the_mitsuhiko

It's for using lsp to do things like symbol renames. It's cheaper to call an LSP server to do those kind of simple refactors than to spend tokens on the model doing it.

Is OpenCode pushing the LLM into doing these refactors via LSP? I found that making these refactors with ast-grep is quite efficient.

It can also use LSP to build context without having to grep around a ton. You don't have to grep and hten pull lines and make guesses for a function. You just ask the LSP for it.

I guess I was not particularly successful with having the agent actually do that. I think what fff is attempting to do (with fuzzy expansion of slightly inaccurate greps) results in better results to me, but even that is debatable.

To be fair, we do not have evals for this today, but we're generally looking into token efficiency and how likely models are at using tools the right way, and we did not find a ton of evidence of LSP helping, even for finding data.

But I would love to see some sessions from people where they have success with LSP data for either refactoring or looking up, because this would be super useful to better understand out blind spots.

How you deal with LSP? One of the best things in OpenCode is LSP integration.

I would love to learn how LSP helps you. I did various basic evals with LSP and I did not find it to improve the agent performance at all. Maybe on newer models this gets better because they are now RLing with LSP in the context, but at least in the past having a commit hook that runs lints and typecheck at the end, is more token efficient than having every edit be injected extra LSP results. You are almost guaranteed to be wasting tokens since partial edits are unlikely going to result in type checks passing.

(Disclaimer: I work on Pi)

We are going to address this. Not by loading the agent but by finding a way to provide official plugins or blessed plugins. But we’re not yet sure what the right approach is.

Suggestion for Pi: capitalize tool names for the Sonnet/Opus models (edit -> Edit, bash -> Bash, ...).

Pi is doing that.

When you use a Codex subscription in an agent like Pi or OpenCode, all the requests and tool call execution go through a sandbox owned by Codex app server

That is not the case. There are some subtle differences between subscription and regular inference API, but not to the degree that behaviors change entirely. In Pi we're doing tests against both API and subscription API regularly to see how they behave.

That kills the KV cache unfortunately. In some models (Gemini) I also doesn’t work because there is a signature on the model messages.

The problem with hyper targeting harnesses to models is that you end up locking yourself quite quickly into special behaviors of models, and you make your sessions non transferrable. That can be an acceptable trade-off and I know people who do that.

Those things are still illegal for minors. I find that to be a rather weak argument.

Will money save you from heartache?

Probably yes. If you consider dating to be a numbers game, money moves things into your favor greatly and will give you a chance to move past.

From the pain of losing a loved one?

Money will help greatly, because for many people getting over loss involves therapy which costs money.

From being lonely?

Money for many people buys companionship.

From having no respect from your peers?

Given that people associate respect with money excessively, probably yes as well.

From losing your health to incurable cancer?

Your chances of surviving cancer greatly scale with money.

Sure, there are limits to it, but absolutely access to money is tremendously pushing things towards happiness. I think it would be disingenuous to pretend otherwise, at least in a capitalist society we all live in.

Money can very much buy happiness. Most of the things that make you unhappy can be remedied with money. How much money you need to accomplish that depends though.

If you’re arguing for abolishing the nation state or any concept the establishes right and obligations based on you being born there, then it would help your cause to not entangle it with racism.

Racism in one form or another is at the very core of the nation state. I disagree with the very notion that rights should be established with a razor that cuts by citizenship. It’s a blunt tool that might work to some degree but clearly the US shows that citizenship is an insufficient measurement to the positive contribution to GDP or entrepreneurship.

It would also be more interesting if you can explain how you imagine we’d even get to a place where there is only a single earthly jurisdiction with free movement etc. now THAT would be an interesting thought experiment.

I don’t know the path to there. That does not mean I can’t see that as a goal worth perusing and to see things they take humanity further from that goal as a regression.

it seems to imply that this was your red line and when you became aware that this administration might be racist and nationalist?

If you read my blog you should have seen plenty of content before to get an idea why my red lines are. I even have a separate blog on that entirely: dark.ronacher.eu. My line is not here.

This is standard practice for nation states. To call it racism is braindead and lazy.

Author here. Just because something is standard practice does not mean this should be a goal. I do not want to live in a world where the north star is dividing people based on their birthplace lottery. I might accept it as a temporary local maximum that countries are optimizing towards but that does not mean I have to subscribe to that as a general destination or even the most optimal one.

I did link this now to the entry in the science fiction dictionary.

But also if you go to the talk page you can see discussions that lead to the change of the description of that page from "slur" to "derogatory term". That discussion was, from what I can tell, about the fact that slurs are for human targets and not machines. A lot of things are derogatory, particularly when applied to tools, and we would not bat an eye in the discussion. Something makes (at least some) people feel different about clanker.

It does, and the use of "clanker" I've seen was generally meant as a slur. It's not a neutral word in any context I've seen it in.

I can tell you that within the agentic coding community plenty of people are using it as a way to refer to coding agents by their users. That's a pretty neutral use.

True, but I also think that "clanker" is fairly far towards one end of the spectrum, not really near the middle.

Until I saw the comments on the most recent hackernews discussion on it I would never have made that guess.

Divisive language just sounds ugly to me because it aims to divide.

The issue we have with LLMs is that people are starting to treat their conversations with these things as a substitute to human interactions. Division very much is the point because they are not humans. There is a question to me about what the approach is but not being precise about their nature does not seem wise to me.

Dogs are not inanimate objects and even legally they carry a lot of rights. None of this should apply to an LLM.

Well, no, not really. You can't use a word that sounds like a slur, and usually is meant to be a slur and complain that it's getting polluted.

What makes something into a slur is when people are perceiving it as such. So yes, it's the usage of the term that is what makes it that. Would clicker be a slur? Or slacker? Or hummer and dinger? Where is the line. I'm assuming the line emerges from what society does with a word.

Then don't! There are degrees between "the agent decided" and "clanker"

But seemingly that is only something that emerges from the discourse. You seem to draw a line somewhere that is different from where I would have drawn it. Impossible to figure out a priori.

And yet Pi has done a few things that were quite transformational. A lot of recent agentic libraries explicitly credit Pi for design ideas.

We’re so early in this technology phase, now is the time to tinker and explore. At one point that window will close.

I agree but can we stop pretending like machines are in the same group as humans? They are not.

The project has templates and that's one of the giveaways to see that a issue bypassed it. Take for instance this issue from 5 hours ago as an example: https://github.com/earendil-works/pi/issues/4970

It does not follow the template, it's made by a user who is also active in the openclaw repo and it's full of slop analysis.

I think that depends on how much weight you give to that word. You might attribute a lot of weight to it, but I don't. Language changes and its what we make of it. Maybe if a prevailing opinion arrives that people respond very negatively to that word, then I will have to rethink my choice of words, but right now I prefer attempting to establish it as a more neutral term for a machine.

My hunch is that this term is loaded not because it's seen as a slur, but because people are currently attempting to give LLMs a soul and they are developing a certain unease about some of the consequences that come with it. For some the LLM is developing from a soul-less tool to a something with a personality, something they bond with, that's where some of those lines are blurring. I just don't think this is healthy at all.

The reason racism is a problem is because humans subdivide humans. A machine is not human. Putting these things on the same line is in fact exactly the issue I have with this.

So no. I cannot stress enough how much I disagree with the idea that this is in any way related to the n-word precisely because this is addressed as non humans.

That we are seriously considering if these things should be in any ways be equated to humans to me is a problem. They are not. They cannot and must not be given rights or responsibilities. They are dumb tools and if we lose sight of that, we can go down really problematic paths.

I have even on the receiving side of agent psychosis for more than a year now (because people talk to the LLM and sometimes the LLM tells people to talk to me) and there is only darkness there.

Where to me this discussion completely goes in a questionable direction is the idea that there are slurs against inanimate objects and that you would put that into the same category as derogatory terms against humans.

The reason you don’t insult other humans should be obvious, but none of that translates to machines. From my perspective it’s a grave mistake that we’re even remotely entertaining the idea right now that LLMs are sentient.

From that perspective I really quite appreciate the term. That said, I can see that it really seems to trigger some folks. I’m not sure we have found the right language to refer to what we have.

The term was adopted in Star Wars and has been used in science fiction before and after it too. It probably sits somewhere on some scale but it has always been used in literature as a term for machines and from that perspective I find it appropriate.

A machine does not have feelings. If you call your power drill a piece of shit you are not overstepping any bounds either.

I know that there is a lot of AI psychosis going on, but I don't have to subscribe to that. I consider it very problematic that we have some industry leaders (eg: Anthropic with their model wellbeing) that are seriously throwing that type of thinking into the room.

Author here. This visceral reaction to the term fascinates me. You’re not the only person here that mentioned it and it makes me wonder where that is coming from.

We’re naming a machine here, not a human being. Even if it was a slur (I don’t consider it one) it would be directed at a piece of equipment.

I find it fascinating because I don’t think people would bat an eye if that term was used for a hammer or keyboard. Yet somehow it changes when applied to an LLM based machinery. Craftspeople often apply jargon to their tools, much of which is neutral to negative.

Yes. It's Berkeley Mono. I use that one, Commit Mono and Mono Lisa depending on how I feel :)

Sure. But Mythos or not, the developments since that post was written, legitimate 11 more vulnerabilities have been found by AI tools.

And yet [1]:

Not even half-way through this #curl release cycle we are already at 11 confirmed vulnerabilities - and there are three left in the queue to assess and new reports keep arriving at a pace of more than one/day.

11 CVEs announced in a single release is our record from 2016 after the first-ever security audit (by Cure 53).

This is the most intense period in #curl that I can remember ever been through.

[1] https://www.linkedin.com/feed/update/urn:li:activity:7463481...

Note the lack of an upper bound

Since uv needs a singular resolution that's entirely intentional. In npm you can install diverging resolutions for different parts of the tree but that is not an option with Python. I had to make the same decision in Rye and there is just no better solution here.

If an upper bound were to be supplied you would end up with trees that can no longer resolve in practice. Some package ecosystems in Python even went as far as publishing overrides for old packages that got published with assumed upper bounds that ended up wrong.

Don't forget that you cannot know today if your package is going to be compatible or incompatible with a not yet released package.