70 years old, 5'9" 140 lbs, 25 years of caloric restriction. PRs: 20 pullups, 4 strict muscle ups. Legless rope climb.

Good news is good news; bad news is good news too 😁. But no one seems capable of asking a simple question: Why would Tesla need AI5, AI6, ... AI N if unsupervised FSD is going to be solved on AI4, eh? The answer is that, as I have been saying for months, unsupervised FSD will never be released on AI4. And don’t tell me it’s for humanoids—a car moving at 100 mph has drastically different latency requirements than a humanoid at 5 mph.

Elon Musk’s Risky Bet on Unsupervised Autonomy: A Reckoning for Tesla In October 2024, Elon Musk unveiled the Tesla Cybercab, a sleek, steering-wheel-less robotaxi that embodied his audacious vision for unsupervised general autonomy. The reveal was a culmination of a strategic pivot that began roughly a year ago, when Musk scaled back Tesla’s consumer car business—deprioritizing new plants and models—to go all-in on the Cybercab and Full Self-Driving (FSD) technology. Musk’s gamble rested on a single, unwavering belief: Level 5 autonomy, where vehicles operate without human intervention in any scenario, was imminent. “Tesla will soon release unsupervised FSD,” Musk would say, predicting cars “driving around by themselves, with no people in them” on public roads. Yet, as 2025 draws to a close, the reality is stark: unsupervised autonomy remains elusive, Tesla’s core business is faltering, and customer trust is fraying. Musk’s bet, while visionary, may have misjudged the complexity of autonomy and the cost of neglecting Tesla’s foundation. The Mirage of Level 5 Autonomy For over a decade, Musk has promised that fully autonomous driving is “next year.” In 2016, he claimed all Tesla vehicles had the hardware for full autonomy. In 2019, he predicted robotaxis would be operational by 2020. Each deadline has slipped, and the Cybercab was replaced with Robotaxi’s June 2025 trial in Austin, touted as a milestone, fell short of Musk’s “unsupervised” promise. X posts from June 2025 highlight the gap: users noted that Tesla delivered a supervised system with a human driver ready to intervene, far from the “nobody in the car” vision Musk hyped. The core issue is that Level 5 autonomy demands a human-like understanding of the physical world—an AI capable of navigating edge cases like construction zones, erratic pedestrians, or sudden weather changes. No AI lab, not even Tesla’s, has cracked this. Musk’s approach—scaling compute power and training data—assumes autonomy is a resource problem. But as critics argue, artificial general intelligence (AGI), the foundation for true autonomy, isn’t something you can force-schedule. It’s akin to trying to “make a baby in one month with nine women,” as the saying goes. The complexity requires breakthroughs, not just brute force. Tesla’s FSD, while advanced, still struggles with basic scenarios, as evidenced by the National Highway Traffic Safety Administration’s October 2024 investigation into 2.4 million Tesla vehicles after four reported FSD-related collisions. The Cost of the Pivot: Declining Sales and Eroding Trust Musk’s focus on autonomy has come at a steep price for Tesla’s core business. In January 2025, Tesla reported disappointing sales, with vehicle deliveries declining for two consecutive years. While Musk had projected 20% to 30% sales growth for 2025 in late 2024, the company quietly dropped that forecast in its Q4 results. Global deliveries fell from 1.81 million in 2023 to 1.79 million in 2024, with 2025 on track for further declines as competitors like BYD and legacy automakers flood the market with affordable EVs. Neglecting new consumer models and production capacity has left Tesla vulnerable, with its aging lineup—led by the Model 3 and Model Y—losing ground. Customer goodwill, once a Tesla strength, is also eroding. Tens of thousands of owners who paid $8,000 to $15,000 for FSD licenses feel misled by unfulfilled promises. A February 2025 class action lawsuit in California, now awaiting class action designation, accuses Tesla of misrepresenting FSD’s capabilities and the performance of its “Hardware 3” computer. The suit echoes a broader discontent: a June 2025 report noted 10,000 Australian Tesla owners joined a similar “phantom braking” lawsuit, highlighting global frustration. These legal battles threaten not just financial penalties but Tesla’s reputation as a trustworthy innovator.

I just took FSD 13.2.9 (2025.20.3) out for the first time. Within the first 2 miles, it made 4 serious mistakes. I didn’t intervene in any of them because the situations were too quick to react to safely. 1. I was at a stop, making a left turn onto a two-way street with partially obstructed visibility. There was a pickup truck coming from the left. The car hesitated, then darted out in front of the truck, close enough to be scary. A human driver would not do this, considering that no time was spent waiting for a safe opportunity, and the truck was the only vehicle on the road. 2. Less than half a mile down the road, it stopped at a red left-turn arrow light, then suddenly made the left turn without waiting for the arrow to turn green. 3. I was approaching an intersection where a bus was blocking the right lane. The left turn lane was occupied by a car waiting for the light. The car in front of me recklessly squeezed between the bus and the left-turn car. To my horror, my car followed the same risky maneuver. 4. About 300 yards down the road, there was an intersection with 3 lanes. The leftmost lane could only turn left, the second lane could go left or straight, and the right lane could only turn right. The car took the leftmost lane but then went straight. I stopped using FSD 13.2.8 around town a while back, but I don't recall it being as bad as 13.2.9.

Why True Unsupervised FSD Might Require AGI I’ve been reflecting on the current state of autonomous driving, especially after spending time with Tesla’s latest Full Self-Driving (FSD) software. While the progress is impressive, I’m starting to think that true unsupervised FSD — the kind that needs no human oversight — might be impossible without AGI-level reasoning. Here’s why. The Illusion of Competence Modern FSD systems are excellent at handling the routine. Highways, lane keeping, traffic lights — these are mostly solved problems under good conditions. But real-world driving is messy. It’s full of ambiguous, rare, and high-stakes scenarios — the so-called “edge cases.” These are things like: A construction worker waving you past a blocked lane with a non-standard hand signal. A child darting out from behind a van. A stop sign that's been spray-painted and partially obscured. Black ice forming just after sunset. These events aren’t just rare; they’re often unpredictable and demand contextual understanding, common sense, and even moral judgment. In other words, the kind of flexible cognition that humans naturally apply when we drive. Beyond Narrow AI Today’s FSD systems — no matter how advanced — are still forms of narrow AI. They rely on deep learning, large datasets, and heuristics. They can mimic decision-making, but they don’t understand the world. Yes, we can throw more data at the problem. Yes, we can train on millions of miles. But unless a system can reason like a human, it’s still just pattern-matching. And pattern-matching fails when reality throws something new at you. AGI: The Missing Link? This leads me to believe that fully autonomous driving — with no need for human intervention — might actually require Artificial General Intelligence (AGI). Only an AGI could: Understand novel, complex, or contradictory situations. Make split-second decisions using general knowledge and real-world reasoning. Weigh risk, interpret intent, and adapt on the fly — just like humans do. The Fragile Hack Without AGI, every workaround feels like a band-aid. More sensors, more data, better models — they help, but they don't fundamentally solve the problem. They just delay the failure to a rarer edge case. And if your system works 99.9% of the time, but fails catastrophically 0.1% of the time, it’s still not safe enough. That’s why, to me, anything less than AGI-level reasoning for unsupervised FSD feels like a fragile hack — impressive, but inherently dangerous.

I take back the flip "fool's errand" comment. Obviously, FSD v13 is very useful.

This is no different than teaching a humanoid to just move from point A to point B while avoiding obstacles. Without a multimodal AGI brain, such robot has limited utility.

My conversation with grok3 reinforced the correctness of my decision to sell all my Tesla shares at $332. You can skip to the FSD discussion. x.com/i/grok/share/ExX1sOERs…

8 minutes of pure gold!

There is a massive uniparty grift machine there in DC.

🇺🇸🫡

💯😀

The American people have had enough!! Absolute MIC DROP.. 🔥🔥💯

🇺🇸THESE ARE THE 158 DEMOCRATS WHO VOTED AGAINST DEPORTING SEX OFFENDERS Alabama: -Terri Sewell California: -Pete Aguilar -Ami Bera -Julia Brownley -Salud Carbajal -Tony Cárdenas -Judy Chu -Jim Costa -Mark DeSaulnier -John Garamendi -Robert Garcia -Sylvia Garcia -Jimmy Gomez -Jared Huffman -Ro Khanna -Sydney Kamlager-Dove -Barbara Lee -Ted Lieu -Zoe Lofgren -Doris Matsui -Kevin Mullin -Grace Napolitano -Nancy Pelosi -Katie Porter -Linda Sánchez -Adam Schiff -Brad Sherman -Norma Torres -Mike Thompson -Maxine Waters Colorado: -Jason Crow -Diana DeGette -Brittany Pettersen -Joe Neguse Connecticut: -Rosa DeLauro -John Larson -James Himes Delaware: -Lisa Blunt Rochester Florida: -Kathy Castor -Sheila Cherfilus-McCormick -Lois Frankel -Maxwell Frost -Darren Soto -Frederica Wilson -Debbie Wasserman Schultz Georgia: -Sanford D. Bishop Jr. -Lucy McBath -Henry “Hank” Johnson -Nikema Williams -David Scott Hawaii: -Ed Case -Jill Tokuda Illinois: -Sean Casten -Danny Davis -Jesús “Chuy” Garcia -Jonathan Jackson -Raja Krishnamoorthi -Robin Kelly -Delia Ramirez -Janice Schakowsky -Mike Quigley -Bill Foster -Brad Schneider -Lauren Underwood Indiana: -André Carson Kentucky: -Morgan McGarvey Louisiana: -Troy Carter Maine: -Chellie Pingree Maryland: -Steny Hoyer -Glenn Ivey -Kweisi Mfume -Jamie Raskin C.A. Dutch Ruppersberger -John Sarbanes -David Trone Massachusetts: -Jake Auchincloss -Katherine Clark -Bill Keating -Seth Moulton -Ayanna Pressley -Richard Neal -Lori Trahan -James McGovern Michigan: -Dan Kildee -Debbie Dingell -Rashida Tlaib -Shri Thanedar -Haley Stevens Minnesota: -Betty McCollum -Ilhan Omar -Dean Phillips Mississippi: -Bennie Thompson Missouri: -Cori Bush -Emanuel Cleaver New Hampshire: -Ann Kuster New Jersey: -Andy Kim -Rob Menendez -Donald Norcross -Bonnie Watson Coleman -Frank Pallone New Mexico: -Melanie Stansbury -Teresa Leger Fernandez New York: -Jamaal Bowman -Adriano Espaillat -Hakeem Jeffries -Yvette Clarke -Gregory Meeks -Alexandria Ocasio-Cortez -Jerrold Nadler -Nydia Velázquez -Paul Tonko -Dan Goldman -Ritchie Torres -Grace Meng -Joseph Morelle North Carolina: -Alma Adams -Valerie Foushee -Deborah Ross Ohio: -Shontel Brown -Joyce Beatty -Greg Landsman Oregon: -Earl Blumenauer -Suzanne Bonamici -Valerie Hoyle Pennsylvania: -Madeleine Dean -Mary Scanlon -Summer Lee Rhode Island: -Gabe Amo South Carolina: -James Clyburn Tennessee: -Steve Cohen Texas: -Greg Casar -Veronica Escobar -Joaquin Castro -Sylvia Garcia -Lloyd Doggett -Lizzie Fletcher -Al Green -Jasmine Crockett -Marc Veasey Vermont: -Becca Balint Virginia: -Donald Beyer -Gerald Connolly -Jennifer McClellan -Bobby Scott Washington: -Suzan DelBene -Derek Kilmer -Rick Larsen -Marilyn Strickland -Pramila Jayapal Wisconsin: -Gwen Moore -Mark Pocan Source: Newsweek

FDA pharma are pure evil.

Holy sh*t. I might actually be able to tell the truth on YouTube again. This will send shockwaves through all of big tech and the intelligence agencies

These people are running the government

She's amazing. Take time to listen!