juped

A youtube competitor is a CDN problem, i.e., business deals

In this particular case it means the Nx ecosystem, which is a solid Numpy alternative.

They're fairly large (larger than lowercase letters in the header text, comparable to uppercase). Glad you were able to fix it since you prefer them much larger.

Another angry all-caps rant in an agents file (cf. "NEVER FUCKING GUESS"). As the operator of this tool which you used to delete your production database, you should at least know that angry all-caps ranting pushes the big textual probability engine into the space of things associated with abusive ranting.

I wouldn't do it this exact way either but the benefit is "having any local throwaway integration branch" vs. having none at all. You don't need to do it this exact way to have one.

The "A" in "AGI" doesn't stand for "Apocalypse", you know.

It made some sense as a goalpost when the frontier of "AI" was "a computer plays, specifically, Go really well", now that typical ones are quite general it's just a floating signifier people should probably stop using for anything.

It's interesting to see the strange workflows that come from jujutsu users, as someone who works on git workflows.

There's some counterproductive stuff in there from my perspective but at its core you're keeping up a throwaway integration branch, which is helpful practice if you'll ever care about an integration. It's annoying with git because the interface for updating your throwaway integration branch is very clunky and easy to get wrong.

I think you should revisit the word "just", its presence in the comment you're trying to discuss, and how it's used.

While this is a perfectly fine policy in the space of possible policies (it's probably what I'd pick, for what it's worth) the arguments being given for it leave a bad taste in my mouth.

You can pry my em dash—short for "Emily's dash", after the poet—from my cold dead hands.

Just like with NFTs, this is all going to discredit the actually sensible use cases for years.

Neither! A logistic curve is just an exponential with a carrying capacity - it is still an exponential! There is no reason to believe that AI capability, which grows logarithmically with the handwaved-resources used on it (roughly, this is compute and training data), grows, has grown, or is growing exponentially!

I know this sounds like "the moderate position" to people but you are accepting that something logarithmic is somehow in fact exponential (these are inverse functions of one another) based on no evidence or argument.

Here is Sam Altman, the one man in the world with the most incentive to overstate AI capability, accepting the extremely-well-known logarithmic growth: https://blog.samaltman.com/three-observations

What we see in reality is a basically-linear growth pattern due to pushing exponentially more resources into this logarithm.

Yeah, what you're going to get is more efficient proofs: you can do induction on one case to get results about elementary functions. Not sure where anyone's getting computational efficiency thoughts from this.

I'm productizing the unique Git workflow work I do - this may be the last week I'm ever available for consulting engagements, so head to https://merge-base.com now if you're interested!

Pijul is the future in some ways, but not others - it's a darcs, and I'm afraid darcsen will always be niche.

Jujutsu steals it, though! You should probably give it a look.

Why would you put a second, jankier API in front of your API when you could just use the API?

No.

The cases where bisect fails you are, basically, ones where it lands on a merge that does too much - you now have to manually disentangle the side that did too much to find out exactly what interaction caused the regression. But this is on the rarer side because it's rare for an interaction to be what caused the regression, it's more common that it's a change - which will be in a non-merge commit.

The squash merge workflow means every single commit is a merge that does too much. Bisect can't find anything useful for you by bisection anymore, so you have to get lucky about how much the merge did, unenriched by any of the history that you deleted.

Somewhat Linux-like. You could probably improve it purely from a git perspective by letting subtask dependencies be many-to-many (the commit graph is a dependency graph), but what you have is probably best for your whole Jira workflow.

I agree.

I set merge.ff = false and alias ff to merge --ff-only. I don't use pull but I do have pull.ff = only set, just in case someday I do.

The graph log and the first-parent log serve different purposes and possibly shouldn't be the same command conceptually; this varies by user preference but the first-parent log is more of a "good default", generally. Merges do say "Merge" at the start, after all.

This is what I advise people to do in consulting engagements, too, it's not one of my personal quirks.

It's like switching to jujutsu gives people some kind of mental permission to do this - jujutsu justifies its existence just by that alone, really.

Even even worse, angry all-caps shouting will make it more stupid, because it pushes you into a significantly stupider vector subspace full of angry all-caps shouting. The only thing that can possibly save you then is if you land in the even tinier Film Crit Hulk sub-subspace.

I touch on this a bit in the piece I wrote for normies, it helped a lot of people I know understand the tech a bit better.

No, you can do it on all the major providers for either no or low cost.

Many people use it like this - this is playing to its strengths, rather than trying to work around its weaknesses. "What's the idiomatic X language way to do Y?" gets you a solid, useful answer in seconds.

But it's just a damn good tool, not the apocalypse/the thing that lets you finally fire everyone. So it kind of gets lost in the hype.

>Cornell, for example, had a limited capacity to pay software developers to maintain and upgrade the site, which still has a very no-frills look and feel.

arXiv is doomed. It was nice while it lasted.

Tools eat up so much context space, too. By contrast the shell tool is trained into Claude.

It's the best starting point in many ways!

Now this is a definite human comment, shoehorned in xkcd links are a huge human tell.

It's very unclear to me why AI companies are so focused on using LLMs for things they struggle with rather than what they're actually good at; are they really just all Singularitarians?

I'm baffled that people, unknown to me, have apparently been considering Claude Code, the program, some kind of "secret sauce". It's a tool harness. Claude could one-shot write it for you, lol.

No, the Japanese imperial calendar oddness is that it's otherwise exactly the Gregorian calendar, but the era may change midyear.