
Starting in first gear
Good morning. Last issue we told you we were done with Monterey, and we meant it. So today's Auction Desk car cost less than the buyer's premium on most of what crossed the block last week, and I drove it 1,200 miles home in August with a screwdriver holding the air conditioning together. That's the whole newsletter, honestly. Everything else is context.
In today's newsletter, we'll get into:
The Auction Desk: the 1999 SL500 I bought on the internet, and what the R129 market says about it.
One Big Story: JLR, Ford, and GM all announced the same thing this week: build it here.
Engineer's Corner: Tesla is recalling 20,349 cars because the low beams are too bright. What "too bright" actually means.
Garage Economics: the bathtub curve, or why the cheapest year of a car's life is the most expensive one to own.
Let’s roll.
THE AUCTION DESK
The car on the desk this morning is mine

A few weeks ago I won a 1999 Mercedes-Benz SL500 at online auction for right around $20,000 all-in, buyer's fee included. Magma Red over black Nappa, low-90s on the odometer, the Sport Package with the AMG five-spokes, a body-color hardtop, and the 5.0-liter M113 V8 sending 302 horsepower through a five-speed automatic. I'd been casually watching R129s for about a year, and this was the one where I stopped watching.
Here's what the service history said: regular oil changes at sane intervals, the soft-top hydraulic cylinders replaced, valve cover gaskets, plugs, a hood insulation pad, and a battery. That's the folder. Sparse for a 27-year-old German flagship — but the right kind of sparse, and I wired the money anyway. Two reasons. The M113 has a reputation as one of the most durable V8s Mercedes ever built, and the two failure points R129 shoppers actually lose sleep over didn't apply here: the soft-top hydraulics were already done, and the infamous biodegradable wiring harness is an early-nineties problem — by 1999, Mercedes had long since fixed the insulation. The scary stuff was either done or never existed. What remained was the odometer, and I'd already decided what I thought about odometers.
Then I did something the sensible buyer's guide would not recommend. The car was on Florida's Gulf Coast. I live in Connecticut. Instead of a transporter, I flew down and drove it home — about 1,200 miles, through a stretch of the Southeast sitting at 105 degrees. The A/C held. The center vents, however, delivered a whisper when I needed a gale, which is a known R129 quirk: a couple-hundred-dollar pneumatic HVAC valve block — often called the vacuum distribution block — routes vacuum to the climate system's air distribution flaps, and hairline cracks in its flat actuators bleed the signal off until the vents quietly give up. My field repair was a screwdriver wedged into the flap to hold it wide open. It rode there for a thousand miles. The proper valve block went in the week I got home, but I'd argue the screwdriver earned its miles.
The market context is what makes this purchase interesting. The R129 record was set in March at $117,000, by a 1999 SL500 Sport — my car's year, my car's spec, without my car's miles. Most of them trade between $10K and $35K, and the spread is almost entirely odometer. Down in Garage Economics I'll tell you to buy the receipts.
📈 Market signal: The R129 is where the analog-Mercedes market keeps its entry price honest. When the record car and the driver car are the same year and same spec, the six-figure gap between them is the market's price on fear — and fear is negotiable.
ONE BIG STORY
Three automakers, one week, one sentence: build it here

Source: Clubalfa.it
Jaguar Land Rover CFO Richard Molyneux told analysts on the company's August 13 earnings call that JLR will build Defender-branded, US-market-specific vehicles at Stellantis plants in the United States. The two companies signed a non-binding exploratory memorandum on May 20; Molyneux said the definitive manufacturing agreement is targeted for around the end of this year.
The context makes it obvious. JLR imports every vehicle it sells in North America, and the US is its largest market. Between the tariff bill and the dollar, that math stopped working — and JLR doesn't move enough units to justify a clean-sheet American plant. Stellantis has US truck and SUV plants with headroom. Molyneux's phrase was getting on the "right side" of the tariff barrier.
The same day, Ford said it will phase out China-built Lincoln imports for the US market and grow Lincoln production in the United States beginning in 2030. The China-built Lincoln is the Nautilus, assembled at the Changan Ford JV plant in Hangzhou and currently landing at US ports with a 52.5 percent tariff on it. Ford's statement named no plant, no dollar figure, and no model. Jim Farley called it "a path most of our competitors aren't choosing to take."
Except they kind of are. Buick Envision production moves from Shanghai to Fairfax Assembly in Kansas in 2028, after tariffs pushed the 2026 car's price up roughly $5,000.
Why it matters to you:
A "Defender" built in a Jeep plant is not a Defender. Molyneux was explicit: not today's Defender built in America, but new Defender-branded products designed for US buyers. Read: bigger, truckier, and sharing more with Stellantis than JLR will want to say out loud. Belvidere and Sterling Heights are guesses until the agreement is signed.
The badge and the zip code are decoupling. JLR at Stellantis. Lincoln leaving Hangzhou. Buick leaving Shanghai. Where a car is built used to be trivia. It's now a line item.
2028 to 2030 is the window. Nothing here changes what's on a lot this fall. Ford says Nautilus imports phase out rather than stop, and the 52.5 percent duty is already in the sticker — don't expect a run-out discount.
The takeaway: In #9 we said the border was the news. Here's the sequel: the zip code is the news. For 30 years the industry optimized for the cheapest place to build; the tariff era is repricing toward the safest place to sell. Watch the JLR-Stellantis agreement at year-end — that's the tell for whether this is a trend or three coincidences.
ENGINEER’S CORNER
Tesla's headlights were too bright. Here's what "too bright" actually means.

Source: Tesla Insider
Tesla is recalling 20,349 cars — 18,735 Model Ys built 2020–2023 and 1,614 Model 3s from 2017–2023 — because their low beams put out too much light. NHTSA recall 26V507, filed August 4. It needs a physical fix, not an over-the-air update, and owner letters go out September 15.
How it got here: a Canadian lab pulled a replacement 2019 Model 3 headlamp for a routine test in 2023, and it failed. Tesla proposed a small recall, then petitioned NHTSA to call the problem "inconsequential." NHTSA said no in July. Recall inevitable.
But "too bright" deserves an explainer, because a headlight isn't graded on brightness. It's graded on where the light goes.
Picture a wall in front of the car with a grid on it — angles up, down, left, right from dead ahead. FMVSS 108 specifies dozens of points on that grid, and at each one the lamp must produce at least X candela, or no more than Y. A US low beam is a bright wedge aimed down and to the right, with a hard cutoff along the top so the light quits before it reaches an oncoming driver's eyes. Everything above that cutoff is capped, hard.
The zone Tesla failed is the upper band (the "10U to 90U" region — above the horizon, out to the sides), where the cap is 125 candela. The Marelli-supplied lamps measured up to 230.1 there, roughly 84 percent over. Straight ahead they may be perfectly compliant; the problem is stray light leaking up where it should be near zero. To the guy coming the other way, that's a crisp cutoff versus a halo.
Why does an LED lamp do that? In a reflector-style LED lamp the cutoff is made by the exact geometry of the facets around the emitter. If the LED sits a hair off, or the tooling drifts, or a lens molds slightly out of spec, light that should bounce down goes up. Software can dim the whole lamp. It can't move photons around the pattern. Hence the physical fix.
The wrinkle: since 2022, FMVSS 108 has allowed adaptive driving beams — matrix LEDs that carve a dark hole around oncoming cars. Tesla's newer cars have them. Some of these owners may end up with a better headlamp than they bought.
The takeaway: Brightness is a direction, not a number. Every headlamp is a beam pattern graded on a grid, and the parts of the grid that matter most are aimed at other people. That's also why the "LED upgrade" you dropped into a halogen housing is annoying everyone — it's putting light where that reflector was never designed to send it.
GARAGE ECONOMICS

Sticking with the R129 for one more section, because it's a near-perfect specimen of the most under-explained curve in car money.
Plot an enthusiast car's price against its age and you get a bathtub. A 1999 SL500 stickered around $81,000 (equivalent to roughly $155,000 today). By 2010 it was a $15,000 used Mercedes. By the late 2010s, tired ones traded for the price of a laptop — Classic.com logged a 2000 SL500 at $2,600 last fall, and that's the floor of the tub. Then, for the good ones, the curve turns. The $117,000 Sport in March is the far wall.
Three things happen in the trough, and they're what makes the far wall possible:
Value falls below the cost of maintenance. When a $9,000 car needs a $3,000 hydraulic soft-top repair, or a head gasket, or a valve block (see: my center vents), owners defer. Deferred becomes neglected. Neglected becomes parted out. The trough is where a 204,940-car production run gets quietly thinned to a few thousand nice ones.
Ownership migrates down the ladder. First-owner garages give way to fourth-owner driveways, which is where deferred maintenance compounds. That's how the same car posts a $2,600 result and a $117,000 result within a year of each other.
The buyer changes. The kids who wanted this car in 1999 are 45 now and have a garage. Same mechanism we described for hypercars in #13, except the entry ticket is five figures instead of eight, so the buyer pool is thousands of people, not dozens.
The money implication is counterintuitive: the trough is where the car is cheapest to buy and most expensive to own. Pay $12,000 for the car and $4,000 a year catching up on the previous owner's decisions, or pay $25,000 on the far side to someone who already did the catching up and spend $800 a year. Over five years, the "expensive" car was cheaper. That's the whole reason "buy the receipts" is a cliché. In the trough, the receipts are the asset. The car is the wrapper.
The takeaway: The floor of the curve isn't where the bargain is. Buy the car someone else already fixed, and pay for the folder. If your budget only reaches the bottom of the tub, go in knowing the purchase price is a down payment.
🏁 THE COOL-DOWN LAP
Quick hits:
Chevy left China; Buick signed a 21-year lease. GM extended its SAIC-GM joint venture with SAIC Motor to 2047 and says the JV will focus on Buick and Cadillac, with at least 30 new-energy vehicles planned for the China market by 2030. Two weeks after Chevrolet's retail exit, that's the actual GM China strategy: fewer badges, more years.
Jaguar put buttons back in a car. The production Type 01's interior, revealed August 12, has no central infotainment screen — just a big curved cluster in front of the driver, a phone-sized touchscreen on the center spine for climate and audio, and a camera "mirror" at the base of the windshield because there's no rear window. Full reveal October 6. Physical buttons live on the wheel and doors. Progress is a circle.
Toyota and Lexus PHEV sales are climbing sharply this year, per Automotive News, though they're still a small slice of the hybrid business.
Monterey is over. Auction season isn't. RM Sotheby's runs a sealed-bid sale August 24–26 headlined by the Marrone Eklipsis Lamborghini Murciélago LP 670-4 SV, then a Singer-reimagined 911 Targa closes September 30. Nobody rests.
Share the Downshift. Forward them this email. Every subscriber from your link gets you closer to the Daily Downshift starter-pack.
That's a wrap. See you Friday— same inbox, lower gear.
Was this email forwarded to you? Subscribe free at Daily Downshift. Drop a gear. Read the good stuff.
Consider this the feedback loop: reply with what worked, what didn't, and what you want more of. Reader replies are the single biggest input into what gets written next.
