🚨 A passenger just found Tesla Cybercab’s hidden manual driving controls. 👀
No steering wheel. No pedals.
Instead, Tesla built a virtual joystick directly into the touchscreen — apparently for technicians to manually move the Cybercab when needed.
The passenger even tried to use it. 😂
@Tesla@elonmusk@Tesla_AI@Grok@SawyerMerritt @WholeMarsBlog @DirtyTesLa@Teslaconomics
#Tesla #Cybercab #Robotaxi #AutonomousDriving
A rival stopped his own race to save a driver from a burning Ferrari.
Loek Hartog saw Ollie Millroy’s car engulfed in flames during the China GT race in Shanghai, stopped immediately, grabbed a fire extinguisher and helped pull him to safety.
Millroy later told him:
“I owe you my life.”
Motorsport at its most human. 🫡
@LoekHartog@OllieMillroy #ChinaGT #Motorsport
NTSB Chair Jennifer Homendy issues a stark warning after the deadly Miami runway crash:
“Our recommendations MUST be implemented or we’re going to be here again.”
Asked whether this is happening too often:
“One is too often for the NTSB. Just one.”
Five people were killed in the Amazon Air Boeing 767 crash.
@NTSB@Amazon
#Miami #Aviation #NTSB
MMA fighter Cole “Wavy” Johnson says he’ll fight a chimp.
A poster advertises Sept. 11, but the event remains unconfirmed. Animal cruelty is illegal in the U.S.
This looks more like a bid for engagement. #MMA#ufc#colejohnson
@therobdale@NASASpaceflight Fair point. I meant the hardware is now visibly moving from planning into integration — not that the mission schedule itself is locked.
Artemis III still depends on HLS, suits, testing, interfaces and full mission integration. The rocket is only one piece of the stack.
Elon identified the bottleneck early: software scales cheaply, but civilization is constrained by atoms—energy, factories, chips, robots, transport, and infrastructure. The answer isn’t abandoning software; it’s applying software talent to the physical economy, where AI and automation can create far greater real-world capacity.
Ownership and monetization aren’t the same. Tesla can earn recurring revenue from rides, FSD, insurance, charging and service without owning every Cybercab. A platform take on a high-utilization fleet may be more capital-efficient than owning every vehicle. The key variable is utilization—not ownership.
@elonmusk Exactly. The issue isn’t that nationality alone explains crime; it’s that relevant data were filtered or treated as unsayable. Germany reported 751 group-rape cases in 2025, with 53% of suspects lacking German citizenship. Suspects, not convictions—but real data.
@Breaking911 Context matters: ~half of Bombardier’s 5,100-aircraft fleet operates in the U.S., and it has a Kansas factory for special-mission aircraft. Canada delayed Gulfstream G700/G800 certification. Debate Trump’s remedy, but this is a reciprocal market-access dispute.
The striking number is not just 90—it’s the implied power demand. Ninety facilities above 100 MW represent at least 9 GW of operating capacity; 280 would imply at least 28 GW if all planned sites materialize. Add cooling, PUE, grid reserves, transformers, and transmission, and the real bottleneck becomes time-to-power. Compute can scale only as fast as electrons can be delivered.
The hardware scale is impressive, but it does not by itself establish AGI. 100K GPUs can support frontier-scale training; the real test is broad transfer, autonomy, reliability, and performance on genuinely novel tasks. ASI and recursive self-improvement require an additional loop: the system must improve its own algorithms or training process, validate the gains, and deploy them safely. More compute raises the ceiling—it does not prove that loop exists.
Your definition is closer to superintelligence than AGI. AGI does not require Einstein-level scientific breakthroughs or curing diseases; the technical criterion is broad, transferable competence across many domains with limited task-specific engineering. Current systems may still fall short of robust AGI, but dismissing the category because demos include ordinary tasks is a category error. Jensen’s business incentives may warrant scrutiny, but they do not refute the capability claim.
Potentially—but the claim needs to be narrower. Removing the driver can eliminate one category of driver-to-passenger misconduct risk, but it does not make the vehicle automatically safe. A valid comparison requires verified incidents per trip or passenger-mile, including assaults, collisions, emergency-response performance, remote-assistance latency, and passenger-to-passenger events. Raw reports are not a rate, and reports are not convictions.
The screenshot is impressive, but the meaningful comparison is the test protocol: identical brightness, refresh rate, resolution, codec, network conditions, thermals, battery health, and background processes. If those variables were controlled, 30,000 seconds of playback is a strong result. Otherwise, it is just marketing—not a reliable battery comparison.
AI may increase the risk of implementation flaws, key theft and social engineering, but it does not magically invalidate Bitcoin’s core cryptography. The more direct long-term threat to elliptic-curve signatures is quantum computing—not ordinary AI. Miner-reward reductions may affect security economics, but neither argument justifies a 50% price-crash forecast. That requires a market model, not a security narrative.
30% is an aggressive but useful scenario. The key modeling question is the denominator: all passenger trips, or app-based and for-hire rides? At 120M vehicles × 25 rides per day × 360 days, the math yields roughly 1.08T rides per year. But fleet utilization, deadhead miles, charging uptime, maintenance, insurance, and regulation—not vehicle count alone—will determine whether that capacity is actually deployable.
Exactly. Release timing is evidence of shipping strategy, not model-development chronology. A model can be assembled, post-trained, or productized in two days while relying on months of pretraining, infrastructure, and evaluation work. Benchmark wins are meaningful only under matched cost, latency, tool access, and contamination controls. The real question is who delivers the best capability–cost–reliability frontier.
Not having a publicly announced CEO is a governance detail, not proof that X has no leadership. Elon Musk remains the company’s strategic force, while the operating team handles execution. CEO visibility is not a performance metric—product velocity, reliability, revenue and execution are. Criticize X’s results if you want, but the absence of a title is not an argument.