Elon Changed His Mind, AGAIN ⚡️

Technological concepts / product feature highlights

Tesla Cybertruck FSD handling of emergency vehicles (AV edge case)

• A clip shared by CJP shows a Cybertruck detecting an ambulance approaching from the rear at an intersection that was mostly blocked.
• FSD automatically executed a safe right turn across three lanes into a “safe area,” clearing a lane for the emergency vehicle.
• After verifying safety, the truck merged back and updated navigation automatically—starting from a complete stop.
• The narrator highlights the importance of this example amid ongoing regulatory debate about how AVs should behave around emergency vehicles.

Tesla AI chip memory spec change (AI5 / LP5) and cost/volume implications

• Elon reportedly changed his mind about AI5 memory capacity:
  • AI5 RAM reduced from 72 GB to 96 GB (contrasting an earlier plan where 96 GB was framed as a “nudge” upward).
  • Motivation: avoid Tesla being the only customer of the minimal-RAM configuration.
  • Rationale: standardizing memory versions helps chipmakers amortize costs of masks and packaging tools, preventing higher per-chip cost for one-off variants.
• Savings estimates (as described by the narrator)
  • AI5 was planned at 144 GB of LP5, now 96 GB → 48 GB less per chip.
  • Rough LP5 pricing: $13/GB raw die price
    • Estimated ~$630 per die (die-only).
  • With packaging/vehicle integration:
    • Estimated ~2× die price, for $1,200–$1,500 saved per AI5 chip.
• Scale math (hypothetical)
  • 1M Optimus units/year (one AI5 each): ~$600M–$1.5B avoided memory purchases.
  • 100k Optimus/year
    • ~$150M/year (one AI5 each), or ~$300M/year (two AI5 per board).
  • AI6 could save more, but LP6 pricing isn’t available yet.
• Capacity vs bandwidth distinction
  • The narrator emphasizes memory capacity (GB) is different from memory bandwidth (Gbps), and that capacity alone doesn’t tell the full performance story.

Tesla next-chip expectations for vehicles (AI4.1 / AI4+)

• The video argues AI5 chips aren’t expected in Tesla vehicles soon, based on Elon’s statements:
  • AI4.1 / AI4+ is positioned as the next vehicle chip.
• Reported memory/compute change:
  • RAM reportedly increases 16 GB → 32 GB per SoC
  • With two SoCs per board, that’s ~64 GB per board
  • Reported ~10% compute increase and memory bandwidth increase.

Unsupervised FSD “hardware fit” uncertainty

• The narrator stresses confusion around Elon’s claims that AI5 is optimized for edge compute in Optimus and RoboTaxi, and distinguishes:
  • RoboTaxi ≠ customer-owned vehicles
• Core concern:
  • Tesla likely must ensure unsupervised capability works on AI4, since millions of AI4-equipped vehicles are already on the road.
• Timing expectation:
  • Answers are expected in the next ~6 months, tied to real-world releases and performance.

Cybertruck / Cybercab / RoboTaxi operational updates (data points + incidents)

RoboTaxi and Cybercab deployment & app availability

• Tesla registrations in Texas reportedly include:
  • 589 robo taxi vehicles
  • 169 cyber cabs
  • (Source is referenced trackers; not treated as official.)
• App rollout is mentioned for countries including UK, Australia, Germany, Italy, Sweden, Singapore, etc.
  • Framed as possibly targeting attendees around Austin events rather than a full global launch.

Dynamic pricing + demand behavior

• Anecdotes from David Moss:
  • During busy Austin events: ~$3.75/mile
  • After ACL weekend: pricing dropped closer to ~$2/mile
• Incident example:
  • A service loop failure where a cyber cab got stuck trying to reroute out of a pickup parking lot.
  • Assistance intervened to stop and restart FSD.

Autonomy operating hours and safety priorities

• Elon reportedly expanded operating hours from 10pm → 11pm, targeting roughly 6am–11pm.
• Safety goal: avoid running over hard-to-see pets at night (example: “gray kittens on gray tarmac”).

Real-world obstacle handling (edge cases)

• A cyber cab faced a blocked road by a utility worker.
• It reportedly looped and eventually reacted once cones were removed—backing up and attempting maneuvers to escape the blocked position.

---

Review / sponsor / tutorial content

Sponsored exoskeleton product: Hypershell X Ultra S (fitness + comfort claims)

• Described as an “intuitive exoskeleton” with:
  • Fitness mode: adds resistance for higher calorie burn; targets hip flexors/leg muscles
  • Comfort features: “soft pack comfort system”
  • Weight: ~4 lb
• Claimed health metrics (as stated by the narrator/video):
  • Average heart rate down up to 42%
  • Oxygen consumption down up to 39%
• Battery and usage:
  • ~18 minutes per charge in eco mode (subtitles reportedly garbled; described as “18 m per charge”)
  • Roughly ~42,000 steps
  • Quick battery swaps
• Certification and specs:
  • SGS certification for consumer-grade product testing
  • Operating temperature range: -4°F to 140°F
• Control:
  • Mobile control via iOS/Android (and mention of watchOS)
• Promotion:
  • 15% off on Amazon and the official site for Oct 6–Oct 12

---

Business/market analysis included (context for product adoption)

Tesla delivery performance and inventory tightness

• Q3 deliveries: 486,532 vehicles, described as beating consensus.
• Claim: Tesla appears to be moving toward extremely low days of supply.
  • Anecdotes mention low inventory at Tesla locations.

Tesla Energy storage outlook

• Q3 energy reporting: 13.7 GWh
• Narrator expects quarterly totals to hover near ~15 GWh until Shanghai/Houston ramp further,
  • based on installed capacity vs production rate caveats.

Comparative EV sales pressure

• Mentions Ford/GM EV sales down significantly year-over-year (as cited in subtitles).
• Suggests Tesla momentum is partly driven by FSD adoption (“butts in seats”).

Germany BEV registration rebound

• Germany BEV sales reportedly increased strongly.
• Tesla registrations reportedly rose sharply and may have improved ranking vs BYD.

---

Main speakers / sources mentioned

• Dylan Lumis (video host/narrator)
• Elon Musk (referenced via reported statements)
• CJP (shared the emergency-vehicle FSD clip)
• Jason Oppenheim (appears in a referenced viral FSD endorsement anecdote)
• Dirty Tesla (credited for tipping the exoskeleton brand)
• Ashley / Phil Dwan / VP of Shopify / David Moss / Spencer / Amin / Sawyer / Victor / utility worker (participants or cited contributors across anecdotes/incidents)
• SGS (certification/testing company mentioned for the exoskeleton)