Hacker News Reader

Comment by CrimsonCape | Hacker News Reader

If you use OpenClaw now can you tell us how it's going?

I used it initially. It was great and useful but the api costs were pretty high..

Now, I use claude code directly on my machine with the remote control flag.. so its like open claw exactly. I can be outside and give commands to it via mobile app

caffeinate -is -t 7200 claude --remote-control

then Control + Command + Q ftw

(keep system from sleeping if plugged in, keep system from going idle even if not plugged in, for 2 hours, start claude code w/ remote control enabled, lock my mac w/ the keyboard shutcut)

The Amphetamine app from the App Store keeps a MacBook awake even with the lid closed.

You need an app for that?

I know right? On Linux for example GNOME needs an extension and only KDE seems to have a handy toggle available from the get go. It seems the normal is to NOT make a feature like this easily available.

I guess `caffeinate` doesn't go that far because it can easily lead to overheating.

I've been using it since about February, still using it. It's running on my old M1 MBP. I have a lot of automations on OpenClaw, things like "Every week summarize this data and send me a message about it" or "check when a new version of an NPM package comes out and submit a PR to update it."

It's quite expensive from a value per token perspective. It's quite cheap from a "send one message from my phone and it does something probably productive" perspective. No major security breaches yet.

All of the use cases I see seem like stuff that’s better performed by normal software. Dependabot will watch for new versions and submit a PR for free.

AI for developing software that runs predictably seems more useful than AI performing actions like a program would.

yeah i think the deterministic part (sending email, submit PR) is usually done by a normale script/software, agents are useful for summarizing text in a smart way, expecially if you need to aggregate text from multiple sources and it's natural language (not structured data)

This still feels like made up work. We are using AI to inflate a prompt in to a full email to send, and then the receiver is using AI to summarize it back down.

That definitely happens, but also many people are also using this stuff to get some amazing work done.

I think AI is an incredible competence magnifier. If someone knows what they are doing that competence can get magnified massively. If they do not know what they are doing, that gets magnified just as much. The tricky part is getting people with a low level but real degree of competence, which is all of us when we're starting out, to learn how to reliably magnify that.

in that particular scenario, is certanly a made up work. but sometimes you are aggregating text from lets say 5/10 different inputs, and summarizing those into a single email with AI is pretty niche and efficient. obviously not everything needs an hammer to be solved

I used it to try it out. Security wise I don't trust it at all so I had it running in a VM and local models.

It was able to create an application that Claude created, using the same instructions file. Considerably slower but I was surprised it could do it.

Reply on news.ycombinator.com