Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

OpenClaw’s memory is unreliable, and you don’t know when it will break

blog.nishantsoni.com

Post reply on HN
Print view View original

151–160 of 196 posts

Re: OpenClaw’s memory is unreliable, and you don’t know when it will break

#151

Post by airstrike »

ambewas wrote: post #124
loehnsberg wrote: post #122

As long as there's no solution to the long-term memory problem, we will have a "country of geniuses in a data center" that are all suffering from anterograde amnesia (movie: Memento), which requires human hand-holding. I have experimented with a lot of hacks, like hierarchies of indexed md files, semantic DBs, embeddings, dynamic context retrieval, but none of this is really a comprehensive solution to get something…

You're right to be skeptical. Without a way to actually implement how the human brain processes experiences into a consolidated memory, we won't be able to solve the long term memory problem at all. Not with the current technology. An LLM context is a pretty well extended short term memory, and the trained network is a very nice comprehensive long term memory, but due to the way we currently train these networks, an…

> I think this is probably the most interesting field of research right now. Actually understanding in depth how the brain learns, and figuring out a way to build a model that implements this.

This field of research has been around for decades, so who's to say when there'll be a breakthrough.

In fact, LLMs are great despite our very limited understanding, and not because we had some breakthrough about the human brain.

Re: OpenClaw’s memory is unreliable, and you don’t know when it will break

#152

Post by nonameiguess »

Rekindle8090 wrote: post #68

My biggest issue with OpenClaw is everyone talks about doing things with it but doesn't explain what it actually is doing. First of all is not an LLM, you're beholden to an api or local llm limitations. Second of all it's always calendars, email replies, summarizing. You do not need an LLM for that, and an LLM doesn't make it easier either. It sounds like executive cosplay, not productivity. Everything I see people t…

I'm gonna die shaking my fists at clouds pretty soon, but amen. I had a guy in work Slack evangelizing about this in January when it first came out and he's talking about automatically ordering groceries and getting reminders to do laundry and shit and I just don't get it. Sure, that's a use case, but I figured out how to feed, clothe, and house myself without digital assistance 25 years ago. How'd you live this long if you can't eat without a machine telling you what to buy and how to cook?

I can't even tell if these replies are in fact just astroturfed bot armies flooding us with marketing or there really is an entire generation of people out there right now who can't do anything unless their phone is telling them what to do.

And where are the outcomes? Okay, you've got OpenClaw telling you every few hours how many calories you've had so far today. Have you gotten leaner? Faster? Stronger? Healthier or fitter by any quantifiable objective metric at all? Or are you just doing exactly what you did before but now your phone is scripting it for you?

Re: OpenClaw’s memory is unreliable, and you don’t know when it will break

#153

Post by osigurdson »

sailfast wrote: post #79

I dunno - my boss has deployed a couple of claw agents that are pretty good at doing SWE and SRE work. They’re available for the whole company to use, and they save us a ton of time. Pretty decent use case! Personally I haven’t found claw agents replace anything really for personal use outside of commercial tools I’d pay for to handle scheduling and stuff, but I also haven’t tried / trusted too many new use cases out…

Just curious, what is the delta with this and Claude Code / Codex CLI for the SWE work?

Re: OpenClaw’s memory is unreliable, and you don’t know when it will break

#154

Post by osigurdson »

jmward01 wrote: post #53
bryanlarsen wrote: post #45

Earlier quoted context omitted.

Conventional LLM's are moving fast too. The argument is that OpenClaw isn't any more useful than conventional LLM's, and I suspect it will always be true because the conventional LLM's will gain any useful capabilities.

I think openclaw provides a unique feature of a standardized host environment for a persistent assistant. This is different than the chat interfaces that are presented by anthropic/openai/others that give you a 'while you are here' assistant interface and is very different from the idea of trained llm weights and ways of serving them up like llama.cpp and others. There really is something unique here that will evolve…

Agree. It is like just having one continuous chat session with ChapGPT forever. Of course they do have memory already (at least ChatGPT does). I ended up turning it off however, because it kept bringing irrelevant stuff into the current convo.

Re: OpenClaw’s memory is unreliable, and you don’t know when it will break

#155

Post by osigurdson »

theturtletalks wrote: post #37

The hype around OpenClaw is a bit confusing but I think I figured it out. For most coders, Claude Code in the terminal was an important event. Letting it access code and change files directly. For normal users, they didn’t see the power is that. OpenClaw runs Pi in a terminal and exposes the chat thru Telegram or any chatting app. This gave the ah-ha moment to non-coders that coders had had for 6+ months prior.

I'm confused about it as well. I've installed OC locally and also on a VM. I don't get it so far. But then again, I'm not willing to give it all of my passwords which is probably why I'm not seeing much value. It isn't just non-coders that see OC as a game changer however.

Re: OpenClaw’s memory is unreliable, and you don’t know when it will break

#156

Post by windexh8er »

sixhobbits wrote: post #107
operatingthetan wrote: post #20

I'm using openclaw as a personal development bot, which is pretty useful. It pings me throughout the day using crons to complete tasks and follows up on them. But aside from that, it is a very unreliable piece of software. I'm constantly having to fix it, or track down correct configurations. It can just decide to randomly edit it's own config, uses incorrect json keys and then the whole thing is dead. Or it blows th…

I had the same experience, now I just have a simple claude code session running with the new channel feature so it can take input and give output via a telegram bot. It's not as fun with SOUL.md etc but so far much less janky.

You realize that SOUL.md is nothing more than prompt injection, right? It's not a magical configuration file that gives an LLM a "soul". It's just anthropomorphizing a part of the prompt. It's also an expanded burn on tokens and, potentially, your money.

But if you think you need an agent framework to use a prompt you're going to love this one simple trick...

Re: OpenClaw’s memory is unreliable, and you don’t know when it will break

#157

Post by azmz »

LaurensBER wrote: post #26
operatingthetan wrote: post #20

I'm using openclaw as a personal development bot, which is pretty useful. It pings me throughout the day using crons to complete tasks and follows up on them. But aside from that, it is a very unreliable piece of software. I'm constantly having to fix it, or track down correct configurations. It can just decide to randomly edit it's own config, uses incorrect json keys and then the whole thing is dead. Or it blows th…

Amen. I love the concept but I've never hosted such a terrible piece of software. Every update breaks something new or introduces another "anti-feature" that's enabled by default. The documentation is often lagging behind and the changelog has such a low signal to noise ratio that you need a LLM to figure out what upgrading will break this time. For now I've just given up on updates and I've been patching bugs direct…

[flagged]

Re: OpenClaw’s memory is unreliable, and you don’t know when it will break

#159

Post by stitched2gethr »

operatingthetan wrote: post #59
BeetleB wrote: post #52

Earlier quoted context omitted.

> None of those things require openclaw. OpenClaw was never meant to be a tool that could do things you couldn't do without it. Also, whenever someone points out you could accomplish something without it, he underestimates the effort needed. In the examples I'm thinking of, someone simply asked OpenClaw to do something, had a few back and forths with it, and it was done. I have yet to see someone say "Oh, I can do th…

> the convenience is an order of magnitude higher than anything else. You offered nothing to support this. My openclaw is realistically just an agent in discord versus the CLI. That's not an "order of magnitude" more convenient. Anthropic already has a tool for it https://code.claude.com/docs/en/remote-control

> Anthropic already has a tool for it Yes, but Anthropic built this tool after OpenClaw, because of OpenClaw.
Post reply on HN