🚀
Cursor fuses a VS Code fork with advanced AI capabilities that go way beyond autocomplete. It generates entire functions from natural language prompts, answers questions about your codebase, performs smart refactoring across files, and even features an "Agent mode" for autonomous coding tasks. SOC 2 compliant with privacy modes for sensitive code.
THE COMPETITION 🥊
Cursor battles GitHub Copilot (Microsoft), Amazon CodeWhisperer, Codeium, Tabnine, and Replit Ghostwriter in the crowded AI coding space. It differentiates through deeper IDE integration, enhanced capabilities beyond simple suggestions, and its 2024 acquisition of Supermaven brought in proprietary model tech to reduce reliance on external AI providers.
FINANCING 💰
Cursor's meteoric rise attracted elite investors—OpenAI Startup Fund, Andreessen Horowitz, Thrive Capital, and tech luminaries like Stripe's Patrick Collison. After an $8M seed (2023), $60M Series A (August 2024), and $100M Series B (December 2024), it hit a $2.6B valuation. Word is they're now courting investors at a staggering $10B valuation.
THE FUTURE ⭐⭐⭐⭐⭐
From $0 to $150M annual recurring revenue in under two years, Cursor's growth trajectory is nearly unheard of. The team's relentless innovation and deep tech expertise position them to dominate the AI-assisted development space. Expect expanded enterprise features, specialized models, and possibly acquisitions of complementary tools. They're not just building a product—they're reshaping how the world creates software. 😎
