A weekend of clear thinking about what New York could become over the next century: its purpose, its economy, and its streets.
The Retreat
A small gathering to think seriously and optimistically about New York's next hundred years.
What we'll discuss →A quiet lake house in Athens, NY: a calm place to think about a loud city.
See the venue →Short talks and long, open conversation. A discussion, not a conference of slide decks.
See topics →The retreat has concluded. Registration is closed, but the conversations and materials remain here.
Registration status →Stops on the Line
The Through Line
Before the particulars: what is the city's actual purpose for the next hundred years? A prosperity engine, a cultural icon, a living museum, a laboratory for how millions live well together? Every other stop on this line branches off the answer, including the harder question of who has to agree before any of it gets built.
The Economy Line
We'll discuss Jonathan's essay The Jagged Boom through the lens of New York: how uneven AI progress, astonishing successes beside embarrassing failures, reshapes the city's economy and everyday life, with real welfare gains that outrun measured GDP.
Read the essay ↗The Transit Line
Restore rail on the Rockaway Beach Branch, or pave it into a linear park? A real case study in the time horizons of public-transit paybacks, and how a city should weigh investments whose returns arrive decades after the ribbon-cutting.
QueensLink ↗The Autonomy Line
A once-in-a-century opening: as NYC transitions to self-driving cars, the vast acreage now devoted to parking gets freed up. What do we build with it: housing, parks, bike lanes, more public space? And how do we make sure we don't waste the moment?
The Resilience Line
Rising seas and Sandy-scale storm surge are the century's clearest physical threat. Do we build the big harbor barrier, a sea wall across the Narrows that only works as a joint NY–NJ project, or defend block by block with marshes, berms, and raised shorelines? What does a storm-proof New York actually cost, and who pays for it?
The Growth Line
Imagine 16 million New Yorkers instead of 8. What would it take to double the city without ruining it, and would that make it greater or just bigger? A forcing function for every hard question about housing, zoning, height, and who gets to call New York home.
Read a dystopian short story about what can go wrong ↗The Regional Line
New York doesn't end at the city line. Through-running regional rail that treats NJ Transit, the LIRR, and Metro-North as a single system, plus a real Northeast Corridor, Boston to Washington in a few hours, could knit the whole megaregion into one metropolis. What would that do to where people live, work, and belong?
Transportation Modeling
Based on the workbook in the data directory: compare each mode's generalized cost per passenger-km while checking whether it can carry the corridor demand.
Longer horizons reward high-upfront-cost modes with long operating lives.
Higher values give faster modes more credit through the time-cost term.
Fixed delays matter more on short trips and fade on longer trips.
Modes over this door-to-door time are marked insufficient.
Modes below this hourly capacity are marked insufficient.
Madrid and Seoul build near $100M/km; New York near $1,500M/km. This slider is governance, not physics.
Removes the parking search: car delay falls from 16 to 4 minutes and lane capacity rises with tighter headways.
Generalized cost per passenger-km vs. trip distance (log scale, current assumptions). Dashed line marks the current trip distance.
The bars show each mode's generalized cost per passenger-km relative to the most expensive mode under the current assumptions. Status requires enough capacity and travel time within the limit. The curves ignore the sufficiency filters — they show cost only.
Money cost spreads each mode's lifecycle cost (construction plus operations over the horizon) across the passengers it actually carries: min(corridor demand, capacity) × service hours × operating years × trip distance. An empty subway is not cheap per rider. Time cost adds in-vehicle time to a fixed delay for access, waiting, and parking. Waits for scheduled modes follow frequency: service sized to the corridor means big-batch vehicles come less often, so bus, BRT, and subway waits equal half the implied headway, bounded between 2 and 10 minutes. "BRT, exclusive ROW" models a busway on a fully grade-separated right-of-way — an abandoned rail corridor, for example — where buses run at near-subway speeds with no intersections.
generalized cost = money cost / pkm + value of time × ((distance ÷ speed) + delay hours) ÷ distanceAssumptions most worth attacking: the 10-minute parking search, a single value of time for all riders, demand treated as constant across service hours, and construction cost as a slider rather than a law of nature.
Known limitation: the model prices existing trips. It cannot see the trips a mode creates by making previously impossible trips possible — which may be the largest benefit of building anything.
QueensWay vs QueensLink
The Rockaway Beach Branch corridor, as a timing problem. Each option opens in a different year, ramps at a different speed, and serves a different mix of today's residents versus the people who move in later. Resident churn cuts both ways: it dilutes the train's benefits (they accrue to future residents) but concentrates the bike corridor's (each mover is drawn from people who chose the corridor for it).
30 years is an appraisal. 100 years is an identity.
Higher rates favor whatever opens soonest. Since the unit here is people served, discounting means valuing future people less — set it to zero if that bothers you.
Share of households that turn over each year.
Self-selection: people who move to a corridor with a great e-bike spine are not a random sample. Incumbent adoption starts at 4%.
Copenhagen's rebound was conversion, not churn: existing residents took up riding as they were exposed to riders. Adoption follows an S-curve toward the same ceiling movers reveal.
The bike corridor opens in year 2 and the busway in year 4 — NYC-realistic, not press-release. Slip scales with project size, so treat this slider's upper half as rail's delay scenario.
Off: benefits count for whoever lives on the corridor, ever — the city as a century-scale institution. On: benefits count only while they land on people who live there today — the median resident's view. This toggle is a philosophy, not a parameter.
Thousands of residents regularly served per year. Dotted verticals mark opening years.
Running average cost per person-year of service (log scale): all dollars spent so far — construction, then operations — divided by all person-years delivered so far.
The unit is people: how many residents each option regularly serves (at least weekly), counted every year and summed as person-years. It's an egalitarian metric by construction — a Sunday rider and a transformed commute count the same — and rail's induced development literally adds people to the count. The per-$M figure is the number that once made me ambivalent about QueensLink; the totals are the numbers that changed my mind back.
Each option has an opening year, a ramp, and a mature number of residents regularly served, out of roughly 300,000 living within a mile of the corridor. The busway ramps to 50,000 riders over five years (the parallel Q52/53 buses already carry on the order of 15–18,000 people a day on a hostile street). Rail ramps to 65,000 riders over three years — QueensLink's 75,000 daily-ride projection, in humans — then adds 25,000 new-resident beneficiaries over fifteen years, because induced development is slow. The e-bike corridor serves adoption share × corridor population. Adoption grows through two channels: churn replaces average residents with self-selected movers, and incumbents convert through social contagion — at a rate proportional to their exposure to existing riders, following an S-curve toward the ceiling the movers reveal. Copenhagen's rebound from ~10% to majority cycling was mostly the second channel: the same residents, converted in place over two decades, because protected infrastructure made the bike the fastest option and everyone knew someone who rode.
Opening years are NYC-realistic rather than best-case: two years for a paved corridor (design, review, and procurement take longer than paving), four for the busway (the corridor's ~20 derelict bridges are the long pole, not the pavement), twelve as the rail median. Megaproject base rates say roughly nine in ten large projects run late, and absolute slip scales with duration — so the small projects carry months of schedule risk while rail carries years; the rail slider's upper range is the delay scenario. The combined option counts the busway's riders plus 85% of the bike corridor's, assuming modest overlap between the two.
The cost chart spends each option's capital evenly across its construction years, then adds annual operating costs once open ($3M for the bike corridor, $25M for the busway, $28M combined, $70M for rail — running buses and trains for decades costs multiples of building the thing). The line is a running average: every dollar spent so far divided by every person-year delivered so far, which is why rail plunges from off the chart as its ridership accumulates against a fixed capital bill.
The current-residents toggle multiplies each year's benefit by (1 − churn)^t — the fraction of today's residents still present — and restricts bike adoption to the contagion channel, since self-selected movers are by definition not today's residents. Today's residents still convert; they just can't be replaced.
Assumptions most worth attacking: the mover-adoption ceiling (is 20% self-selection plausible, or is housing choice too multi-causal?), the rail extensive margin at 40 (roughly twice the busway's land-value capitalization, per the empirical literature), the equity blind spot — the bike corridor's benefits go to people healthy and willing enough to ride, and mostly skip the Rockaway peninsula, whose residents were the point — and the flatness of person-years itself, which counts a pleasant Sunday ride and a life-changing commute as one person each. If you think depth of benefit should matter more than breadth, this is the metric to argue with.
Weekend Timetable
Sessions are short and conversational, and everything bends around the lake, the weather, and the group. Nothing here is mandatory.
In the Same Spirit