I'm reminded of the famous story of (I think) the central beam in a building at Oxford. The story goes something like:

The central beam was beginning to fail and the Oxford administration knew they needed to replace it. When they went around for quotes, no one could replace the beam because it was 100 ft in length and sourced from an old growth tree. Such logs were simply unavailable to buy. To solve the issue, the staff begin to look at major renovations to the building's architecture.

Until the Oxford groundskeeper heard about the problem. "We have a replacement beam," he said.

The groundskeeper took the curious admins to the edge of the grounds. There stood two old growth trees, over 150 feet tall.

"But these must be over 200 years old! When were they planted?" the admins asked.

"The day they replaced the previous beam."

Here's another good example of a series of slow experiments: the cosmic distance ladder.

https://www.youtube.com/watch?v=YdOXS_9_P4U https://en.wikipedia.org/wiki/Cosmic_distance_ladder

You can compute the distance to the moon if you know the radius of the earth by looking at how long lunar eclipses take, data gathered over years of observations.

Eratosthenes computed the radius of the earth by clever trigonometry in ancient times, and Aristarchus computed that a 3.5-hour lunar eclipse indicates that the moon is ~61 earth radii away.

Once you have the distance to the moon, you can compute the size of the moon by measuring how long it takes the moon to rise. It takes about two minutes, and so the radius of the moon is about 0.0002 of the distance to the moon.

By cosmic coincidence, the sun and the moon appear to be approximately the same size in the sky, so the ratio of radius/distance is approximately the same for the sun and the moon. If you measure phases of the moon, you'll find that half moon is not exactly half the time between the full moon and new moon. Half moon occurs not when the moon and the sun make a right angle with the earth, but when the earth and the sun make a right angle with the moon.

You can use trigonometry to measure the difference between the half-time point between new/full moon, and the actual half moon, giving you an angle θ. The distance to the sun is equal to the distance to the moon divided by sin(θ).

To get θ exactly right, you need a very precise clock, which the Greeks didn't have. It turns out to be about half an hour. Aristarchus guessed 6 hours, which was off by an order of magnitude, but showed an important point: that the sun was much larger than the earth, which was the first indication that the earth revolved around the sun. (Aristarchus' peers mostly didn't believe him, not simply out of prejudice, but because the constellations don't seem to distort over the course of a year; they were, as we now know, greatly underestimating the distance to nearby stars.)

Next, you can compute the shape of the orbits of the planets, by observing which constellations the planets fall inside on which dates over the course of centuries. Kepler used this data first to show that the planetary orbits were elliptical, and to show the relative size of each orbit, but with only approximate measures of the distance to the sun (like the θ measurement above) there's not enough precision to compute exact distances between planets.

So, scientists observed the duration of the transit of Venus across the sun from near the north pole and the south pole, relied on their knowledge of the diameter of the earth, and used parallax to compute the distance to Venus, and thereby got an extremely precise measurement of the earth's distance to the sun, the "astronomical unit." It took decades to find the right dates to perform this measurement.

The cosmic distance ladder goes on, measuring the speed of light (without radar) based on our distance to the sun and the orbit of Jupiter's moon Io, using radar to measure astronomical distances based on the speed of light, measuring brightness and color of nearby stars to get their distance, measuring the expected brightness of variable stars in nearby galaxies to get their distance, which provided the data to discover redshift (Hubble's law), measuring the distance to far away galaxies (and thereby showing that the universe is expanding).

In my free time, I have taken to trying to prove the Collatz conjecture.

People much smarter and more educated than me have failed at this quest, so I will nearly certainly fail at it, but that's not really the point in my mind. Even if I'm not the one to actually prove it, I can at least try and contribute to the body of work towards proving it. Mathematics is, more than nearly anything else, the result of generations building upon previous generations work. It's never "done", always growing and refining and figuring out new things to look at.

I have a few ideas on how to prove Collatz that I have not seen done anywhere [1], and usually (at least for me) that means it's a bad idea, but it's worth a try.

One of the greatest things about humans is our willingness to have multi-generational projects. I think maybe the coolest thing humans have ever done was eliminate smallpox, and that took hundreds of years.

[1] Which I'm going to keep to myself for now because they're not very fleshed out.

    The 2nd Ave Subway in Manhattan, with
    preparatory construction beginning in
    1942. First phase opened in 2017.

    While every project is unique, it is not
    immediately clear why digging a subway
    on the Upper East Side is twenty times
    more expensive than in Seoul or ten
    times more expensive than in Paris.

here's a even more damning look: https://www.vitalcitynyc.org/articles/why-it-costs-4-billion...

edit: I've been on a tirade about this subject this week. https://www.brethorsting.com/blog/2025/07/state-capacity-and...

A friend of mine once wrote a dictionary[1]. It has all the (normal) one syllable words in English, defined using only other one syllable words. He decided to work on it by focusing on one letter per year, so A was in 1991, B was 1992, and the book was finished in 2017, 26 years later.

It's not even a very long book - only a few hundred pages - but I'm sure if I tried to do the same thing all at once, I'd probably have lost interest around B or C, so I suppose it was a worthwhile strategy.

[1] It's not online anywhere as far as I know, sorry.

The SAS is a joke, putting its name on the same list as actually impressive feats like the proof of Fermat's Last Theorem insults everything else on the list. It's the most expensive subway line worldwide per mile, ever, despite the existence of technology that made tunnelling easier. Inflation adjusted, it costs more per mile than hand-digging one of the PATH tubes with 1900s technology [1]. Its cost and duration are almost entirely due to politics and not technical and logistical challenges, including the MTA political fiefdom fighting the Park Board political fiefdom, make-work-program labour spending, staff paid to have their thumbs up their asses in the tunnels [2], deep-bore tunneling instead of cut-and-cover to avoid political fighting, and MTA departments wanting their miniature fiefdom dug into the ground at each station [3]. The SAS is a project that should bring great shame to everyone in charge and everyone who stood around in the tunnels getting paid to do nothing.

[1] https://en.wikipedia.org/wiki/Uptown_Hudson_Tubes (tunnel happens to be about a mile and it cost 21 million 1905 dollars)

[2] https://www.nytimes.com/2017/12/28/nyregion/new-york-subway-...

[3] https://pedestrianobservations.com/2023/12/09/the-mta-sticks...

I dunno. I think we should separate out the stuff that fundamentally has to take a long time, like the pitch experiment, from stuff like Notre Dame, which just took a long time because they lacked the resources to do it all at once. Like OK, it takes a long time to build a big church because you need to find all the right rocks or whatever. But the pitch, that’s the universe taking a long time to tell us something.

(I’m being flip for comedy/emphasis sake, of course Notre Dame is pretty impressive too).

This reminded me of an old comic or meme about people’s expectations about science that went like:

Protester: What do we want??

Crowd: High quality, double blinded, N of 100000, 20 year longitudinal, preregistered studies!!

Protester: When do we want it??

Crowd: Now!!!

Related by content (OP says "This page is a riff on Patrick Collison's list of /fast projects"):

Fast - https://news.ycombinator.com/item?id=36605912 - July 2023 (298 comments)

Fast (2019) - https://news.ycombinator.com/item?id=30872279 - March 2022 (97 comments)

Fast - https://news.ycombinator.com/item?id=21848860 - Dec 2019 (291 comments)

Fast · Patrick Collison - https://news.ycombinator.com/item?id=21355237 - Oct 2019 (5 comments)

--

Also related, if only by title, this from yesterday:

Fast - https://news.ycombinator.com/item?id=44736967 - July 2025 (418 comments)

I think of The Art of Computer Programming...

https://en.wikipedia.org/wiki/The_Art_of_Computer_Programmin...

Kind of amusing to have this at the top of the front page considering “Fast” was there yesterday

Kudos to the OP for writing this.

That PC post always irked me. Not because it showed positive examples of going fast but because it felt slightly demeaning to teams/projects that move slowly on purpose, with intent.

The Art of Computer Programming has been a work in progress since 1962.

That’s longer than some of the list items.

Two more to add to the list:

– The Voyager probes were launched in 1977 and reached interstellar space in the 2010s. https://en.wikipedia.org/wiki/Voyager_program

– The oldest bonsai have been in training for centuries. https://en.wikipedia.org/wiki/Bonsai

Most democracies have elections every 4 or 5 years. That is good, in that we can get rid of underperforming politicians and parties. But it is bad, in that there isn't a lot of incentive for politicians and parties to plan over a longer timescale than 4 or 5 years.

China has the opposite problem. It can plan and finance long term projects. But there is little prospect of peacefully changing the leadership.

>A fun question: of these projects, which required a long time, and which could have been greatly accelerated?

Pretty much everything on the list is a research study of a long-term process that is inherently impossible to accelerate.

From the list, only the Second Avenue Subway and the Sagrada Familia unambiguously qualify as projects that could be greatly accelerated. The SAS was not under active construction for the vast majority of the time between 1942 and 2017 — actual construction only happened for a couple years in the early 70s, then another couple years in the late 80s, and finally from 2011-2017. The fits and starts were due to a combination of bureaucratic red tape, economic woes, and gross incompetence. The Sagrada Familia has also seen only intermittent construction over the last century, primarily because of lack of funding.

I personally find these examples underwhelming. Most of them are processes that require time, like the pitch drop experiment.

I suspect that the things in our lives that truly have value and take a long time aren’t easy to identify as projects. No one person starts it with a clear idea of where it will end. Investment in future capabilities. Knowledge gathering without clear application or business model. Strengthening institutions and traditions of human rights to ensure that no one group can arrest history.

This goes hand in hand with the Lindy effect[0]. Some of the examples given in the article are a testament to it.

[0]: https://en.wikipedia.org/wiki/Lindy_effect

I'm reminded of the quote: “Never give up on a dream just because of the time it will take to accomplish it. The time will pass anyway.”

There is a restaurant (or, perhaps, food purveyor is more accurate) in Japan that has lasted for several centuries. It's been owned by one family, I believe. When Covid hit and their clientele disappeared, they just continued to pay their staff and mostly closed operations until the pandemic was brought more or less under control. They had enough money socked away that they survived this period unscathed. I wish I could remember the name of the place.

Anyway, that's my dream - to own and run a small family business that can support the family even in times of extended crisis. I have no interest in unicorns or IPOs or buyouts or any of that.

Neural network was introduced in 1950s. However, the neural architecture, the compute and data required for them to be efficient has been only in last decade.

From perceptron to transformers (few hidden layers to 480B parameters), from multicore CPUs to distributed GPUs and WWW/social media has all contributed to the growth of Artificial Intelligence.

This has took almost 50+ years and so many iterations along the way.

There’s also John Cage’s “As Slow as Possible” https://en.m.wikipedia.org/wiki/As_Slow_as_Possible

The point of https://patrickcollison.com/fast is not that everything has to be fast, but that you can probably do it faster than you think. Quoting https://nat.org/: "time is the denominator"

I like this list (and Collison's)!

One thing I would say -- the Sagrada Familia definitely didn't have to take the incredible time it has. Maybe not a good example of something that could only be done over the long term. Gaudi didn't prioritize it, and a civil war ruined it.

It is, however, an example of something beautiful that did take a long period of time.

You can watch the Pitch drop experiment [1] live here http://thetenthwatch.com/feed/

[1] https://en.wikipedia.org/wiki/Pitch_drop_experiment

I had to enable third-party JavaScript and resource loading to see a rendering of… “1.0 × 10⁻²¹”. Sometimes people use TeX math markup superfluously.

Language itself is an interesting problem. We have texts that are ancient and some are unreadable and others are readable. I personally can't understand old variants of English while (American) English is the only language I speak.

There is so much assumed in our use of language that it can be largely unintelligible without detailed historical context. The first time I heard the term "in the car park" I chuckled at the thought of an amusement park for cars... "parking lot" only came a few thoughts later. We drive on parkways. We play in the park. We park in the lot. Lots are reading this sentence twice. Give this paragraph to a school-kid in just 100 years and it will seem like gibberish. Word.

As an academic, I fairly resonate with this message. Also notice that most examples he noticed are from academia/science endeavors. I see academia as probably the only place where slow projects are expected and even encouraged; think of PhD students working on basic science problems, often supported for 5-7 years at end (of course close to minimum wage).

This is not to hide that all slow undertakings are good or anything. Often because of inefficient executions or bureaucratic hurdles, academic suffers. But, I am trying to highlight the observation that how a slow and steady progress is the typical modus operandi for an academic lab/group. A famous saying comes to mind: Rome isn't built in a day.

Regarding LIGO, if anyone finds the sensitivity of LIGO as shocking as I do, here's a 2002 lecture from Kip Thorne explaining how it's achieved.

https://www.youtube.com/watch?v=mGdbI24FvXQ&t=495s

This video is one of about 60 recorded in a year long series of lectures that were delivered at Caltech early on in the project. They are archived by Pau Amaro Seoane at this address https://astro-gr.org/online-course-gravitational-waves/

I think that pretty much everything we work on (as tecchies) is the endpoint of a very long timeline.

Every advancement stands on the shoulders of those that came before. Maybe we can run an LLM, because some Roman architect figured out how to make an aqueduct stay up in a seismically-active area.

If you watch James Burke's Connections[0], you get a feel for it (I think some of them are a bit of a stretch, but I really enjoyed it).

[0] https://en.wikipedia.org/wiki/Connections_(British_TV_series...

> Many Cathedrals were built over more than a century. An example is Notre Dame, over 1163-1345.

Or Cologne Cathedral, which took more than 600 years to complete. Though actual build times were a bit shorter (1248–1560, 1842–1880).

Long-term projects make me strangely proud to be a human (for all of our faults and foibles). "A society grows great when the old plant trees in whose shade they will never sit."

For open source, SQLite has pledged long term support through 2050: https://www.sqlite.org/lts.html

I imagine it will go on for much longer, though!

I will post this in defence of speed.

https://jsomers.net/blog/speed-matters

I think in the western world, Art, and music are both long term projects. So much so that we seem to have "reinvented" music at least twice. Once after the Greeks into classical western music, then again when jazz went into tonal harmony.

At least parts of it are "scientific" and "directed," see the Lydian Chromatic concept for example https://en.wikipedia.org/wiki/Lydian_Chromatic_Concept_of_To...

In case the author is here…

Here’s my pedant nitpicking: La Sagrada Familia is not a cathedral. It’s just a regular church, albeit a large and impressive one.

Some are projects that have a changing variable over a long period of time (Framingham Heart Study, E. coli long-term evolution experiment) or strive to exist a long time (Clock of the Long Now). I would argue that these projects -- their process, data collection methods, and goals -- may have been developed quickly, in a short amount of time. Their longevity is proof that the original project was well established. But the same could be said of the invention of the wheel, shoe, sliced bread, etc

I assume this is a response to: https://news.ycombinator.com/item?id=44736967

If we look beyond problems that humans solve, well, evolution of diverse and specialized species seems to require time (and be undone by humans going fast)

A few other proposed entries:

- https://en.wikipedia.org/wiki/Svalbard_Global_Seed_Vault

- https://en.wikipedia.org/wiki/Voyager_program (the twin spacecraft that have since left the heliopause)

Quadratic equations took something like from the ancient Greeks to the middle ages, afaik.

Made me think of https://99percentinvisible.org/episode/as-slow-as-possible/

The Crazy Horse Memorial has been going since the 1940s. It's progressing nicely.

https://en.wikipedia.org/wiki/Crazy_Horse_Memorial

Interestingly, the post titled "Fast" made the front page yesterday: https://news.ycombinator.com/item?id=44736967

The Sagrada Familia has been criticized as a symbol of bureaucratic inertia, some critics insinuating that the delays are deliberate for financial interests.

Making the argument for "Medium" - https://news.ycombinator.com/item?id=44750838

I'd add https://en.wikipedia.org/wiki/Centennial_Light to this list.

I love the story of the Framingham Heart Study, it's one I've referenced a lot when I talk to people and organizations about how they might not have the data they need and how important data collection is.

Nice post! This rhymes with the ideas Cal Newport presents in Slow Productivity.

> I suspect many key open source systems (Linux, Wikipedia) will still be around in 100 years.

Bet FORTRAN will still be around. Maybe PHP, as well. Def C.

. My work aims to help create systems which support creativity and discovery. Currently, my main projects are working on metascience, programmable matter, and tools for thought. In the past I've worked on quantum computing, open science, and artificial intelligence, and there's a lot of crossover with my current interests. Bio (2020).

The 2nd av subway is a bad example… that’s just a masterclass in political incompetence.

Also, all my side-projects.

(No, I will NOT use LLM for side-projects. That defeats the purpose of side-projects for me!)

I guess tomorow's front page top article will be called "Steady" ?

Missing the California high-speed rail on their list of examples.

The bitcoin block chain should be on this list

Cool list, but to be a party pooper:

> Will Unix Time or TCP/IP ever be replaced? Modified: sure.

UNIX time is already being replaced with a 64 bit value instead of signed 32 bit. TCP/IP has already been replaced, that's QUIC over IPv6 which is what my computer uses every time it connects to Google.

I mean you can claim IPv6 is still "IP" because it shares the same first two letters, but IPv6 is different enough to be easily considered a different protocol.

Well, in 1980 when I got my hands on the first "powerful" benchtop computer that I had complete control of, I started a project to do a little machine learning so that I would one day have some of the foundation I needed to handle the data more intelligently in the future.

That's what I always wanted to have a computer for and why I took Fortran in college to begin with.

I knew I wasn't going to reach the "intelligence" part within a short number of years, for one thing I had figured out it would be much faster on a more specialized chip than a CPU, plus it would require megabytes of memory when I only had kilobytes, and way more storage space too.

I couldn't be spending years concentrating on this while waiting for hardware technology to progress, and for survival otherwise I gravitated to a niche within a natural science career that would not be replaced by the AI which I expected to be rapidly approaching from those who had a much better head start both financially as well as in computer science.

Now by the time the late 90's rolled around, I already had my own company for a number of years (knew that was going to take decades too so I had started in that direction as a teenager) and by then had more than one computer. Woo hoo!

And megabytes! Oh Yeah!

Plus Office '97 which put the "paperless office" within reach even though paperwork was my primary deliverable product.

With Y2K looming I decided to use some of the megabytes in the more powerful PC to try a bit of the old ML again, with much more of a bent for AI this time. I had already been pitched in the early '90's by neural net vendors but I wasn't ready for that. After a few more years of consideration I had a much better idea of the groundwork I would need, separated the raw automation from the intelligent input I was making and that was a good milestone in efficiency right there. I was barely able to get a bit of my concept from 1980 put on to a "powerful" DOS/Windows platform when it crashed and set me back a couple months before starting to get hammered into eventual submission by years of stacked natural disasters.

Growth had been halted but by this time I was pretty mature and charge enough per page that I personally wasn't going to be the one that needed any more automation than I already had. When I first started I could afford to type each page manually on a (intelligent) typewriter to begin with, and I could make even more at today's prices now doing that again if I had too. This was another marginally positive trend that was not very obvious, and it was so marginal that was when I accepted that I would have to actually outlive most of my contemporaries if I was going to make very much of it.

Whew, that wasn't easy and it took a while too.

Even if total automation doesn't make me any more than total manual effort, it is the kind of thing that the bigger multinational groups could really take to the bank. So I've always kept it in mind, I knew about it all along, that's where I got started.

Anyway, there's still a blank tab on an XLS spreadsheet where the tabs to the left are all the "very important" data which I ruminate about then do a little typing accordingly before hitting the button. Then the tabs to the right get populated sequentially and filtered until the final tab spits out a file that gets emailed to the client. It comes straight from Excel with letterhead and fonts virtually indistinguishable from Word. At the beginning with MS-Word I was faxing with a dedicated land line plugged directly into the PC, now email or not when the client prints it there are very few ways to tell the difference from when I would fax them a signed page from my typewriter too.

It took quite a while to reach the point today where AI might be getting close enough in my lifetime to where I could train it to fill in that blank tab for me.

It would have to be about perfect though.

Patience, my friend.

There is/was an experiment to domesticate foxes that began in the early 1950s; something that can only be done slowly generation after generation.

The article conflates a few different "slow" projects, rather than the premise which is efforts that required decades to come to fruition.

He mentions projects started long ago but are still ongoing, like the Sagrada Familia. Then there's innovations from long ago which are still being used, like Linux. Also, he includes ideas which took decades to finally be implemented, like LIGO.

In my opinion, none of these examples are particularly good at demonstrating, "What problems can human beings only solve over a very long period of time?", except for Fermat's Last Theorem.

All technology builds on that which came before, step by step. You can trace Unicode directly back to Morse Code, via various steps like ASCII, Telex, Baudot Code, etc. But the original goal of Morse wasn't to display emojis.

I'd say General Relativity might be a good example, starting with Newton's efforts to quantify the forces of the real world, ending with Einstein's explanation of spacetime. But again, it's not as clear of a problem as Fermat's Last Theorem which was a single problem that required centuries to solve.

AI may be a good example as well, starting with the advent of the digital computer. The very first scientists who worked with them like von Neumann immediately looked forward to the day of an electronic brain. It's taken nearly a century so far and is still underway.

Figuring out a good reason to colonize the solar system

An optimal manufacturing and logistics network for the solar system

Inventing replicators and dispensing with capitalism

How to game HN: always write rebuttals

nitpick:

> That's an accuracy comparable to measuring the distance to the Sun to an accuracy of one atom.

This does not exist, because "the location of the Sun" cannot be defined to the precision of one atom, as the Sun is constantly changing shape and size on a much, much larger scale (easily half the orders of magnitude of the distance to be measured).