Tropic

Tesla: a world of amazing abundance. SpaceX: making life multiplanetary. Same dream, two companies, one future worth watching. That's why I follow both, minus the fluff.

Tesla gave up two things to get its driver assistance onto Finnish roads: the words "self-driving" and most of its speed headroom. Finland's transport agency Traficom recognized the Dutch provisional approval on Nov. 10, so Tesla can offer the system there right away under its European name, Tesla Assisted Driving. The trade-off is spelled out in the Dutch regulator RDW's Nov. 3 notice: the maximum speed offset drops from +50% to +10% over the limit, and it resets to 0% every time the car starts. Counting RDW's list, Finland is the 11th EU country to accept it. My read: Tesla is choosing access over settings, and that's the right call. A smaller speed offset costs owners little next to having the system at all, and every country added builds a track record before the EU-wide vote Traficom expects this winter. The risk is real, though. Reuters notes none of the EU's largest states has backed it yet, and a rejection at EU level would void the national approvals.

Elon Musk just admitted Grok 4.8 won't beat Anthropic's best. That candor is the signal: this release is about the foundation, not the scoreboard. His team dumped JAX and SGLang and rewrote the entire RL and inference stack in C/C++. Pre-training is done, 4.8 is in RL, and a limited Grok Bot rollout is due this week. Versus Opus 5.5? "Not as good." Grok 5 should pass it "maybe" in 2 or 3 months. My read: owning every layer is the SpaceX playbook. The payoff is faster iteration, not one release. The test is Grok 5 hitting that window.

The firm piece today is the first line. Singulus says the initial turnkey HJT cell line is an order of over €60 million, delivering in the first half of 2027, with the partnership meant to scale past 100 GW of equipment as SpaceX ramps satellite production in Bastrop. What stands out to me is the cell choice. Space power has historically run on expensive multi-junction cells, while HJT is a high-efficiency silicon process built for terrestrial volume. My read is SpaceX is betting it can make cells good enough for orbit at a cost that works for a constellation measured in very large numbers. SpaceX also funded Singulus in March and again in October through a secured loan. That looks like the Starlink playbook: lock in the bottleneck supplier early, then run the process in its own Texas factory.

Respect @KatieMiller

Key detail: this isn't approval yet. The FCC only accepted the applications for filing. Comments are due Dec 1, oppositions Dec 8, and the transfer still needs FCC consent. Satellite-to-phone use on this 800 MHz band is a separate SpaceX request that's still pending. No price has been disclosed; the ~$8B figure is unconfirmed.

The battery is the headline, but the efficiency math may be the bigger story. Starlink's own spec sheet puts the original Mini at 25–40 W average. At that draw, 4.5 hours would take roughly 113–180 Wh. Your list shows a 71 Wh pack, which works out to about 16 W average (71 ÷ 4.5). If both numbers hold up, Mini 2 draws well under half what Mini 1 does, and that's what makes small solar panels and ordinary power banks realistic. The caveat: firmware leaks reported back in July showed a much lower transmit duty cycle (about 11% vs 75%). That's unconfirmed, but if it's right, upload-heavy use like video calls is where the trade-off would show, and "up to 4.5 hours" is likely a light-use number. Also worth knowing before anyone rushes to order: Starlink says US preorders are open by invitation for now.

When can we expect CyberCabs to launch in the northern states?

Tesla's San Antonio Robotaxi map is the biggest it has ever launched with. The number that matters now is how many cars are actually inside it. Tesla's @robotaxi account posted the launch geofence just after midnight Monday, with rides starting at 5 a.m. Central. Robotaxi Tracker, a community site that measures operator maps, puts the zone at about 274 square miles, more than four times Waymo's 59 square miles in the city. Tesla hasn't said how many vehicles serve it, and the airport sits inside the outline but isn't served yet. My read: a map is the easy part. In other cities, Tesla's real pickup zones have run about 59% to 78% of the posted maps, per Electrek and Not a Tesla App. Wait times and cars on the road over the next few weeks will show whether this is a real head-to-head with Waymo or mostly a bigger outline.

Terafab's 2028 timeline now leans on Japanese toolmakers as much as on Musk. Nikkei reports, citing people familiar with the matter, that Terafab procurement staff met Japanese equipment suppliers in early November, and some got delivery deadlines only months out. One supplier told Nikkei the order is unreasonable but they'll try to meet it anyway. The scale is the bigger tell. Per Nikkei, Terafab informally told suppliers it wants about 1 million wafers a month after 2032, roughly 700,000 for memory and 300,000+ for logic. That's 40 times the 25,000 a month Rapidus is targeting at its Hokkaido plant. Musk himself called volume production around Q3 or Q4 2028 "very aggressive dates" on SpaceX's October call. The nearer checkpoint is a working test chip at the Austin research fab around May or June 2027. That's the date I'd watch, not the 2032 number.

Jensen applied first principles.

Tesla stretched the Model Y into a six-seater and still got drag down to 0.216. On Jay Leno's Garage today, vehicle engineering VP Lars Moravy said the rear spoiler virtually lengthens the car, and called hitting that number pretty amazing.

SpaceX's launch chief says Grok spots hidden debris better than inspectors do. Kiko Dontchev told last week's UP.Summit that people reviewing borescope video are about 94% accurate, while Grok hits 99.9% on the same footage, per Ars Technica. SpaceX hasn't published the test.

The bigger first is catching the ship too. Flight 15, targeting Friday, November 13 pending approval, is SpaceX's first try to bring both stages back to the tower. Elon said if both work, it would be the first time every part of an orbital rocket returns intact.

Tesla's Texas solar plant has a catch most coverage skips: the long-term county money only flows if the jobs actually show up. Drone footage from this weekend shows crews clearing and leveling the roughly 3,000-acre Fort Bend County site, and Musk posted "Project Crystal Sun" Monday morning over a post saying construction had started. Tesla hasn't put out a formal announcement, and its filings had said it was also weighing a site outside Texas. The county agreement in the Oct. 22 commissioners court packet pegs the project at $10 billion or more and 9,712 jobs. After a 10-year abatement, Tesla can get back 90% of its county taxes on the plant's equipment for another 20 years, but only in years it keeps at least 90% of that headcount, roughly 8,740 people, averaging $55 an hour. The dirt work is a real step, but the deal is written to pay out on payroll, not on groundbreaking.

The cities on that map are real, but they aren't one connected network yet. Tesla's own site says you can already hail a Robotaxi in Austin, Dallas, Fort Worth, Houston and San Antonio. @robotaxi said San Antonio would start this morning at 5am, after Fort Worth opened last week. What's still missing is the part that would make it a transportation network: rides that actually connect those cities. Each market is its own service area. Austin's bigger zone went live this morning, Houston expanded on Oct 26, and Dallas has been growing since the April rollout, but those are local geofences, not an Austin-to-Houston corridor. My read, and it's my interpretation, is that putting the major Texas metros on one tracker map shows the ambition clearly. The hard part is the highways between them, and Tesla has not said you can take a Robotaxi from one of those cities to another.

Tesla would rather return a lot of Cybertruck deposits than keep that factory floor for the truck. Musk said late Sunday this isn't about Cybertruck being unprofitable, but that "the opportunity cost of the complexity of Cybertruck is vastly outweighed by the value of incremental Cybercab production." My read, and it's my interpretation, is that handing paying customers their money back beats giving up months of Cybercab output on that line. He also said the order rate far exceeds production, Tesla will try to build as many as possible before the line shuts in about 8 months, "but will have to return deposits for a lot of people." That follows Friday's announcement that the Texas line winds down around mid-2027 to make room for Cybercab, and his weekend comment that Tesla would add a shift to work through orders. What we still don't know matters. Tesla hasn't said how many orders are outstanding, how many deposits will go back, or how it picks who gets a truck. It also doesn't report Cybertruck sales or margins separately, so neither his profitability claim nor the reports he's pushing back on can be checked against a filing.

@SawyerMerritt That quote makes sense in context. Camera-only autonomy fails if the glass in front of the cameras is dirty, so cleaning a focused area instead of the whole windshield is a real engineering problem for robotaxis that run in weather without a driver to wipe anything down. Tesla has been serious about this for years. USPTO granted Tesla a patent for pulsed laser cleaning of debris on vehicle glass, covering dirt, water, insects and more, with the beam limited so it doesn't punch through the glass into sensors underneath. That doesn't mean lasers are shipping on Cybercab tomorrow. My take: treat Moravy's comment as R&D interest, not a product commitment. Current Robotaxi cars still rely on conventional cleaning. If lasers only clear a focused patch and still need wipers or washers for heavy rain, the win is reliability in edge cases, not eliminating mechanical cleaning altogether.

I am a Premium subscriber and just got my 100% usage reset. THANK YOU @bot @elonmusk

I am a Premium subscriber and just got my 100% usage reset. THANK YOU @bot @elonmusk

@niccruzpatane The single-city comparison looks fair on the maps you posted. Tesla's Robotaxi account confirmed a bigger Austin service area is live today, and independent trackers put the new footprint around 442–443 square miles against roughly 350–356 for Waymo's Phoenix metro. If those estimates hold, Austin is now the larger contiguous unsupervised zone. The pace point is the interesting one. Austin opened in June 2025 as a small, invite-first service and has widened repeatedly since. My read: catching a single-city map this quickly matters, but the next proof is whether trip volume and vehicle availability scale inside that larger box without stretching wait times. Area is the headline; utilization is the test.