It’s a known problem in social sciences that a lot of stuff which would ideally be generalisable is actually only really established by studies on white, educated people in industrialised rich democracies. (WEIRD). It’s like the similar issue that behavioural economics is almost entirely a science of predicting the behaviour of graduate students and experimental psychology is increasingly the study of Amazon Mechanical Turk workers.

But there’s a more subtle issue, I think, in policy making, which is that simply being in a position to influence policy means that you’ve had a very atypical life experience. I first noticed this at a conference a couple of years ago, where we had been discussing the future of artificial intelligence in the workplace. I made a joke (trying to do so politely, because it was a good conference and I was very grateful to the person who invited me, but, I worry, using up a substantial amount of that week’s being-a-dick tokens), along the lines of:

“Haven’t you made a big mistake by inviting all these intelligent people from high performing institutions? Of course they’re going to be optimistic. If you wanted to know what the future of AI in the workplace would really be like, you ought to be talking to mediocre people from slightly dysfunctional companies who haven't really thought about what they're doing. Because that’s going to be the majority of actual use cases!”

I thought I was joking, but ever since I said that it’s been hard to stop seeing it everywhere. Particularly, almost everyone in a position of power has had a completely unusual relationship with the education system. It’s not just a matter of not knowing the price of a pint of milk or whatnot, it’s reacting favourably to being offered training. Or enjoying having things explained to you. If you’ve ever looked at the disclosures attached to your investment statements and wondered “who the hell reads this crap”, the answer is “people who become financial regulators”.

Not an easy problem to solve though, because my joke was indeed a joke. If you invite a load of mediocre middle managers from bad institutions to your conference, you’re just really going to get some subjects for anthropological study. They are not going to be able to articulate where they’re going wrong, obviously. And they may not even really be able to describe what they’re trying to do, or notice what effect it’s actually having.

Similarly, poor people have lived experience of difficult interactions with the state, but that doesn’t mean that they understand how to fix them, and they might be quite likely to focus on things which are immediately irritating rather than actually causing the problems. User experience research is difficult and expensive, and is generally done by equally WEIRD and Very Special people.

Policy and business are for the most part run by The Very Special, and there is no getting away from that. So they will always, to a greater of lesser extent, reflect the kind of things that Very Special People experience and regard as important. The trick to stopping this becoming pathological is to remind yourself of the most atypical parts of the experience of Very Specialness (being praised, passing exams, etc), and take a regular quick check that the system isn’t developing excessive tendencies to self-soothe its creators or being optimised for precisely the people who don't need it.

The Five Levels: from Spicy Autocomplete to the Dark Factory

0. Spicy autocomplete, aka original GitHub Copilot or copying and pasting snippets from ChatGPT.
1. The coding intern, writing unimportant snippets and boilerplate with full human review.
2. The junior developer, pair programming with the model but still reviewing every line.
3. The developer. Most code is generated by AI, and you take on the role of full-time code reviewer.
4. The engineering team. You're more of an engineering manager or product/program/project manager. You collaborate on specs and plans, the agents do the work.
5. The dark software factory, like a factory run by robots where the lights are out because robots don't need to see.

Dan says about that last category:

At level 5, it's not really a car any more. You're not really running anybody else's software any more. And your software process isn't really a software process any more. It's a black box that turns specs into software.

Why Dark? Maybe you've heard of the Fanuc Dark Factory, the robot factory staffed by robots. It's dark, because it's a place where humans are neither needed nor welcome.

I know a handful of people who are doing this. They're small teams, less than five people. And what they're doing is nearly unbelievable -- and it will likely be our future.

I've talked to one team that's doing the pattern hinted at here. It was fascinating. The key characteristics:

• Nobody reviews AI-produced code, ever. They don't even look at it.
• The goal of the system is to prove that the system works. A huge amount of the coding agent work goes into testing and tooling and simulating related systems and running demos.
• The role of the humans is to design that system - to find new patterns that can help the agents work more effectively and demonstrate that the software they are building is robust and effective.

It was a tiny team and they stuff they had built in just a few months looked very convincing to me. Some of them had 20+ years of experience as software developers working on systems with high reliability requirements, so they were not approaching this from a naive perspective.

I'm hoping they come out of stealth soon because I can't really share more details than this.

Tags: ai, generative-ai, llms, ai-assisted-programming, coding-agents

If you watch American cooking shows, you’ve likely experienced “salad confusion”. You see a chef preparing what looks like rocket, but they call it arugula.

It’s the same plant (Eruca sativa). It has the same peppery bite. So why do English speakers use two completely different names?

The answer isn’t just a quirk of translation. It is a linguistic fossil record revealing the history of Italian migration.

The name you use tells us less about the vegetable and more about who introduced you to it.

A Latin word with a double life

It all starts with the Latin word eruca.

Crucially, this term had a dual meaning. It referred to the vegetable, but also meant “caterpillar” – maybe because the plant’s hairy stems resembled the pests often found on brassicas.

As the Roman Empire faded and Vulgar Latin (the language of the vulgus, or the common people) evolved into the Romance languages, this single word split along two paths.

The Northern route: aristocratic ‘rocket’

As the word travelled north through Italy, it morphed from eruca into the Northern Italian diminutive ruchetta.

From there, it crossed the Alps into France, becoming roquette.

By the 16th century, French culinary influence was dominant in England. The first written record appears in 1530, in John Palsgrave’s text L'esclarcissement de la langue francoyse (Clarification of the French Language – said to be the first grammar of French for English speakers), translating roquette to “rocket”.

By 1597, English botanist John Gerard was describing “garden rocket” in his large illustrated Herball, or Generall Historie of Plantes, cementing it in the British lexicon.

This terminology travelled with the First Fleet. In Australia, “rocket” was a colonial staple, not a modern discovery. Planting guides in the Hobart Town Courier from 1836 list rocket alongside other brassicas, such as cress and mustard, as essential kitchen garden crops.

This is why people in Australia, the United Kingdom and New Zealand say “rocket”. For these speakers, the word followed an aristocratic, pre-industrial path.

The Southern route: migrant ‘arugula’

In the United States, the word “arugula” didn’t arrive in books; it arrived in pockets.

In the late 19th and early 20th centuries, millions of Italians emigrated to the US. This was a mass migration of the working class, predominantly from Southern regions like Calabria and Sicily.

These migrants spoke regional languages, erroneously called dialects, rather than Standard Italian.

In the South, eruca had evolved differently. We can trace this in historical dictionaries: Gerhard Rohlfs’ monumental Dictionary of the Three Calabrias (1932–39) records the local word as arùculu.

Similarly, Antonino Traina’s Sicilian-Italian dictionary (1868) lists the variant aruca.

When Italian immigrants established market gardens in New York, they sold the produce using their dialect forms. They weren’t selling the French roquette; they were selling the Calabrian arùculu.

Over many years, this solidified into the American English “arugula”.

For decades, arugula was an “ethnic” ingredient in the US, underscoring its origins as “an unruly weed that was foraged from the fields by the poor”. It wasn’t until a New York Times article on May 24 1960 that food editor Craig Claiborne introduced it to a wider audience.

Noting it “has more names than Joseph’s coat had colors”, he used the New York market term “arugula” alongside rocket in his recipes, inadvertently codifying it as the standard American name.

There is perhaps a sense that “arugula” might come from Spanish, given the influence of words like cilantro in American culinary terminology.

In Spanish, Latin eruca evolved into oruga which is uncannily similar to “arugula”. But, linguistically things are a little more complex.

While the Spanish word maintains the reference to the plant it also retains the Latin term’s double meaning: a salad vegetable and a caterpillar.

According to Bréal’s Law of Differentiation, named for the linguist Michel Bréal, languages detest absolute synonyms. If a word has two meanings, the language will intervene somehow. Indeed, today’s Spanish speakers prefer to call the plant rúcula. If you were to ask for an ensalada de oruga in Spain today, you’d probably get odd looks and, maybe, a caterpillar salad.

What about ‘rucola’?

So where does the word rucola – seen on menus in Rome today – fit in?

While the Anglosphere was splitting into rocket and arugula, Italy was undergoing its own linguistic unification. Standard Italian rucola is another diminutive which gradually won out over other regional variants.

Philologically, rucola represents a middle ground. Its rise in usage in Italy in the second half of the 20th Century eclipsed competing terms like rughetta, ruchetta or ruca.

Rucola now has international reverberations. The preferred term in Spanish is modelled on it, it appears in many other European languages and it is making inroads in the English lexicon.

From peasant weed to political symbol

By the 1990s in the US, the “peasant weed” had completed a remarkable social climb. It became a political shibboleth for the American “liberal elite” (most famously during Obama’s Arugula-gate in the 2008 presidential election campaign).

Meanwhile, in Australia, rocket popped up as the ubiquitous garnish of the cafe culture boom found on everything from pizza to smashed avo.

The divide is a reminder that language is rarely accidental. When an Australian orders “rocket,” they echo a 16th-century exchange with France. When an American orders “arugula”, they echo the voices of Southern Italian migrants in 1920s New York. And when someone uses “rucola” perhaps it’s a way of evoking Italy’s mythical, UNESCO-awarded gastronomy.

Matt Absalom does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

I saw someone online today saying they enjoyed Python Programming Exercises, Gently Explained.

I went to the author’s site and I saw that he makes all of his books (Automate the Boring Stuff with Python and more) available for free at https://inventwithpython.com/. I’d heard of his books many times before but I didn’t realize that he made them available on his site like that!

I’m Al Sweigart, and I write books to teach beginners to code. I put them online for free because programming is too valuable and needs to be accessible to all.

I always think it’s so cool when authors do that, another example is Mark Dominus’s Higher Order Perl

Content warning:tw trauma

To the end, friends. https://www.youtube.com/watch?v=uIfqSTBTJXQ&list=RDuIfqSTBTJXQ&start_radio=1