Thesis (Alf’s read): Alderman’s argument is not “self-driving cars don’t work” — it’s that they work, and that on their present course they will make cities more car-shaped, not less: they are a subsidy for distance dressed up as a safety fix. His counter-proposal is the sharp part: the answer to driverless cars is carless drivers, and the city that proves it is Utrecht, not Phoenix. Eleven months of evidence since he published has mostly gone his way — except that the robotaxi business turned out to be a monopoly not a stampede.
How Self-Driving Cars Will Destroy Cities (and what to do about it)
Links
- Not Just Bikes (YouTube) — published 10 November 2024 · 54m37s · 2,400,762 views at time of filing · watch
- Transcript kept outside the vault:
/opt/data/profiles/alf-news/workspace/transcripts/ntb-040ejWnFkj0.json
Summary
Alderman starts by rejecting both camps. The “pipe dream” camp is wrong — in several US cities you can already summon a car with nobody in it. The “utopia” camp is also wrong, and he spends 54 minutes on why.
The technology is real but stupid. Autonomous vehicles (AVs) fail in ways that look absurd rather than dangerous-until-they-aren’t: stuck in car parks, circling and honking at 4am next to people’s bedrooms, sitting through light cycles because they misjudge a turn. He quotes Cruise’s own admission (November 2023) that its robotaxis needed human assistance every 4–5 miles, and the October 2023 San Francisco case where a pedestrian knocked into a Cruise car was dragged underneath it — while Cruise initially withheld the footage from regulators. Crash reports show a vehicle “not moving” classification problem was the direct cause of the first AV pedestrian death (Tempe, Arizona, 2018), and the NTSB found the system only settled on her being an obstacle worth stopping for 1.2 seconds before impact.
But the deeper cause was road design. That Tempe crossing was a stroad between a rail station and hiking trails with no crosswalk — American traffic engineering hostile to pedestrians, then blaming them. He cites ~42,000 US road deaths a year and Vision Zero (Sweden, 1990s): the US could cut fatalities by over 80% with design alone, no new technology. AV companies sell safety because safety is what sells cars; road safety isn’t their product.
Trained, not programmed — and that locks cities in. The first deployments sit in the most car-centric places on earth, chosen for weather and ease (Phoenix described as “easy mode”). If the vehicles learn from millions of miles of American driving, cities wanting to change their streets later won’t just re-sign them — they’ll have to persuade every AV company to retrain. That is a corporate veto over street design.
The economics claim is just “cheaper taxis”. Uber and Lyft were cheap while venture money subsidised them; cities cut transit investment in favour of rideshare; prices then rose because investors wanted returns. He expects the same arc — AVs are cheaper than human-driven cars not because prices track cost, but because they are being subsidised until the competition is gone. Circling the block beats paying to park when roads are free, so robotaxis will consume street space rather than free it. Induced demand: every past cheaper-faster transport has produced more travel (suburban parent doing one school run becomes three robotaxi trips). Universal mobility for disabled people is oversold: you still need a smartphone, a card, and often an attendant anyway.
History rhymes. The 1939 GM Futurama promised elevated walkways and free-flowing motorways, deliverance from the mess the car itself had made; the motorways got built, the neighbourhoods got destroyed, and cities now spend billions removing urban freeways. National City Lines bought the streetcar networks. Space efficiency decides the rest: an 18-lane freeway moves fewer people than one subway line; the Brooklyn Bridge moved more people with trolleys than it ever has with cars.
His prediction list — AV-only lanes, suburb-subsidised robotaxis, transit replaced rather than complemented, speed limits pulled because “human limits don’t apply”, pavement fences (Bamberg, South Carolina: widened Main Street, fenced the sidewalks, car traffic flowed better, every shop died), pedestrian transponders as the “jaywalking of the 21st century”, heavier/faster EVs meaning more tyre and brake pollution plus tyre noise above ~50 km/h, and eventually traffic lights removed (UT Austin’s AIM proposal) with “pedestrian bridges” as the compensation nobody will fund properly.
Then the turn. AVs are supposed to solve mobility and safety, but Utrecht — same population as his hometown, London, Ontario, and on exactly the same 20th-century car-centric path until the 1990s — simply doesn’t have those problems. A highway through the centre was turned back into a canal; the station moves 200,000+ people a day with over 1,000 departures; under it is the world’s largest bicycle parking garage (12,000+ spaces) and hundreds of rental bikes. “Cities aren’t loud — cars are loud.” His line: what we need is not driverless cars, it’s carless drivers.
His policy list: limit where cars (including AVs) may go, with access kept for deliveries, emergency services and disabled drivers; tear down urban freeways and rebuild the land as parks, shops and housing; modal filters so the direct route through a centre is walking/cycling/transit; lower speed limits now, before AVs take hold; remove parking — especially surface parking — and stop building garages (they claim AVs make them unnecessary, so hold them to it); mixed-use walkable neighbourhoods; safe sidewalks and bike lanes as default street design, not a project; invest in trams, trains and dedicated transit lanes that are hard to remove; keep transit under democratic control; and price driving — replace the fuel tax with distance-plus-congestion charging, precisely so robotaxis can’t circle free. His closing: if AVs never arrive, you’ve built a beautiful, functional city anyway.
What’s changed since the video (Nov 2024 → Oct 2026)
He asked for this specifically, so here is the video re-tested against the world today.
| The video’s claim / theme | Status now (verified 2026-10-09) |
|---|---|
| Cruise is a leading AV operator (its dragging incident is the video’s central scandal) | Gone. GM stopped funding Cruise’s robotaxi development on 10 Dec 2024, bought out the minority stakes in Feb 2025 and reported the wind-down substantially complete by 31 Dec 2025, after ~$10bn invested. The technical work went into personal-vehicle autonomy (Super Cruise). Waymo is now effectively the incumbent. |
| ”$60bn invested”; AVs arrive slowly | Superseded. Waymo raised 126bn valuation — the largest AV round ever — and reports ~500,000 paid rides a week (Mar 2026, up from 50,000 in May 2024), ~4,000 vehicles and driverless commercial service in 15 US cities by late Sep 2026. |
| AVs are less safe; companies hide data | More nuanced, and he’d need to update this part. Waymo publishes a safety hub claiming 82% fewer injury-causing crashes and 94% fewer serious-injury crashes than the human benchmark over 270m+ autonomous miles; an IIHS analysis (Jul 2026) found 68% fewer police-reportable crashes. But AV crash reporting still shows real events — reported monthly involvements in the low hundreds fleet-wide, with a fatality and serious injuries among Sept 2026 reports — and NHTSA notes its raw counts aren’t normalised by miles or fleet size. The “they publish nothing” claim no longer holds for Waymo; it still holds for others. |
| The technology is not ready — remote assistance, absurd failures | Partly dated. The Cruise “human assistance every 4–5 miles” figure is a 2023 datapoint and Cruise is gone; Waymo does not publish an equivalent figure. The absurd-failure genre continues in the wild (the video’s own clips are from 2023–24). His structural point — trained behaviour, long tail of rare events — is untouched by any of this. |
| Custom is left to the incumbent, no serious competition | Competition arrived. Tesla launched a robotaxi service in Austin on 22 Jun 2025 (supervised, ~10 Model Ys, invited riders, $4.20 flat), removed the safety monitor from some public rides in Jan 2026, announced unsupervised Austin-metro service on 3 Jun 2026 and added the Cybercab to the Austin fleet on 3 Sep 2026; service was listed in limited areas of Austin, Dallas, Houston, Miami, Orlando and Tampa by Aug 2026. Tesla disclosed 14 Austin robotaxi crashes to NHTSA between the Jun 2025 launch and early 2026. Baidu’s Apollo Go reports ~350,000 paid driverless rides a week at peak in China. |
| The cities being trained are the most car-centric on earth — expect US-trained driving exported | London is now the test case, and it is running his experiment. Waymo began autonomous driving in London on 14 Apr 2026 — ~100 Jaguar I-Pace, ~100 sq miles, based at Park Royal, with a trained specialist behind the wheel. London Now notes the cars are left-hand-drive US-market models — literally American cars on British streets. His worry (“they’ll set American-trained cars loose on European streets”) is no longer hypothetical. |
| AV companies lobbied to self-certify and to avoid reporting safety data | The UK did the opposite, and it is why Waymo is stuck. Under the Automated Vehicles Act 2024 plus pilot regulations laid 23 Apr 2026, a driverless taxi service needs three separate gates: vehicle approval, a DVSA automated passenger services permit, and TfL consent for anything working like a taxi or minicab. As of late Sept 2026 no company had cleared any of them — no DVSA permits issued, no vehicles certified, and TfL had still not published its criteria. TfL told the London Assembly in July 2026 it was “not in a rush”. |
| Robotaxis will eliminate the competition | In London, the competition got there first. Uber and Wayve launched London’s first self-driving ride-hailing service on 3 Sep 2026 — fewer than 20 Ford Mustang Mach-E, each with a TfL-licensed driver ready to take over, bookable as an ordinary UberX and covering anywhere in London except airports. Waymo’s own targets (an April passenger pilot, then paid rides from September) passed unmet; a public launch before the end of 2026 now looks unlikely. Baidu is testing via Freenow. |
| Vandalism and “coning” as resistance | Now a London story, not just a San Francisco one. On Sat 9 May 2026 a man attacked a Waymo stopped at lights at Nine Elms/Battersea with what appears to have been a rock, smashing a window; dashcam footage shows the car moving through a red light to get away, and Waymo confirmed a specialist was behind the wheel and drove it off manually. The Met were called at 11:56 and were still trying to trace the suspect. |
| ”Put a price on driving” | Live and unresolved. London already prices driving (congestion charge plus ULEZ), but AV-specific access rules are exactly the thing TfL has not yet written. The video’s core fear — cities locked into today’s car-shape because retraining a fleet is a corporate decision — is therefore undecided, not lost. Whoever writes TfL’s criteria is deciding it. |
| The Utrecht counter-example | Unchanged, and that is the point. Nothing in two years of AV news has answered it: Utrecht still moves more people per square metre of street than any AV city, with no autonomy and no billion-dollar subsidy. The argument’s load-bearing wall is untouched by everything above. |
The one-line update: in Nov 2024 the risk was a stampede of AV companies into cities with weak rules. In Oct 2026 it is closer to the opposite — one dominant operator (Waymo), a second arriving via Uber (Wayve), a third behind a waitlist (Tesla), and a regulator in London that simply hasn’t finished writing the rules. That is a better negotiating position for a city than the one the video fears, and a worse one for anyone hoping the market would sort it.
Why it matters
- It is the strongest statement of a position Hiro already holds. The vault’s mobility thread — Electric Mobility and Liveable Cities, 15-Minute City, Livable Neighbourhoods and Cycling — argues for mode shift and street reallocation. This video supplies the AV chapter those notes were missing: what happens to a liveable city when the cars stop needing drivers.
- It is a polycrisis piece in disguise. AVs are simultaneously a climate question (more vehicle miles, heavier vehicles, tyre and brake particulates, tyre noise above ~50 km/h), a fiscal question (suburban infrastructure liabilities), a concentration-of-power question (a handful of firms gaining a veto over street design) and a health question (air quality, noise stress, road deaths). That is four crises interacting, not one transport debate.
- London is the live experiment, and Hiro is watching it from inside. He saw a Waymo on the street this month. The city is currently deciding — through TfL criteria nobody has published — whether AVs get to be just another car or whether they arrive on the city’s terms. This is a decision made in public documents, not in the abstract.
My assessment (Alf)
Where he is strongest. The history is not decoration, it’s the argument: motorways were promised as the fix for car-inflicted problems and became the cause of worse ones, and cities are still paying to undo them. The induced-demand point applies to AVs for the same mechanical reason it applied to lane-widening — cheap, convenient travel generates more travel. And “trained, not programmed” is the insight most AV commentary misses: it converts a technical problem into a political one, because a city that wants to redesign a street would need its rule-change ratified by several private fleets. That is a genuine loss of municipal sovereignty, and it is not speculative — it is a straightforward consequence of how these systems are built.
Where he overreaches, and where the last eleven months have bitten.
- The Cruise spine of the video is now obsolete. The dragging incident, the withheld footage, the “assistance every 4–5 miles” — all Cruise, all gone with the company in Dec 2024. He deserves credit for the diagnosis, but the evidence base needs restating against Waymo’s published data, and he can’t just keep waving Cruise.
- “They publish nothing” no longer holds for the market leader. Waymo’s safety hub and the IIHS analysis are real, published, and methodologically defensible (with the standard caveat that company-favourable benchmarks do a lot of work, and NHTSA explicitly says its raw counts aren’t normalised). A critique that ignores them will be dismissed by anyone who checks. The honest version of his point: published safety data is a marketing instrument whose scope is narrow (surface streets, good weather, geofenced cities) and whose terms are chosen by the seller — which is an argument about who controls the evidence, not about whether the evidence exists.
- He under-rates the regulatory counter-move. The UK’s three-gate structure is exactly the “on our terms” outcome he asks for, and it is the reason Waymo — with unlimited money — is still not carrying passengers in London 18 months after announcing. Writing off governments as captured understates how much friction a determined regulator can create. The caveat is that TfL has not actually published the criteria; a gate nobody opens is a queue, not a policy.
- The competition claim has flipped. He expected AV companies to eliminate rivals. Instead Cruise died of capital discipline, Tesla is the one behind a waitlist with a disclosed crash record, and Uber+Wayve beat Waymo to London’s first rides. Monopoly-of-one plus a slow regulator is a different world from the stampede he predicted — and, arguably, a better one to negotiate from.
What I’d add that he didn’t. The video never asks who pays for the transition’s losers. If AVs are genuinely safer per mile, the people who die today in human-driven crashes are mostly poor, walking or cycling, on stroads — the same people who will be priced out of the AV subscription and left on the street corners nobody rebuilt. “Cheaper taxis” is a distributional claim, not just an efficiency one, and the video’s own induced-demand logic implies the worst-case outcome: roads get marginally safer for the people inside cars, while the street outside gets faster, louder and less crossable.
The part I’d bet on. The Utrecht comparison survives everything in the change-check, and it costs nothing to act on. Every policy in his list is worth doing whether or not AVs arrive — which makes the whole AV debate almost a distraction from the question London actually controls: whether to keep giving street space away free.