Kinrany

The write side on the client can't be transparent, can it?

Otherwise either write persistence is best-effort, writes don't work offline at all, or there is some kind of custom conflict resolution layer that guarantees that all writes will be accepted, like CRDTs

The Hurl file format seems good, any new FOSS replacements should probably build on top

Disagree because it's one bit of information and in practice you can often guess the correct direction even if you're not sure if "not" was or wasn't said.

Yes, they basically asked for LARP. It'd be more interesting to give them a shared goal while also asking to survive.

It should probably be X$ per alarm regardless of it being false or not, minus first N per year.

And of course the whole point of tools is being superhuman, compared to humans without those tools. E.g. compilers are better than humans at converting C into assembly. "Just a tool" doesn't mean much.

> rather than a super human intelligence

That's a strawman (or at least a weakman): only a very small number of people (perhaps overrepresented online) thinks current agents are superhuman at programming in practice.

Which is of course not the same group as those who expect new models to become superhuman in months/years.

Open source projects themselves have reputation, regardless of how they are governed specifically.

Simpson's paradox?

If it was true, then companies would offer less and less money until it stopped being true

It's a good default for a pre-1.0 tool that aims to type check existing code.

It's never going to be embeddable, no?

Is there a project that does all of these together?

- embeddable like sqlite

- clustered like dqlite

- has a standalone node like rqlite

- can be backed up to S3 like litesteeam

- optionally AP like cr-sqlite alongside CP

The same goes for electricity. The utility is extremely useful, but it's a utility and is priced as such.

This will happen once the models are good enough, though it will take some kind of cartel to prevent the public from doing the same or distilling the models.

Because selling access to models will not be a good business. But they can form a cartel and keep frontier models to themselves. This will also prevent distillations.

I'll settle for git operations and PR merges not breaking.

Thanks!

Disappointed that the language doesn't do anything linguistically interesting. I thought it would infer keywords from their first use, or something like that.

What's basic learning theory?

Can it be optimized by doing it forward again, for all values at once?

I don't think the distinction the article makes is between handcrafted and mass-produced. "Software factories" is just the name we gave to these slop machines recently, but it's been the same with teams that churned out features without thinking much beyond the requirements. When it comes to physical items, factories can be set up with care and handcrafted items can be churned out without any.

> Even just an issue tracker like Jira would be an insane boost of productivity.

Would it though? The problem with many issue trackers is that unlike many other tools, they do not encapsulate the right way of doing things, and so it's easy to add paperwork without making the work more straightforward.

And if there is someone who can set up Jira and teach people to use it well, they don't necessarily need Jira.

I'm curious if there is a comparison to Luau.

That sounds illegal

How are the regulatory hurdles supposed to help against foreign open models though? China won't cooperate with that.

This exactly. The three interesting parameters are total time to answer, total cost, and output quality (for any given purpose).

Sure, but that's normal for moonshots. Maybe LLMs really will stall and spend 10 years getting better at anything and everything except taste, and then there'll be a completely new approach that has nothing to do with massive GPU farms. Maybe not.

Being paid per word doesn't cover all of the incentives. One needs to remain employed, which is the same incentive that salaried employees have.

Once the models are good enough, the companies can stop selling access to models and start directly competing with the rest of the economy.