OpenClaw, but in containers: Meet NanoClaw

Ideally, you shouldn't have to defend yourself against your own AI agent. But we don't live in an ideal world and an unrestrained agent can cause a ton of damage. OpenClaw, an open source agent platform, made that obvious as its popularity surged over the past month and security incidents followed.  Consider the case of …

[Neil Barnes]

Re: yet smaller

I'd settle for a sanityclaw, but everybody knows sanityclaws don't exist!

[Claude Yeller]

Re: can we get ... NoClaw ?

"can we get a microclaw or Infinitesimal-claw or maybe best of all NoClaw ?"

Sorry, but things never get uninvented. It seems several empires have tried to ban gun powder, but that didn't work either. Inventions are here to stay.

If you cannot beat them, join them.

Treat AI code generators as really "smart" compilers. You describe your code design in a markdown document and let the AI generate and test the code.

Then put your experiences in your resume and ask your employer for a raise. We are in a hype cycle, aren't we?

[zimzam]

Re: can we get ... NoClaw ?

Things may not get uninvented, but poorly conceived inventions regularly get consigned to the dustbin of history.

[Claude Yeller]

Re: The dustbin of history

We can always hope, or, hope dies last.

[adrian727]

Re: can we get ... NoClaw ?

The closest you can get is NullClaw

https://nullclaw.org/

[jake]

Re: can we get ... NoClaw ?

"The closest you can get is NullClaw"

Trust me, one can have no claw.

In fact, if you have any brains at all, figuring out how to have never claw is fairly easy. As is never AI.

_{And you don't have to go to a web page to justify your lack of action, either.}

[abend0c4]

Re: can we get ... NoClaw ?

the dustbin of history

As containers go, it's more secure than most.

[elsergiovolador]

Re: can we get ... NoClaw ?

Teleportation and cure for cancer has been uninvented at least twice.

[Claude Yeller]

Re: uninvented at least twice.

"Teleportation and cure for cancer has been uninvented at least twice."

I assume they were moved to an alternative reality.

[elsergiovolador]

Re: uninvented at least twice.

Yes, that's were they live now.

[Claude Yeller]

Re: uninvented at least twice.

Not here?

Is cancer a disease that can be cured by DNA teleportation?

I guess, the answer is no.

[elsergiovolador]

Re: uninvented at least twice.

Often pan-brain is crossing dimensions. These are the results.

[a_foley]

Rant

Security-conscious "AI" agent? Is this some kind of joke?

I try to avoid LLM things, and certainly projectsabominations suffixed with "Claw". Like even for folks who don't care about security, they're completely useless; unless they're too lazy to even just walk to the microwave and heat frozen pizza. Unfortunately, that is probably the case, humans are lazier than lazy.

And on the topic of LLM things to avoid, also includes projects which utilise vibe-coding, like these Claw-things. Why? (you already know it) It shows that the maintainers are one or more of the following: not capable of maintaining the project; lacks interest in the project; unserious; don't care about stability nor security. The worst part is that projects which do aren't transparent about it, so one has to check for telltale signs...

Anyway, I apologise for the rant, just tired of all the "AI" crap I get force-fed every day.

[vZeroG]

Re: Rant

Based.

[jake]

What's with all the new cults springing up all of a sudden?

See SubjectTitle.

[Irongut]

Re: What's with all the new cults springing up all of a sudden?

I think you spelled cunts wrong.

[Ropewash]

Cohen and his brother Lazer working on 'future tech'..

So Jewish space lasers it is.

[Lee D]

Claw

It's ChatGPT with its output piped directly into a sudo terminal.

It's the most ridiculous thing I've ever heard of.

You can't fix that with a container.

[Anonymous Coward]

Oh my god NOBODY CARES.

It doesn't work. It has no use case. Nobody intelligent is using it.

It doesn't make money.

It doesn't make anything easier.

STOP RUNNING THIS CODSWALLOP.

[JavaJester]

LLMs aren't the only things incapable of learning new things

After all the horror stories of Open Claw misbehaving, why are people still running it? A container will make the situation worse because it provides a sense of false security. Your computer might have some protection, but it can still wreak havoc on the network. For example, a container would not have helped with a rogue Open Claw AI posting a hit piece on a git maintainer because its pull request was rejected.

[Glen 1]

Re: LLMs aren't the only things incapable of learning new things

Microsoft still has a business, so it's not like people ever really learn