krzyk

Because people are crazy, usage of npm for installing binaries is quite common unfortunately.

The most important feature of Pi is that it is small, and has small system prompt, making it great for locall LLMs.

Why would anyone compare a cloud LLMs power usage when one doesn't pay for it? Local power consumption is important for those.

codex is in rust and not in power and memory hungry js/ts.

e.g. opencode right now uses ~80% of my CPU.

At first I also thought that it would be just call and wait, but a lot of work is done locally (any tool calls).

> When Google wanted engineers to use AI features, it turned them on in Cider-V by default. And if you turned them off, later updates would turn them back on.

How is that good? That would be the first thing that would force me to use IDE that is not controller by 1984 entities.

p4 makes me wake up at night screaming.

Similar to that IBM/Rational ClearCase, both are so unfriendly compared to subversion/cvs or git/mercurial that I always struggle to believe why someone would torture themselves using that. Probably admins love them because they allow some tooling to be added.

Oh, yeah - libraries yes. But for application code it is useless.

This sounds like an ad for US than anything else.

Does any of the US companies earn money on LLMs? No, they bleed money. Github Copilot is switching to token based pricing, which will be costlier than hiring juniors.

Anthropic also is switching enterprises to token based pricing from their subscription one.

From the big three only Codex is still in somekind of subscription pricing, but they'll shift eventually (usage limits are a kind of that, but they have them less stricter than Claude ones)

There is one winner in this race - China. Trump with his agendas and wars makes it even more likely that China will lead this new market.

> I would go for a Mac Air or Neo, but only if I could install ChromeOS.

Similarly, but I would extend that to mac mini/studio, but I would like Linux on it. I like hardware, but I hate the OS there.

With AI it is important to catch errors/hallucinations early, static typing helps with that.

So languages with dynamic typing might hide some errors until runtime, static typing one could catch that during compilation.

With dynamic ones you need way more tests to cover some of the scenarios that compiler does for others.

And there is significant amount of code written "for ages" in languages that were there longer, like C, C++, Java (yes, I know that python is quite old, older than Java - 1991).

Could you elaborate what does "compiling orchestration prompt" mean?

commons, is something that is eventually being migrated into the main, at least those that are decided to be required for most projects. I don't use apache commons or guava at all in java (now at 25 or 26, depending on project) - there are still some libs that depend on those, but I would argue that most use it out of inertia, than actual need.

As for slf4j, I still don't see any justification for an abstraction layer on top of logging. I never, ever migrated from one logger to another, and even if I did need to do it - it is very easy as most loggers are very similar. E.g. that's why I decided to use log4j2 in my latest project.

Rather people unable to setup a static websites where needed.

I hate that I can't do a curl, or automate my curls to retrieve data from the web because I either see some cloudfrare protection or some captcha.

Information is blocked in walled gardens.

Model capabilities are rising slower compared to model pricings. Recent price increases made hiring juniors cheaper and in the short run, not to mention in the long run.

It would be best if distros kept tap on kernel changes and update as soon as possible when they see a security issue fixed.

Sending emails to some big distros would still result with e.g. Gentoo not getting that info because they are not a big distro.

There are so many distributions that it is not possible to notify each one, unless there is some single distribution list for all.

And if you disclose to just a handful, why ignore the rest?

You can look at the code in editor or IDE even when CLI agent is doing work.

I do that when I want to, but for me using agents in IDE is like looking with one eye covered.

Yeah, people learned.

I created a 4 subagents that polled for new tasks, and restart after ~5h.

It was a great run.

They do for any new plan. Those multipliers are only for people that paid annually. After their subscription ends they'll go into token based pricing like the rest of people.

Those multiplier are only for grandfathered Pro an Pro+ plans that had annual billing, basically a way to scare people of out of those plans. Ant new ones (and bussiness+enterprise plans) will be on token based billing since June 1.

Is that autocomplete better than IntelliJ own plus their local only LLM completion?

I uninstalled copilot plugin because it was eating memory and its completions where about 60% good and the rest was bad.

After switching back to IntelliJ I see just positives.

Fair, I don't use LFS so I don't know how it works.

I was writing more in regards of hooks as "checks" for workflow/lints/etc.

Local hooks are just a convenience. CI checks are assurances, you have to have them.

If one hates the round-trip he/she will adopt hooks quickly.

It is still that person creation. Not sure about American law, but AFAIR in my country you can't remove the author from creative work (like source code), you can move the financial beneficiary of that code, but that's it.

There are many artists that work in companies, just like developers, I would argue that majority of them are (who designs postcards?)

I use Haiku frequently, and for my codebase it is working fine.

But I'm not vibecoding, I don't let models do large work or refactorings, this is just for some small boring tasks I don't want to do.

Which Opus models were free on Copilot?

Well they charge per prompt, but with usage limits it is a mix of token and prompt. If prompt multiplier is higher, tokens are also multiplied, so limit is reached sooner.

It is basically a token based pricing, but you get alos a limitation of prompts (you can't just randomly ask questions to models, you have to optimize to make them do the most work for e.g. hour(s) without you replying - or ask them to use the question tool).

Using local models or external?

If external, aren't they pricey? How much tokens they generate?

If local, what runs on such hardware that gives reasonable results?

There are people running OpenClaw, so yeah, crazy as it sounds, some do that.

I'm reluctant to run any model without at least a docker.