Solving problems with LLMs is like solving front-end problems with NPM: the “solution” comes through installing more and more things — adding more and more context, i.e. more and
more packages.
LLM: Problem? Add more context.
NPM: Problem? There’s a package for that.
…
As I’m typing this, I’m thinking of that image of the evolution of the Raptor engine, where it evolved in simplicity:
This stands in contrast to my working with LLMs, which often wants more and more context from me to get to a generative solution:
…
Jim Nielsen speaks to my experience, here. Because a programming LLM is simply taking inputs (all of your code, plus your prompt), transforming it through statistical analysis, and then
producing an output (replacement code), it struggles with refactoring for simplicity unless very-carefully controlled. “Vibe coding” is very much an exercise in adding hacks upon hacks…
like the increasingly-ludicrous epicycles introduced by proponents of geocentrism in its final centuries before the heliocentric model became fully accepted.
This mess used to be how many perfectly smart people imagined the movements of the planets. When observations proved it couldn’t be right, they’d just add more
complexity to catch the edge cases.
I don’t think that AIs are useless as a coding tool, and I’ve successfully used them to good effect on
several occasions. I’ve even tried “vibe coding”, about which I fully agree with Steve Krouse‘s observation that
“vibe code is legacy code”. Being able to knock out something temporary, throwaway, experimental, or for personal use only… while I work on
something else… is pretty liberating.
For example: I couldn’t remember my Google Sheets API and didn’t want to re-learn it from the sprawling documentation site, but wanted a quick personal tool to manipulate such a sheet
from a remote system. I was able to have an AI knock up what I needed while I cooked dinner for the kids, paying only enough attention to check-in on its work. Is it accessible? Is it
secure? Is it performant? Is it maintainable? I can’t answer any of those questions, and so as a professional software engineer I have to reasonably assume the answer to
all of them is “no”. But its only user is me, it does what I needed it to do, and I didn’t have to shift my focus from supervising children and a pan in order to throw it together!
Anyway: Jim hits the nail on the head here, as he so often does.
“For years, starting in the late ‘70s, I was taking pictures of hitchhikers. A hitchhiker is someone you may know for an hour, or a day, or, every so often, a little longer, yet,
when you leave them, they’re gone. If I took a picture, I reasoned, I’d have a memory. I kept a small portfolio of photos in the car to help explain why I wanted to take their
picture. This helped a lot. It also led me to look for hitchhikers, so that I could get more pictures.
“I almost always had a camera… I finally settled on the Olympus XA – a wonderful little pocket camera. (I’ve taken a picture of the moon rising with this camera.) One time I asked a
chap if I could take a photo, and he said, “You took my picture a few years ago.” I showed him the album and he picked himself out. “That’s me,” he said, pointing…”
…
Not that hitch-hiking is remotely as much a thing today as it was 50 years ago, but even if it were then it wouldn’t be so revolutionary to, say, take a photo of everybody you give a
ride to. We’re all carrying cameras all the time, and the price of taking a snap is basically nothing.
But for Doug Biggert, who died in 2023, began doing this with an analogue camera as he drove around California from 1973 onwards? That’s quite something. Little wonder he had
to explain his project to his passengers (helped, later on, by carrying a copy of the photo album he’d collected so-far that he could show them).
A really interesting gallery with a similarly-compelling story. Also: man – look at the wear-and-tear on his VW Bug!
Some time in the last 25 years, ISPs stopped saying they made you “part of” the Internet, just that they’d help you “connect to” the Internet.
Most people don’t need a static IP, sure. But when ISPs stopped offering FTP and WWW hosting as a standard feature (shit though it often was), they became part of the tragic process by
which the Internet became centralised, and commoditised, and corporate, and just generally watered-down.
The amount of effort to “put something online” didn’t increase by a lot, but it increased by enough that millions probably missed-out on the opportunity to create
their first homepage.
Seventy years ago, residents of this part of London would take shelter from V1 and V2 bombs in a tunnel beneath my feet. And today, I’m going down there to take a look!
It started with a fascination after discovering a little-known stone circle near my new house. It grew into an obsession with the history of the place.
Two years later, our eldest was at school and her class was studying the stone age. Each of three groups were tasked with researching a particular neolithic monument, and our eldest was
surprised when she heard my voice coming from a laptop elsewhere in the class. One of her classmates had, in their research into the Quoits, come across my video.
It turns out “local expert” just means “I read the only book ever written about the archaeology of the stones, and a handful of ancillary things.”
And so this year, when another class – this time featuring our youngest – went on a similar school trip, the school asked me to go along again.
I’d tweaked my intro a bit – to pivot from talking about the archaeology to talking about the human stories in the history of the place – and it went down well: the
children raised excellent observations and intelligent questions1,
and clearly took a lot away from their visit. As a bonus, our visit falling shortly after the summer solstice meant that local neopagans had left a variety of curious offerings – mostly
pebbles painted with runes – that the kids enjoyed finding (though of course I asked them to put each back where they were found afterwards).
But the most heartwarming moment came when I later received an amazing handmade card, to which several members of the class had contributed:
I particularly enjoy the pencil drawing of me talking about the breadth of Bell Beaker culture, with a child
interrupting to say “cool!”.
I don’t know if I’ll be free to help out again in another two years, if they do it again2: perhaps I
should record a longer video, with a classroom focus, that shares everything I know about The Devil’s Quoits.
But I’ll certainly keep a fond memory of this (and the previous) time I got to go on such a fun school trip, and to be an (alleged) expert about a place whose history I find so
interesting!
Footnotes
1 Not every question the children asked was the smartest, but every one was gold.
One asked “is it possible aliens did it?” Another asked, “how old are you?”, which I can only assume was an effort to check if I remembered when this 5,000-year-old hengiform monument
was being constructed…
2 By lucky coincidence, this year’s trip fell during a period that I was between jobs, and
so I was very available, but that might not be the case in future!
This wonderful project, released six weeks ago, attempts the impossible challenge of building a Civilization-style tech tree but chronicling the development and interplay
of all of the actual technological innovations humanity has ever made. Even in its inevitably-incomplete state, it’s inspiring and informative. Or, as Open Culture put it:
Our civilization has made its way from stone tools to robotaxis, mRNA vaccines, and LLM chatbots; we’d all be better able to inhabit it with even a slightly clearer idea of how it
did so.
This video, which I saw on Nebula but which is also available on YouTube, explores a hypothetical alternate history in which the Schuman Plan/European Coal & Steel Community never happened, and the knock-on effects lead to no EU, a more fragmented Europe,
and an ultimately more-fractured and more-complicated Europe of the late 20th/early 21st century.
Obviously it’s highly-speculative and you could easily come up with your own alternative alternative history! But the Twilight Struggle player in me as well as the alternate
history lover (and, of course, European Union fan) especially loves the way this story is told.
It’s worth remembering that for the last half-millenium or more, the default state of Europe has been to be fighting one another: if not outright war then at least agressive
economic and political rivals. Post-WWII gave Europe perhaps its longest ever period of relative peace, and that’s great enough that all of the other benefits of a harmonised
and cooperative union are just icing on the cake.
EU Made Simple is a fantastic channel in general, and I’d recommend you give it a look. It ties news and history in with its creators
outlook, but it’s always clear which bits are opinion and it’s delightfully bitesized. For Europeans-in-exile in this post-Brexit age, it’s hopeful and happy, and I like it.
There’s a small house ( ⌂ ) in the middle of IBM’s infamous character set Code Page 437. “Small house”—that’s the official IBM
name given to the glyph at code position 0x7F, where a control character for “Delete” (DEL) should logically exist. It’s cute, but a little strange. I wonder, how did it get there?
Why did IBM represent DEL as a house, of all things?
It’s highly accessible, so even if you’ve only got a passing interest in, I don’t know, text encoding or typography or the history of computing, it’s a great read.
My love of the yesterweb forced me to teach myself just-enough Blender to make an animation for a stupid thing: an 88×31 button representing “me” (and, I suppose, my blog, whenever I
next end up redesigning its theme).
On Wednesday, Vodafone
announced that they’d made the first ever satellite video call from a stock mobile phone in an area with no terrestrial signal. They used a mountain in Wales for their experiment.
It reminded me of an experiment of my own, way back in around 1999, which I probably should have made a bigger deal of. I believe that I was the first person to ever send an email from
the top of Yr Wyddfa/Snowdon.
Nowadays, that’s an easy thing to do. You pull your phone out and send it. But back then, I needed to use a Psion 5mx palmtop, communicating over an infared link using a custom driver
(if you ever wondered why I know my AT-commands by heart… well, this isn’t exactly why, but it’s a better story than the truth) to a Nokia 7110 (fortunately it was cloudy enough to not
interfere with the 9,600 baud IrDA connection while I positioned the devices atop the trig point), which engaged a GSM 2G connection, over which I was able to send an email to myself,
cc:’d to a few friends.
It’s not an exciting story. It’s not even much of a claim to fame. But there you have it: I was (probably) the first person to send an email from the summit of Yr Wyddfa. (If you beat
me to it, let me know!)
You’re probably familiar with the story of George and Robert Stephenson’s Rocket, a pioneering steam locomotive built in 1829.
If you know anything, it’s that Rocket won a competition and set the stage for a revolution in railways lasting for a century and a half that followed. It’s a cool story, but
there’s so much more to it that I only learned this week, including the bonkers story of 19th-century horse-powered locomotives.
The Rainhill Trials
Ten teams submitted applications to enter the Rainhill Trials, but only five actually took part. Four of these were the steam locomotives illustrated above.
Over the course of the 1820s, the world’s first inter-city railway line – the Liverpool & Manchester Railway – was constructed. It wasn’t initially anticipated that the new railway
would use steam locomotives at all: the technology was in its infancy, and the experience of the Stockton & Darlington railway, over on the other side of the Pennines, shows
why.
The Stockton & Darlington railway was opened five years before the new Liverpool & Manchester Railway, and pulled its trains using a mixture of steam locomotives and horses1.
The early steam locomotives they used turned out to be pretty disastrous. Early ones frequently broke their cast-iron wheels so frequently; some were too heavy for the lines and needed
reconstruction to spread their weight; others had their boilers explode (probably after safety valves failed to relieve the steam pressure that builds up after bringing the vehicle to a
halt); all got tied-up in arguments about their cost-efficiency relative to horses.
Nowadays, a train can be cancelled and a paying customer might barely get a half-hearted apology and a spot on a crowded rail replacement bus. But back in 1826 even the crew
of a broken-down train might be offered a copious allowance of beer to keep them motivated. Scan from page 119 of The North Eastern Railway; its rise
and development, by William Weaver Tomlinson.
Nearby, at Hetton colliery – the first railway ever to be designed to never require animal power – the Hetton Coal Company had become so-dissatisfied with the reliability and
performance of their steam locomotives – especially on the inclines – that they’d had the entire motive system. They’d installed a cable railway – a static steam engine pulled the mine
carts up the hill, rather than locomotives.
This kind of thing was happening all over the place, and the Liverpool and Manchester Railway Company were understandably cautious about hitching their wagon to the promise of steam
locomotives on their new railway. Furthermore, they were concerned about the negative publicity associated with introducing to populated areas these unpopular smoke-belching engines.
But they were willing to be proven wrong, especially after George Stephenson pointed out that this new, long, railway could find itself completely crippled by a single breakdown were it
to adopt a cable system. So: they organised a competition, the Rainhill Trials, to allow locomotive engineers the chance to prove their engines were up to the challenge.
When the Stockton & Darlington line began serving passengers in 1826, their advertisements only ever showed passenger coaches being pulled by horses, never steam locomotives.
The challenge was this: from a cold start, each locomotive had to haul three times its own weight (including their supply of fuel and water), a mile and three-quarters (the first and
last eighth of a mile of which were for acceleration and deceleration, but the rest of which must maintain a speed of at least 10mph), ten times, then stop for a break before doing it
all again.
Four steam locomotives took part in the competition that week. Perseverance was damaged in-transit on the way to the competition and was only able to take part on the
last day (and then only achieving a top speed of 6mph), but apparently its use of roller bearing axles was
pioneering. The very traditionally-designed Sans Pareil was over the competition’s weight limit, burned-inefficiently (thanks perhaps to an overenthusiastic
blastpipe that vented unburned coke right out of the funnel!), and broke down when one of its cylinders cracked2.
Lightweight Novelty – built in a hurry probably out of a fire engine’s parts – was a crowd favourite with its integrated tender and high top speed, but kept breaking
down in ways that could not be repaired on-site. And finally, of course, there was Rocket, which showcased a combination of clever innovations already used in steam
engines and locomotives elsewhere to wow the judges and take home the prize.
But there was a fifth competitor in the Rainhill Trials, and it was very different from the other four.
Cycloped
When you hear the words horse-powered locomotive, you probably think of a horse-drawn train. But that’s not a locomotive: a locomotive is a vehicle that, by definition, propels
itself3.
Which means that a horse-powered locomotive needs to carry the horse that provides its power…
If this isn’t the most-zany railway vehicle you’ve ever seen, please share what beats it.
…which is exactly what Cycloped did. A horse runs on a treadmill, which turns the wheels of a vehicle. The vehicle (with the horse on it) move. Tada!4
You might look at that design and, not-unreasonably, decide that it must be less-efficient than just having the horse pull the damn vehicle in the first place. But that isn’t
necessarily the case. Consider the bicycle which can transport itself and a human both faster and using less-energy than the human would achieve by walking. Or look at wind
turbine powered vehicles like Blackbird, which was capable of driving under wind
power alone at three times the speed of a tailwind and twice the speed of a headwind. It is mechanically-possible to improve the speed and efficiency of a
machine despite adding mass, so long as your force multipliers (e.g. gearing) is done right.
I’ve long loved this 2010 photo of Blackbird, simultaneously showing a flag (blowing left, with the wind) and a streamer (blowing right, as a result of the wind-powered
vehicle’s speed) demonstrating that it is travelling against the wind, but significantly faster than the wind.
Cycloped didn’t work very well. It was slower than the steam locomotives and at some point the horse fell through the floor of the treadmill. But as I’ve argued above, the
principle was sound, and – in this early era of the steam locomotive, with all their faults – a handful of other horse-powered locomotives would be built over the coming
decades.
Over in the USA, the South Carolina Canal and Railroad Company successfully operated a passenger service using the Flying Dutchman, a horse-powered locomotive with twelve seats
for passengers. Capable of travelling at 12mph, this demonstrated efficiency multiplication over having the same horse pull the vehicle (which would either require fewer
passengers or a dramatically reduced speed).
This strange contraption was eventually replaced with a steam train, under the understanding that improvements in steam locomotive technology would continue to develop faster than
advancements in techniques for the selective breeding of horses.
As late as the early 1850s, people were still considering this strange approach. The 1851 Great Exhibition at the then brand-new Crystal Palace featured Impulsoria, which
represents probably the pinnacle of this particular technological dead-end.
Capable of speeds up to 20mph, it could go toe-to-toe with many contemporary steam locomotives, and it featured a gearbox to allow the speed and even direction of travel to be
controlled by the driver without having to adjust the walking speed of the two to four horses that provided the motive force.
The reins now arriving on platform one is the Mane Line service to Carlisle. Mind the gallop. Stand clear of the hackamore.
Personally, I’d love to have a go on something like the Flying Dutchman: riding a horse-powered vehicle with the horse is just such a crazy idea, and a road-capable
variant could make for a much better city tour vehicle than those 10-person bike things, especially if you’re touring a city with a particularly equestrian history.
Footnotes
1 From 1828 the Stockton & Darlington railway used horse power only to pull their
empty coal trucks back uphill to the mines, letting gravity do the work of bringing the full carts back down again. But how to get the horses back down
again? The solution was the dandy wagon, a special carriage that a horse rides in at the back of a train of coal
trucks. It’s worth looking at a picture of one, they’re brilliant!
2 Sans Pareil’s cylinder breakdown was a bit of a spicy issue at the time because its
cylinders had been manufactured at the workshop of their rival George Stephenson, and turned out to have defects.
3 You can argue in the comments whether a horse itself is a kind of locomotive. Also – and
this is the really important question – whether or not Fred Flintstone’s car, which is propelled by his feed, is a kind locomotive or not.
4 Entering Cycloped into a locomotive competition that expected, but didn’t
explicitly state, that entrants had to be a steam-powered locomotive, sounds like exactly the kind of creative circumventing of the rules that we all loved Babe (1995) for. Somebody should make a film about Cycloped.
Hypothetically-speaking, what would happen if convicted felon Donald Trump were assassinated in-between his election earlier this month and his inauguration in January?
There’ve been at least two assassination attempts so far, so it’s not beyond the realm of possibility that somebody will have another go at some point1.
Hello, Secret Service agents! Thanks for visiting my blog. I assume I managed to get the right combination of keywords to hit your watchlist. Just to be clear, this is
an entirely hypothetical discussion. I know that you’ve not always been the smartest about telling fiction from reality. But as you’ll see, I’m
just using the recent assassination attempts as a framing device to talk about the history of the succession of the position of President-Elect. Please don’t shoot me.
If the US President dies in office – and this happens around 18% of the time2 – the
Vice-President becomes President. But right now, convicted felon Donald Trump isn’t President. He’s President-Elect, which is a term used distinctly from
President in the US Constitution and other documents.
This card was pretty-much nerfed by Wizards’ ruling that Presidents-Elect, Vice-Presidents etc. were not (yet) kinds of President.
It turns out that the answer is that the Vice-President-Elect becomes President at the inauguration. This boring answer came to us through three different Constitutional Amendments,
each with its own interesting tale.
The Twelfth Amendment (1804) mostly existed to reform the Electoral College. Prior to the adoption of the Twelfth Amendment, the Electoral College members each cast two
ballots to vote for the President and Vice-President, but didn’t label which ballot was which position: the runner-up became Vice-President. The electors would carefully
and strategically have one of their number cast a vote for a third-party candidate to ensure the person they wanted to be Vice-President didn’t tie with the person they wanted to be
President. Around the start of the 19th century this resulted in several occasions on which the President and Vice-President had been bitter rivals but were now forced to work
together3.
While fixing that, the Twelfth Amendment also saw fit to specify what would happen if between the election and the inauguration the President-Elect died: that the House of
Representatives could choose a replacement one (by two-thirds majority), or else it’d be the Vice-President. Interesting that it wasn’t automatically the Vice-President,
though!
It didn’t happen like this. In real life, there was a lot less singing, and a lot more old white men.
The Twentieth Amendment (1933) was written mostly with the intention of reducing the “lame duck” period. Here in the UK, once we elect somebody, they take power
pretty-much immediately. But in the US, an election in November traditionally resulted in a new President being inaugurated almost half a year later, in March. So the Twentieth
Amendment reduced this by a couple of months to January, which is where it is now.
In an era of high-speed road, rail, and air travel and digital telecommunications even waiting from November to January seems a little silly, though. In any case, a secondary feature of
the Twentieth Amendment was that it removed the rule about the House of Representatives getting to try to pick a replacement President first, saying that they’d just fall-back on the
Vice-President in the first instance. Sorted.
Just 23 days later, the new rule almost needed to be used, except that Franklin D. Roosevelt’s would-be assassin Giuseppe Zangara missed his tricky shot.
The Twentieth Amendment (1967) aimed to fix rules-lawyering. The constitution originally said that f the President is removed from office, dies, resigns, or is
otherwise unable to use his powers and fulfil his duties, then those powers and duties go to the Vice-President.
Note the wording there. The constitution said that if a President died, their their duties and powers would go to the Vice-President. Not the Presidency itself. You’d
have a Vice-President, acting as President, who wasn’t actually a President. And that might not matter 99% of the time… but it’s the edge cases that get you.[foonote]Looking
for some rules-lawyering? Okay: what about rules on Presidential term limits? You can’t have more than two terms as President, but what if you’ve had a term as Vice-President
but acting with Presidential powers after the President died? Can you still have two terms? This is the kind of constitutional craziness that munchkin US history scholars get
off on.[/footnote]
It also insisted that if there’s no Vice-President, you’ve got to get one. You’d think it was obvious that if the office of Vice-President exists in part to provide a “backup” President
in case, y’know, the nearly one-in-five chance that the President dies… that a Vice-President who finds themselves suddenly the President would probably want to have one!
But no: 18 Presidents4served without a Vice-President for at least some of their
term: four of them never had a Vice-President. That includes 17th
President Andrew Johnson, who you’d think would have known better. Johnson was Vice-President under Abraham Lincoln until, only a month after the inauguration, Lincoln was assassinated,
putting Johnson in change of the country. And he never had a Vice-President of his own. He served only barely shy of the full four years without one.
Anyway; that was a long meander through the history of the Constitution of a country I don’t even live in, to circle around a question that doesn’t matter. The thought randomly came to
me while I was waiting for the traffic lights at the roadworks outside my house to change. And now I know the answer.
Very hypothetically, of course.
Footnotes
1 My personal headcanon is that the would-be assassins are time travellers from the
future, Chrononauts-style, trying to flip a linchpin and bring about a stable future in which he wasn’t elected. I
don’t know whether or not that makes Elon Musk one of the competing time travellers, but you could conceivably believe that he’s Squa Tront in disguise, couldn’t you?
2 The US has had 45 presidents, of whom eight have died during their time in office. Of
those eight, four – half! – were assassinated! It’s a weird job. 8 ÷ 45 ≈ 18%.
3 If you’re familiar with Hamilton, you’ll recall its characterisation of the
election of 1800 with President Thomas Jefferson dismissing his Vice-President Aaron Burr after a close competition for the seat of President which was eventually settled when
Alexander Hamilton instructed Federalist party members in the House of Representatives to back Jefferson over Burr. The election result really did happen like that – it seems that
whichever Federalist in the Electoral College that was supposed to throw away their second vote failed to do so! – but it’s not true that he was kicked-out by Jefferson: in fact, he
served his full four years as Vice-President, although Jefferson tried to keep him as far from actual power as possible and didn’t nominate him as his running-mate in 1804. Oh, and in
1807 Jefferson had Burr arrested for treason, claiming that Burr was trying to capture part of the South-West of North America and force it to secede and form his own country: the
accusation didn’t stick, but it ruined Burr’s already-faltering political career. Anyway, that’s a diversion.
4 17 different people, but that’s not how we could Presidents apparently.