solarkraft

In popular media: https://youtube.com/watch?v=45sJJpS1OEM&t=3m28s

Did you just say “his”?

I wouldn't read too much into it. I've gotten this popup for a model I hadn't had access to yet before and it resulted in what you describe.

I’m so tired by the AI labs’ 100 agent based products. I understand that they’re still figuring out form factors, but does everything have to be incompatible with everything else?

Why shouldn’t one be able to worry about 2 bad things at the same time?

What bothers you the most about Codex? Looking from the outside, it looks like a pretty capable harness with a pretty good client.

They’ve been accepting this for years and so have users

That’s what I like about it. It signals a kind of ambition open source projects usually lack.

And this doesn’t even have to mean having to idly wait 20-30 minutes at a fast charger: Lots of shopping and work places have chargers now.

I’m not claiming this to be an entirely solved issue, but the situations in which EVs make sense are already very broad and keep expanding. The city I live in has lantern chargers (admittedly not everywhere), some others are building them into curbs. Apartment parking spaces and garages are getting chargers installed. If you live rurally, chances are high you can install your own.

Sure they are, but do you think they are continuing training to improve a model after release without bumping the version number, presumably only to game the benchmarks?

It's a whole meme category to make the AI overview help with absurd scenarios: https://knowyourmeme.com/memes/where-is-mama-ai-overviews

Is anybody tracking these quality changes? All I've seen so far are accusations (quite a few at this point) but not really any actual data.

> but it does add value

There is SO MUCH value to be added on top of archives. Browsing archive.org kind of sucks, but there is so much amazing material on there.

Curation makes all the difference.

You can do cool experiments with them. But nothing a typical consumer is likely to care about.

Same plan as for a power outage: Go outside (hacker news), chat with your neighbors and wait.

The models tend to not be rewarded for not doing that.

Or Kimi K3? /s

Some people are resource constrained.

This part I find a little disappointing. The plan is the perfect boundary to cut context which still isn’t free, even if models can handle a lot of it now.

But that’s just the value of planning, not having it be a special mode.

I really wonder what I’m missing because my general rule to plan first unless explicitly instructed works perfectly fine with all models I use.

“I want to ...” / “Let’s ...” -> Plan

“Do X” -> Actually Act.

But then again I also have it configured to only ever answer questions instead of inferring them to be instructions (which I’ve seen others do differently).

I agree that couldn’t is a tad too far, you could always have tested more. But even with more testing you could’ve missed this, completely irrespective of the use of AI.

Awesome! I believe NixOS is perfect for systems you just want to work and track reproducibly, like appliances (which a lot of server software is supposed to be). I’m glad people smarter than me are actually able to realize its advantages =)

In the same space, also check out openDesk initiated by the German Government (https://www.opendesk.eu/en) and La Suite by the French one (https://lasuite.numerique.gouv.fr/).

AFAIK, there has been some collaboration with the German government on La Suite, not sure about the others.

Then maybe the need for your site is gone. That’s a good thing in the sense that the problem is solved. If your goal is “do something in the cocktail space” (which is also roughly the companies’ goal I would say), indeed you should do what the AI can’t, which I’d say is be human touch. An AI can’t give your opinion or view.

Here’s an easy proxy: Did you feel the need to build it because nothing to your liking existed? Did it take some effort to build? Do others feel the need to use your thing for those reasons?

Then it’s sufficiently non-obvious. I believe there’s still a pretty big field of things that fall into this category. Making proper things is a lot more than writing some code.

That’s exactly what I do. OpenAI + Opencode Go subs, 0 interest in Claude.

Async, as I understand it, means that the agent never has to actively wait for things to complete, so it can do work while it waits and can receives the result at some later point which it can then react to. I don’t know whether Eve has that feature, it would be cool. Should not meaningfully influence its ability to be deployed serverlessly.

Totally agree, Qwen 3.8 is a pain to wait for. Glimmer much less so!

It’s a library and there are usage examples.

> but clearly more documentation is better than lack of

I tend to disagree. This makes it very high signal. So far my agent has been successful at navigating the repo.

Right! It’s not really rocket science. I’ve been thinking of building a general “wait” command that just blocks until some event occurs. Still more “active” waiting, but as far as I understand the efficiency problem comes from polling.