1. Compile times. OCaml compiles very quickly (think Go) and this gives you a fast feedback loop. It lets the LLM rip through the implementation.
2. OCaml has an idiomatic approach of using interface files for all modules that can be accessed outside their libraries, and interface files give LLMs a great précis of exactly what is wanted. They then just have to follow the types and fill in the blanks to get the implementation. It's nearly the perfect use case.
Both force the model to reason about types too much. The more constrained type system plus global inference merely holds them to what they already wrote, with very fast feedback.
Rust is fine if you need it but most things don’t and ocaml has more convenient abstractions for “regular” work. If you need rust you need rust but ocaml isn’t that far off in perf.
Haskell type system is too expressive, it itself becomes a place for the agent to make mistakes and get bogged down.
For web frontend, it's derivative programming language, rescript, is also a secret weapon because of types, fast compilation times and first class support of React. Using React in rescript is much nicer than TypeScript.
Yes I contributed code to a paper on it this spring. It’s hard to measure and “studying” it just means feeding money into frontier models. If anyone has a few billion tokens for a couple masters students hmu so they can get it published.
I'm interested too - besides the paper, is it your personal experience that working with llms on ocaml is superior to other languages? I thought with the sheer amount of training data languages like python and java would be "ideal"
A couple software examples off the top of my head are Richard Hipp with SQLite and Matthew Ahrens with ZFS. Both almost incomprehensibly sophisticated projects executed well over decades.
Neither are anything like a solo effort in the way these examples are, and it's hard to say what they'd be like without the long individual involvement.
Yeah that comment isn't quite fair but Berry does have a massive blindspot around labor. His entire like, deal, is built on the nearly-unmentioned-in-his-entire-works fact that he's describing a world of agrarian landowners who do relatively little productive work themselves. We can't all be gentleman farmers!
The fact that his writings are still so good and useful and interesting speaks incredibly highly of his powers of perception and skill as a writer. But it makes them nearly useless in a "how ought we to be" conversation, which is one of the main things his works engage with. So it really is a huge problem with his entire project.
I think, at least for the next few years, certain preferences and consensuses were accidentally embedded in the models to your benefit.
They love htmx! Claude will independently suggest using it instead of more general js frameworks for smaller projects. A gpt model told me, and I wrote this down because it was striking, "htmx won the argument it was making."
It's the same thing with rust. I'm an ocaml expert and ocaml genuinely fits a lot of the problems I work with better regardless of that. But because of when the models were trained and what they were trained on, a certain sort of alt-consensus opinion is frozen into them, and it believes rust is a better choice for nearly everything!
I think you're good for the next few years at least. It's an open question what happens after that dead internet theory recursive loop of llms trained on llm output etc so who knows you might be good forever.
We will see. Changing things around a bit (especially flipping attribute inheritance to explicit by default) may confuse a lot of models. Oh well, it's the right thing ¯\_(ツ)_/¯
Please don’t misunderstand my comment, I wasn't trying to diminish the brilliant work you’ve done on HTMX, which I’ve personally benefited from. I could have probably worded it better than I did. I'm also not saying you _did_ misunderstand it, only asking that you please don't. I was just sharing an observation that I realize is not everyone’s experience. I’m still a fan + supporter, even though I now reach for JS more often because the pain of writing it has largely disappeared. I’m still not a fan of it by any means, but LLMs have been a major mitigator. I'm always happy to see the HTMX updates when they come out.
> Palestine Action has been designated a terrorist group in the UK as well after sabotaging military aircraft
This is already an extreme perversion of justice all on its own. Damaging military equipment is absolutely outside any useful definition of terrorism. This is simply armed conflict.
Damaging property for political ends has been part of the definition in the UK, and I think the US, for decades. I would say the case for targetting military equipment is stronger than for damaging most other things.
> This is simply armed conflict.
Armed conflict by residents and citizens against the state is usually a serious criminal offense everywhere.
> Damaging property for political ends has been part of the definition in the UK, and I think the US, for decades.
Does that include property in Gaza, the West Bank or Lebanon? Because if the people who spray painted something are "terrorists", I want to know what the people who wiped Rafah off the face of the earth are called.
Israelis and others who support this have shown that the lesson they took from the Holocaust is that they believe it's morally fine as long as they're the ones doing the killing.
Well the definition is flawed, or as the British say, the law is an ass.
Terrorism is widely understood to be the deployment of military force against civilian populations. Of course states do this all the time, but 'state terrorism' is basically dependent on whether politicians in country A align with country B or not; if they don't, they'll call out state abuses in country B as state terrorism, if not they'll say it's law and order.
Regardless of these nuances, extending the definition to include any kind of property damage is just laughably opportunist.
"Any kind" is quite the load bearing argument there. Let me remind you Europe is under an ongoing land invasion, which is why fucking with its defensive capabilities is unlikely to solicit much sympathies.
Bad argument. If you sabotage military equipment in Europe, obviously you also indirectly lower the military potential of that army somewhere else. The planes in Europe aren’t there to be paperweights, they are either being maintained there for service or are there as a deterrent and need to be reinforced by other planes when damaged.
Bad argument. You're saying the French were wrong to sabotage Hitler's airplanes because they might've been used after the war to defend Germany from the USSR.
If you think European defense is analogous to hitler defending Germany in WWII, we should have started the discussion at a very different point. Do you really think that?
> Armed conflict by residents and citizens against the state is usually a serious criminal offense everywhere.
I think it’s worth considering that you’re much more likely to be a resident or a citizen than a state when you’re deciding how you feel about these kinds of things.
Damaging military equipment is a military end, not a political one. If it were otherwise, then a military attacking a military would definitionally be terrorism. At that point, you've removed the utility of the word "terrorism" from English and made it synonymous with "armed conflict."
Terrorism being done chiefly to change policies by terrorizing people through symbolic acts is a constrained, generally-accepted definition of terrorism.
Contrast with 9/11, where symbols of American capitalism, filled with civilians, were attacked in order to terrify people who have nothing to do with those buildings into supporting changes in policy that hurt the US.
I suppose the UK could declare Palestine Action as combatants and hunt them down and shoot them.. but that doesn't seem very uh, appropriate. I sort of prefer them being arrested and going through courts.
Note that the UK would be committing war crimes in this case, since you are not allowed to murder unarmed soldiers, you are supposed to capture them and put them in POW camps.
Which is typically called „arresting“ when a government does it to their own citizens…
That's also false, you can absolutely kill unarmed combatants so long as they remain combatants. Eg you can bomb an artillery crew even if they don't have sidearms.
To be eligible for capture, a combatant must generally surrender.
> typically called „arresting“ when a government does it to their own citizens
I encourage you to read up on POW camps. There are no attorneys or rights. You may work POWs. There are cases in which food and medical aid can be severely limited depending on how it is sourced. You may hold POWs functionally forever.
Source that you wouldn’t be allowed to kill unarmed enemy soldiers? Imagine a scouting team doing reconnaissance, surely you’re allowed to attack and kill them?
Just because it's part of the government's definition doesn't mean that it's true. It's in the government's interest to define as much as possible as terrorism.
The labeling of groups as "terrorist" is intended to justify authorization for military action against those groups, where, prior to the Global War on Terror, state affiliation or outright sponsorship was typically needed for a military response.
I am not an expert on the legalities. My impression is that, if they are terrorists, you kind of have to catch them in the act, or preparing to act, or providing support for actions. An army, in wartime, can be attacked in broader circumstances.
Are you implying that PA is a military organization that is at war with the UK? I don't think you actually have thought through the implications of that. If this were the case, they would be considered legitimate military targets who can be killed.
That would be a lot worse for them than sanctions.
> Damaging military equipment is absolutely outside any useful definition of terrorism.
I hope no one takes this seriously.
If you're caught sabotaging equipment on a US military installation in many cases they are authorized to use deadly force and you'll be held accountable under the UCMJ. The absolute best outcome is you're designated a terrorist and live to see a prison cell.
> Terrorism used to be defined as attacking civilian targets to aid in political messaging.
Not consistently, and its not a clear definition.
The same groups that attacked civilians in places I have lived in also attacked military targets as part of the same strategy. Terrorism is used to threaten, divert resources from front lines, cause economic damage, etc. Civilian targets are attacked in most wars. Is a car bomb outside a military barracks that different from one in a civilian area? What about a suicide bomb aimed at a sentry on a civilian road? I was about to drive past that one. Are the civilians who were killed collateral damage from an attack on a military target or victims of terrorism?
> people who burn police cars
So people who attack a civilian target. By your own definition they are terrorists if they are doing it for political motives.
Everyone wants this to be it but over and over we discover that the bigger a model is the better it is at all tasks, even ones far outside the domain it was optimized for. IE claude fable is better at writing both code and prose than smaller code- and prose-specific models.
The way vision and language models converge into the same geometric space should be extremely alarming for the "you don't need global knowledge for local tasks" type dreams.
And to be clear I'm not saying that smaller models don't or can't work well, or that we shouldn't be heading in this direction. And it's not quite the case that broad knowledge is strictly necessary. But it never seems to be negative! And so far it is the best way we've found to do... everything. Small models are good to the extent they are like big models, not to the extent that they are small.
On narrow domains, it is very common for small models to match or outperform larger ones at a fraction of the parameter count.
For example in language, this is called the “curse of multilinguality”. Small models that handle a single translation direction can easily outperform big ones that try to handle them all.
In any case, for most tasks the question is not “how many tasks can this model kind of do well” but “given time/cost constraints, what is the maximum level of quality we can achieve”. And for that, small models are usually very competitive.
Yes small models are and will be useful for lots of stuff for several reasons.
But the idea they’d be better than a bigger model is cope, you’re pretty much always better off running the biggest one you can bring to bear within your constraints.
This argument seems somewhat strange to me, and at odds with the recent model releases.
It feels like claiming that because a current generation laptop CPU is slower than a current generation HEDT CPU, we should ignore that the laptop CPU is faster and more efficient than the previous generation HEDT CPU.
Of course a huge system could do more, but these systems are constantly evolving and just in this month we've seen comparatively smaller models approach capabilities that were frontier AI less than a year ago. What you're saying would be true when we eventually find the limits of how much models can compress knowledge. Until then, all I think you're saying is that companies that are willing to burn more money can afford to ignore efficiency.
Fable 5 is actually a lousy writer. Opus 4.6 is the best for writing and prose assessment. Gemini 3 is smarter at reading comprehension but tends to be more unstable in judgement.
Yes. Speaking for the US, if your personal goals include having things like PTO and health insurance you're going to want the degree. Most white collar jobs are off limits without a degree in something, even if entirely unrelated to the job itself.
Not that there aren't other paths, or non-degree-gatekept jobs with those things. But they are harder and fewer and becoming increasingly so.
Programming is one of the few prestigious, highly paid jobs left that doesn't have a pretty hard req on having a degree to get into. That's why I do it. But yes I've also run into all the things you mention.
> if your personal goals include having things like PTO and health insurance you're going to want the degree.
Only ~40% of Americans have completed some level of post-secondary schooling, and only ~30% have a college degree. ~60-70% of jobs do not offer PTO and health insurance? That seems a little hard to believe. The BLS indicates that 79% of jobs offer PTO.
Sure, but that PTO is pretty... not great outside of salaried careers. For instance, for 5 years of work at Starbucks, you can get a maximum of ~2 weeks paid vacation across those 5 years in California (which has specifically friendly legislation to encourage this).
You ever hear of the scottsboro boys? Emmett Till's murderers were also acquitted via jury nullification, it was common for lynchings in the jim crow era. It's a tool.
Sprawl is where he originally started exploring these ideas but Jackpot is where he really focused deep into them and it's even more prescient. It's also bleaker, he has clearly recalibrated his estimate of how much power and control these new entities will have and what they will choose to do with it. Unfortunately I think it's a better representation of the path we're on.
