OpenClaw’s memory is unreliable, and you don’t know when it will break
blog.nishantsoni.comAs 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 that feels as intelligent as what these systems are able to do within their context windows.
I am als a touch skeptical that adjusting weights to learn context will do the trick without a transformer-like innovation in reinforcement learning.
Anyway, I‘ll keep tinkering…
I've used open claw (just for learning, I agree with the author it's not reliable enough to do anything useful) but also have a similar daily summary routine which is a basic gemini api call to a personal mcp server that has access to my email, calendar etc. The latter is so much more reliable. Open claw flows sometimes nail it, and then the next day fails miserably. It seems like we need a way to 'bank' the correct behaviours - like 'do it like you did it on Monday'. I feel that for any high percentage reliability, we will end up moving towards using LLMs as glue with as much of the actual work as possible being handed off to MCP or persisted routine code. The best use case for LLMs currently is writing code, because once it's written, tested and committed, it's useful for the long term. If we had to generate the same code on the fly for every run, there's no way it would ever work reliably. If we extrapolate that idea, I think it helps to see what we can and can't expect from AI.
This is interesting. I haven't used OpenClaw but I set up my own autonomous agent using Codex + ChatGPT Plus + systemd + normal UNIX email and user account infrastructure. And it's been working great! I'm very happy with it. It's been doing all kinds of tasks for me, effectively as an employee of my company.
I haven't seen any issues with memory so far. Using one long rolling context window, a diary and a markdown wiki folder seems sufficient to have it do stuff well. It's early days still and I might still encounter issues as I demand more, but I might just create a second or third bot and treat them as 'specialists' as I would with employees.
I did (using Claude Code) something that sounds very similar to this. It’s a bunch of bootstrapped Unix tools, systemd units, and some markdown files. Two comments:
- I suspect that in this moment, cobbling together your own simple version of a “claw-alike” is far more likely to be productive than a “real” claw. These are still pretty complex systems! And if you don’t have good mental models of what they’re doing under the hood and why, they’re very likely to fail in surprising, infuriating, or downright dangerous ways.
For example, I have implemented my own “sleep” context compaction process and while I’m certain there are objectively better implementations of it than mine… My one is legible to me and therefore I can predict with some accuracy how my productivity tamagotchi will behave day-to-day in a way that I could not if I wasn’t involved in creating it.
(Nb I expect this is a temporary state of affairs while the quality gap between homemade and “professional” just isn’t that big)
- I do use mine as a personal assistant, and I think there is a lot of potential value in this category for people like me with ADD-style brains. For whatever reason, explaining in some detail how a task should be done is often much easier for me than just doing the task (even if, objectively, there’s equal or higher effort required for the former). It therefore doesn’t do anything I _couldn’t_ do myself. But it does do stuff I _wouldn’t_ do on my own.
Right - I think email is a much better UI than Slack or WhatsApp or Discord for that reason. It forces you to write properly and explain what you want, instead of firing off a quick chat. Writing things down helps you think. And because coding harnesses like Codex are very good at interacting with their UNIX environments but are also kinda slow, email's higher latency expectations are a better fit for the underlying technology.
Any chance you might put this on GH? Sounds really interesting.
Maybe but it's so simple I'm not sure it's worth it. You can easily make your own!
What sort of tasks do you have it do for you?
Two categories: actual useful work for the company, and improving the bot's own infrastructure.
Useful work includes: bug triage, matching up external user bug reports on GitHub to the internal YouTrack, fixing easy looking bugs, working on a redesign of the website. I also want to extend it to handling the quarterly accounting, which is already largely automated with AI but I still need to run the scripts myself, preparing answers to support queries, and more work on bug fixing+features. It has access to the bug tracker, internal git and CI system as if it were an employee and uses all of those quite successfully.
Meta-work has so far included: making a console so I can watch what it's doing when it wakes up, regularly organizing its own notes and home directory, improving the wakeup rhythm, and packaging up its infrastructure to a repeatable install script so I can create more of them. I work with a charity in the UK whose owner has expressed interest in an OpenClaw but I warned him off because of all the horror stories. If this experiment continues to work out I might create some more agents for people like him.
I'm not sure it's super useful for individuals. I haven't felt any great need to treat it as a personal assistant yet. ChatGPT web UI works fine for most day to day stuff in my personal life. It's very much acting like an extra employee would at a software company, not a personal secretary or anything like that.
It sounds like our experience differs because you wanted something more controlled with access to your own personal information like email, etc, whereas I gave "Axiom" (it chose its own name) its own accounts and keep it strictly separated from mine. Also, so far I haven't given it many regular repeating tasks beyond a nightly wakeup to maintain its own home directory. I can imagine that for e.g. the accounting work we'd need to do some meta-work first on a calendar integration so it doesn't forget.
I’m doing this exact same thing in my solo saas company, except with Cursor’s Cloud Agents. I can kick them off from web, slack, linear, or on a scheduled basis, so I’m doing a lot of the same things as you. It’s just prompts on a cron, with access to some tools and skills, but super useful.
That unreliability was why I gave up on OpenClaw. I tried hard to give it very simple tasks but it had a high degree of failure. Heartbeats and RAG are lightyears away from where they need to be. I'm not sure if this can be overcome using an application layer right now, but I trust that many people are trying, and I'm eager to see what emerges in the next year. In the mean time I know that they're working very hard on continuous learning - real-time updates to weights and parametric knowledge. It could be that in a year or so, we can all have customised models.
That would be great if that comes to fruition. Investing in a model with weights updates would be like investing in employee training, rather than just giving the same unreliable employee more and more specific instructions.
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 LLM is just fundamentally not able to "move" these experiences to long term, like a human brain does (through sleep, among others).
Once we can teach a machine to experience something once, and remember it (preferably on a local model, because you wouldn't want a global memory to remember your information), we just cannot solve this problem.
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. Because right now, with backtracking and weight adjustments, I just can't see us getting there.
I think if we want to build on what we have, instead of compaction at the end of the context window, the LLM would have to 'sleep', i.e. adjust its weights, then wake up with the last bits of the old context window in the new one, and have a 'feel' for what it did before through the change in weights. I just sense it's not that simple to get there, because simply updating the weights based on a single context sample risks degrading the weights of the whole network.
I like the idea of using small local model (or several) for tackling this problem, like low rank adaptation, but with current tech, I still have to piece this together or the small local models will forget old memories.
Sleep would probably be a part of the equation for consolidating , but there's still the question of how exactly does the brain process the information during sleep in a way that it permanently consolidates the information.
It's not how an llm can work right now, it needs too much iterations & a much bigger dataset than what we can work with. A single time experiencing something and we can remember it. That's orders of magnitude more efficient than an LLM right now can achieve.
Couldn't fitting solve the problem? That's what companies do: take a model as a base and train it on the specific data long enough so that it prefers the new data. Overfitting may be a thing but for personal use, I may want to have it work as I expected, every time.
> 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.
Exactly. It's been around so long and we still don't know how to mimic it.
The way an llm learns is a very interesting way of doing it, but it sure isn't what the brain is doing.
But it's indisputable.. We can get enormous results with this technique. It's just probably not the way forward for faster learning to remediate the issue of context loss.
How well do LoRAs work for this using something like Thinking Machine's Tinker?
Why does a language model have to be monolithic? I think retraining a model is expensive (relatively speaking). Is there some way to bolt on specialization?
That's exactly the issue. Retraining is too expensive & needs too much iteration to work efficiently I think.
I agree. A key to human intelligence is our ability to adjust our weights in real-time. All knowledge becomes parametric knowledge - the knowledge stored inside the model. RAG is a messy workaround which requires making assumptions about what is needed to load from external sources before it is clear what is needed. Agentic loops can go some way to overcome this, but they are resource intensive, slow, prone to mistakes and deviations, and far less accurate. The secret sauce of an LLM is the vectorised weights. RAG is like putting a 1990s Honda Civic engine into a Ferrari. You can do it, but the result is quite terrible.
I think we will eventually end up with models which can be individually trained and customised on regular schedules. After that, real-time.
From my perspective there are some people that have never built real processes in their life that enjoy having some processes now. But agent processes are less reliable slower and less maintenable then a process that is well-defined and architectured and uses llm’s only where no other solution is sufficient. Classification, drafting, summarizing.
I’ve had a Whatsapp assistant since 2023, jailbraked as easy assistant. Only thing I kept using is transcription.
https://github.com/askrella/whatsapp-chatgpt was released 3 years ago and many have extended it for more capabilities and arguably its more performant than Openclaw as it can run in all your chat windows. But there’s still no use case.
It’s really classification and drafting.
This. So many junior engineers showing me AI flows that could just be a script with a few parameter inputs
Why write script when more tokens does job
yeah but unfortunately an AI flow can bring promotions, while scripts won't
I find that it's usually management that ask for such things "because AI".
I mean using AI is a great way to interpret a query, determine if a helper script already exists to satisfy it, if not invoke a subagent to write a new script.
Problem with your "script" approach is how does that satisfy unknown/general queries? What if for one run you want to modify the behavior of the script?
I like to experiment with AI flows to make iteration quicker, then once something work investing in is found, back up and build something that's actually repeatable.
Same thing could be said with SKILL.md yet they are highly useful...
Yes you can automate via scripting, but interacting with a process using natural language because every instance could be different and not solid enough to write a spec for, is really handy.
tl;dr: there's a place for "be liberal in what you receive and conservative in what you send", but only now have LLMs provided us with a viable way to make room for "be loosey goosey with your transput"
I understand but there still is usually 80-95% of the skill flow that you can script out that is repeated. Script it out and simplify your skill, make it more stable, and provide more opportunity to scale it up or down i.e use stronger or weaker models if need be. We should be scripting and forming process first then seeing where we can put AI after that.
The AI for everything thinking is really easy to let infect you. I was trying to figure out how to make some SQL alerting easier to understand quickly. The first thing my brain went to was "oh just shove it into an LLM to pull out the info of what the query is doing". And it unfortunately wasn't until after I said that out loud that I realized that was a stupid idea when you could just run a SQL parser over the query and pull the table names out that way. Far faster, more cost effective, and reliable than asking an LLM to do it.
That’s actually an awesome idea and totally helps to reduce wasting context size - move repeatable instructions to a SKILL.md, and once they’re repeatable and no longer have variability to input, turn it into a tool! Rinse repeat.
Oh nice, you could even eventually turn the whole process including inference into an app so that you’ve cut out the LLM from the whole process saving you execution time
You say that with the wisdom of experience.
But there's still value in people exploring new spaces they find interesting, even if they do not meet your personal definition of pareto-optimal.
Exploring with AI doesn’t lead to the same level of learning. They are doing the equivalent of paying to skip the level up of their character and going to the final boss with level 1 armor
I look at it more like speedrunning a level. You're skipping the parts of the level that take up the most time, some times using hacks. Is it universally as much fun as playing the game? No, just like using AI to prototype might get you to the same place, but without the experience of discovery and blockers along the way.
Fully agree with your comment regarding real processes. Being a Six Sigma Black Belt, studying processes and reducing the errors is critical. The Openclaw processes at the moment scare me.
I'm only at Four Alpha Brown Belt still but once I pass my test I think I'll also understand why this is critical. I can't wait to get scared.
Make a skill.md called six sigma black belt audit refactor and publish it
Vibe process automation
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 through it's context and doesn't know to compact. Then it's just stuck. I can't wait till it matures or something more reliable comes along.
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 directly in the JS when they bother me enough.
If OpenClaw is the future of software I'm honestly a bit scared for the industry.
I'm open to suggestions, I tried Zeroclaw and Nullclaw but they're bad in their own way. I would like something that's easy to run on Kubernetes with WhatsApp integration and most important, stable releases.
> If OpenClaw is the future of software I'm honestly a bit scared for the industry.
I think it's mainly the industry wannabes gathering around a "sexy" brand name again, when they're really more interested in "AI as personal assistants".
OpenClaw just has the most traction despite being a hot mess, because the people hyping it up don't know how bad the codebase is, or because they want to launch something first and switch it over to a more credible alternative after.
Hermes Agent
I know Twitter has been talking up Hermes Agent by Nous Research a lot or id recommend building your own agent off of Pi.
Thanks for this suggestion, I installed it yesterday after seeing this comment and this surely is a breath of fresh air! It appears that everything is designed reasonably well from the ground up. It’s more limited, but what’s there works well.
> easy to run on Kubernetes
that sounds like an oxymoron.
Yeah I’m amazed how much constantly breaks between every release, I’m dreading to update it. I’m using it as a kind of personal assistant that keeps track of all my outstanding work. Morning and evening briefing, can get an overview of what I did per week / month, it’s pretty solid for keeping track of all this.
But it’s so terribly unstable, it’s as if nobody actually tests things before pushing a release. I don’t need 2 updates per day, I just need one every few weeks that’s stable.
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...
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 talking about that's actually productive, it's doing probabilistically when deterministic tools already exist and have for in some cases over 20 years.
You don't need an LLM to put a meeting on a calendar, that's literally two taps with your phone or a single click in gmail. Most email services already have suggestions already built in. Emails have been summarized for 10 years at this point. If you're so busy you need this stuff automated, you probably have an assistant, or you're important enough that actually using general intelligence is critical to being successful at all.
The idea of getting an LLM email response sounds great for someone who has never worked a job in their life.
This comment section is full of llm writen responses too, to the point where its absurd. Noticing how most of them just talk in circles like "But I think many people criticizing the various Claws are missing out on the cronjob aspect. There's value in having your AI do work automatically while you're asleep. You don't even need OpenClaw for that, just a cronjob that runs claude -p in the early morning. If you give your AI enough context about yourself, you get to a point where it just independently works on things for you, and comes to you with suggestions. It doesn't need to be specifically prompted. The environment of data it can access is its own context, its own prompt. With that, it can sometimes be surprising and spooky what you wake up to, without being directly prompted."
This literally isn't even saying anything. This paragraph does not mean anything. It's not saying what its doing, whats happening or what the result is, just "something is happening".
No, you didn't save time using openclaw, you just changed to managing openclaw instead of doing your actual job.
You don't need custom scripts for most things if its actually something that matters, most tools already exist, and if you do openclaw isn't going to help you do it.
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?
> 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?
Some people have developed personal coping strategies for neurodivergence, and can probably do better with some AI assistance.
I myself tend to live by my calendar. Even then, I may eventually forget to follow up on something until it's too late, because I'm busy or overwhelmed with other things.
It would’ve happened eventually anyway, but OpenClaw is basically what kickstarted the beginning of the end of token subsidies. It’s a almost begging to be used wastefully. And agents would miss and lose nothing without it. It’s devoid of a reason to exist.
I don't follow the thinking here. If you are using it for coding maybe, but the main use case of openclaw is as a personal assistant. I'm using a $10 a month minimax subscription for it, and I've never used more than 10% usage of a 5 hour window.
Serious question, and not meant to be snarky, but what is it doing , in terms of a personal assistant, that couldnt be done with some inbox cleanup, a couple of phone shortcuts, and siri/alexa/whatever google has?
Mine is not connected to any of that. It is only connected to my Obsidian vault. Daily habits are reminded and checked off. It asks about calories etc. Think more of a manager or accountability partner than 'email on steroids.' I use it to track personal projects primarily. Lots of other tools could accomplish task tracking, but this one is more proactive than those. If you read my other comments today you will see I am not an openclaw-maxi.
Siri/Alexa/gemini are severely limited in what they can do/access because they wanna protect the brand.
I don't think this is fair (and I say this as someone who doesn't see much a point of OpenClaw). To me it's very obvious that Claude Code itself is the beginning of the end of token subsidies.
When Claude Code was released, there was a community leaderboard where people competed who could waste the most tokens. Let that sinks.
I know people, especially people who write code, like to blame "the other clueless people" for ruining their cheap token plan. But we're not stuck in the traffic. We're the traffic.
Oh, of course we're all burning tokens like crazy. It's just that some are actually producing something of value with them, and some... are running OpenClaw.
I hope not. We need a lot of new data centers. Let the tokens flow like water.
OpenClaw is not just "memory unreliable." The whole OpenClaw is unreliable. Imagine Claude Mythos being released publicly — finding OpenClaw's vulnerabilities would be child's play for him.
I actually quite enjoy the OpenClaw. Although the recent CC crackdown has caused me to try different LLM providers which aren't that reliable but anyways, here are few things I do with it, in all separate groups.
* Telegram Health Group, created an agent to help me track sleep, recommend my supplements based on my location, remind me in the morning and evening to monitor my food. I send it images of what I eat and it keeps track of it. * Telegram Career Group, I randomly ask it to find certain kind of job posts based on my criteria. Not scheduled, only when I like to. * Telegram Coder Group, gave it access to my github account. It pulls, runs tests and merges dependabot PRs in the mornings. Tells me if there are any issues. I also ask it to look into certain bugs and open PRs while I'm on the road. * Telegran News Group, I gave it a list of youtube videos and asked it to send me news every day at 10am similar to the videos.
So far, it's a super easy assistant taking multiple personas. But it's getting a bit painful without CC subscription
Are you concerned about telegram admins having access to that information?
Yeah! And what if Samsung put a bug in the silicon that gives them access to all your stuff without you ever knowing!
Or the CIA has set up inside your closet with a listening device!
You can just have your bot use Telegram topics for this.
I built a special belief-based system recently for my own agent harnesses instead of some similarity based fact storage stuff... which falls flat once conflicting data points enter the system and just increase LLM confusion and make it do weird things. this means learning over time works a bit more like humans do - superseding old beliefs and reconciliating stuff cleanly over time. Also including the building blocks to have a subagent managing it autonomously (with tools/skills/soul). works quite well and very fast given its pure nodejs+sqlite and doesn't eat tokens like crazy or needs any thirdparty embeddings solution. maybe have a look.
> 0 legitimate use cases
My teams currently using it for:
- SDR research and drafting
- Proposal generation
- Staging ops work
- Landing page generation
- Building the company processes into an internal CRM
- Daily reporting
- Time checks
- Yesterday I put together proposal from a previous proposal and meeting notes, (40k worth)
> In every case, when you dig deeper, the story is one of two things: either what they built could already be done with standard AI tools (ChatGPT, Claude, any decent LLM with a simple integration), or it’s aspirational
All your use cases are fairly well handled by conventional LLM's. OpenClaw is a security nightmare, so it's probably worth switching away.
Most of these things I could’ve handled with pen and paper, but that’s missing the point.
None of those things require openclaw. You could accomplish them with something like Google Drive and Claude Code CLI.
> 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 that without OpenClaw" and go ahead and do it within 10 minutes.
Not once.
OpenClaw is flawed, but the convenience is an order of magnitude higher than anything else.
> 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
> You offered nothing to support this.
You've got it inverted. My point is the people saying "You could have done that just as easily with ..." are the ones not supporting it. The commenter has already built that thing with OpenClaw. If someone is saying it could easily have done without it - well, demonstrate it!
It appears that you are confusing who has the burden of proof here. It is the one making the claim contrary to the status quo.
Hint: the status quo is not that openclaw is a tech that is magnitudes better than using LLMs without it.
Listing a bunch of things that are just normal LLM things as reasons why openclaw is great is not making that case.
Burden of proof is on the one making the claim. Status quo has nothing to do with it.
You should revisit the burden of proof then. Status quo is most certainly an important part.
Regardless, their claim was "OpenClaw is flawed, but the convenience is an order of magnitude higher than anything else."
And they attempted to shift the burden when I asked for substantiation.
Status quo influences how good your proof has to be (extraordinary claims require extraordinary evidence) but not who needs to bring it.
Think more deeply about this. https://lastedenblog.wordpress.com/2016/12/13/professor-jame...
What a bizarre article. The morality of recreational torture is not a matter of factual correctness. Burden of proof is not a concept that makes any sense when there’s a disagreement over morality. You can make arguments for your position and those arguments may involve factual claims which can be proven or disproven, but the underlying morality can’t.
And then it ends with that sudden left turn into denouncing atheists as inherently irrational and evil. WTF?
Congratulations on bringing an argument so terrible that I’m actually more convinced of my position after having read it than I was before.
It appears that analytic philosophy may not be for you.
If it thinks I’m evil then clearly they don’t want me.
> Anthropic already has a tool for it Yes, but Anthropic built this tool after OpenClaw, because of OpenClaw.
The difference is I would have to do that myself. It has access to gdrive and cc and does it for me when I send it a message in chat. Sometimes when I’m out I even just send it voicys.
Method 1: run Claude Code in YOLO mode and use natural language instructions to get something done
Method 2: send natural language instructions to OpenClaw to use Claude Code to do the same thing
Sorry my tiny brain says to me method 2 is doing the same thing with extra steps
I can have multiple conversations on multiple topics always accessible via different discord channels, all with a shared memory, without that memory being held in a continually degrading context window.
One channel - reminders for medications, and recording my dosage. Another - "research this fancy new tech thing for me". Another - "let's continue work on that side project we started last week". And then in another - "create a dashboard of my meds dosage using that fancy new tech thing we were talking about yesterday". And of course finally "any urgent emails this morning?".
All without finding, creating, or setting up multiple apps or scripts for each individual task. If I have another idea, I just tell it what I want it to do, or ask it how we can make it happen.
You're contradicting yourself here. Are you controlling it yourself or not? lol
It’s task dependent.
None of those require Claude Code CLI either, you could develop their workflows with a script (bash, python) and any quality LLM.
Just an example of how I would accomplish them. The obsession with openclaw is generally misguided. The 'magic' is the LLM. I'm running an OC instance on a server in my home, I have experience here.
What happens if it makes a mistake? How would you know?
10/80/10
10% done by an assistant that’s been trained on the task (or a dev or me)
80% heavy lifting done by claw
10% review and corrections
I don’t get people’s hate. Let others enjoy it.
I think people are just tired of the fire hose of posts that have been showing up since it came out. It’s so annoying. Why does everyone need to pimp it so hard? It’s like your aunt trying to push Herbalife on you every time you see her.
That, and some people hate being on the receiving end of the output. The old "if you didn't bother writing it, I'm not going to bother reading it".
somehow I've been able to do that for 40+ years using my brain, eyes, fingers, vi , CLIs and shell scripts. no unsolved problems there.
I did too :D
Nothing of what my agents do, we didn’t previously do. But now I can get moderate to good results with a lot less effort. Allowing the business to expand whilst keeping costs controlled.
I was able to buy stuff from home without the Internet as well.
I’ve had some success with claude cli agents at some scale with a memory architecture - but it roughly reads like a massive index, where it crawls through a trail of breadcrumbs to piece together all the info it needs to do a task. It’s fairly tedious to maintain, and it’s always a battle maintaining reasonable context size and token spend.
I’d say it’s like 85% reliable on any given task, and since I supervise it, this is good enough for me. But for something to be useful autonomously, that number needs to be several 9’s to be useful at all, and we’re no world near that yet.
I’m currently watching someone trying and failing to roll openclaw out at scale in an org and they believe in it so much it’s very difficult to convince them even with glaring evidence staring them in the face that it will not work
I've had a crack at this problem in Agent Kanban for VS Code (https://github.com/appsoftwareltd/vscode-agent-kanban). The core idea is that you converse with the agent in a markdown task file in a plan, todo, implement flow, and that I have found works really well for long running complex tasks, and I use this tool every day. But after a while, the agent just forgets to converse in the task file. The only way to get it to (mostly) reliably converse in the task file is to reference the task file and instructions in AGENTS.md. There is support for git work trees and skipping commits of the agents file so as not to pollute the file with the specific task info. There is also an option for working without work trees, but in this flow I had to add chat participant "refresh" commands to help the agent keep it's instructions fresh in context. It's a problem that I believe will slowly get better as better agents appear, and get cheaper to use, because general LLM capability is the key differentiator at the moment.
I partly identify with the article. While I don't use OpenClaw itself, I hacked together my own small Claude-in-a-loop/cronjob, and it seems we're all getting our morning briefings and personalized morning podcasts now.
The other common use case seems to be kicking off an automated Claude session from an email / voicetext / text / Telegram, and getting replies back. I'm emailing Claude throughout the day now, and sometimes it's useful to just forward an email to Claude and ask it to handle the task within it for me.
But I think many people criticizing the various Claws are missing out on the cronjob aspect. There's value in having your AI do work automatically while you're asleep. You don't even need OpenClaw for that, just a cronjob that runs claude -p in the early morning. If you give your AI enough context about yourself, you get to a point where it just independently works on things for you, and comes to you with suggestions. It doesn't need to be specifically prompted. The environment of data it can access is its own context, its own prompt. With that, it can sometimes be surprising and spooky what you wake up to, without being directly prompted.
Give it enough context, long term memory, and ability to explore all of that, and useful stuff emerges.
Just a heads up if you're using claude -p with the sub your account might be in trouble, they're cracking down on this kind of stuff lately.
Enforcement seems to be a combination of string matching for 3rd party sysprompts, heavy usage, and some random factor.
Not clear if there are any hard rules you can stay on the good side of, the only way to be safe seems to be to pay per token. (There goes the ~90% discount!)
--
Also yeah you get ~80% of Claw by shoving Claude Code in a Telegram Bot ;) It's already a general purpose computer use thing, people forget! (And it's a lot better at extending itself than the actual claws, lol)
I think the least illegal and also least bad option is to just use ngrok and tmux tho
If you look at my comment history, you'll see what seems to be someone defending OpenClaw (even though I stopped using it).
I have some issues with the article, but I agree with some of the conclusions: It's great tinkering with it if you have time to spare, but not worth using weeks of your time trying to get a perfect setup. It's just not that reliable to use up so much of your time.
I will say, it's still amongst the best tools to do a variety of tasks. Yes, each one of those could be done with just a coding agent, but I found it's less effort to get OpenClaw to do it than you writing something for each use case.
Very honest question: One of the use cases I had with OpenClaw that I'm missing now that I don't use it: I could tell it (via Telegram) to add something to my TODO list at home while I'm in the office. It would call a custom API I had set up that adds items to my TODO list.
How can I replicate this without the hassle of setting up OpenClaw? How would you do it?
(My TODO list is strictly on a home PC - no syncing with phone - by design).
(BTW, the reason I stopped using OpenClaw is boring: My QEMU SW stopped working and I haven't had time to debug).
> I could tell it (via Telegram) to add something to my TODO list at home while I'm in the office. It would call a custom API I had set up that adds items to my TODO list.
All the existing, commodity todo list apps on the market can't address your use cases?
At least I can't tell there is anything you can't do on your personal phone.
> All the existing, commodity todo list apps on the market can't address your use cases?
Nope. I've custom honed my TODO system since 2009. I'm not switching for some one else's app.
And I don't use phones.
> Very honest question: One of the use cases I had with OpenClaw that I'm missing now that I don't use it: I could tell it (via Telegram) to add something to my TODO list at home while I'm in the office. It would call a custom API I had set up that adds items to my TODO list. How can I replicate this without the hassle of setting up OpenClaw? How would you do it?
What you are looking for is an orchestration platform such as n8n or windmill.dev. You can still have a telegram bot and still use LLM for natural language interaction, but it's much more controlled than OpenClaw. I do exactly what you describe, add todos to my todoist account from telegram.
> I could tell it (via Telegram) to add something to my TODO list at home while I'm in the office. It would call a custom API I had set up that adds items to my TODO list. How can I replicate this without the hassle of setting up OpenClaw?
The general idea is make a simple deterministic program that runs on your PC at home in a never ending loop. Every minute or so, check Telegram for a new message. If a message is received, then the program runs "claude -p" with a prompt, whatever MCP tools or CLI permissions it needs, and the contents of your Telegram message. Just leave the program running on your home computer while you're out, and you're done.
I don't use Telegram, so coding the part to check Telegram would be the hard part. I use email instead, and have the program check every minute for new mail (I leave my email program running and check the local inbox file). I'd already coded up a local MCP server to manage my ToDo list (Toodledo) so Claude just calls the MCP tools to add the task.
For things like TODO, I guess an email will suffice.
However, it was really nice being able to use Telegram and get quick validation. I also had a flow set up where I could send a voice memo. It would take the audio file (ogg), run Whisper, and then pass through an LLM for cleanup, and follow the instructions in my message. Really handy to use while I'm walking around.
I guess I want to create my own OpenClaw like agent, but not with its crazy broad access: Just limited to the functionality I allow, and with the convenience of using Telegram. I don't care about memory, soul, etc.
I did the exact same voice memo thing too, except I had Claude make an Android app to record the file and send it to Whisper. In the end I had the app just email the transcription & trigger Claude that way (ie receiving the email triggers my PC to wake up Claude), rather than sending Claude the audio file directly.
My reverse audio reply loop is convoluted - I have Claude generate its TTS file from Whisper/Mistral, and upload them to a server with an RSS file it updates, so I can play them in my podcast app (AntennaPod), then send me a notification via Pushover that the reply is waiting. I ended up building out an MCP tool for that workflow, so Claude really just calls the MCP tool with the text of what it wants to say, everything else is a deterministic program doing the work.
Memory is really useful to have - it can just be a bucket of searchable Markdown files. It's also useful to have a "reminders to self" Markdown file that Claude reads each time, and that Claude can update. I don't continue the same context window, and that "reminders to self" plus the ability to read previous emails in the conversation seems to be enough to keep the context going for me.
You'll feel better if you know exactly how your Claw is locked down. Mine doesn't have the open email access others are granting, not at all. Claude gets a bit grumpy about that and keeps begging for more access :)
>telegram bot that runs claude -p
https://github.com/a-n-d-a-i/ULTRON
It also supports Codex :)
I felt pretty clever until (1) I found a repo where they used this trick to create a full OpenAI compatible API endpoint[0] (lmao, the VC money distortion field spawning truly comical Rube Goldberg machines), and (2) they started banning "unauthorized" usage of the Claude sub, which trend unfortunately seems to be accelerating recently as their lower value consumers have grown in both number and usage.
I think shoving claude -p in your bash script / cronjob / messaging app bot of choice counts as "unathorized 3rd party harness", but your guess is as good as mine...
(claude -p with per-token billing (i.e. paying 7x more) is allowed though, of course)
-- There's also an Agents SDK (formerly Claude Code SDK?) which is basically just claude -p but with more typing, as far as I could tell.
[0] https://github.com/router-for-me/CLIProxyAPI
[0b] Honorable mention https://github.com/kronael/claude-serve
Basically all modern models can do that, in _any_ conversation. Write a skill that teaches it how to use your API and where the endpoints are. Local models will need a harness (look up pi), but any frontier model can do this out of the box.
Writing a script to make a POST request is something assistants have been able to do for quite a while now.
And if you have a Claude subscription, you can use Dispatch to directly write to your PC's drive, no API needed.
I'd use Obsidian with the sync. Or you can vibe code a telegram bot that calls that API for you in like 50 lines or something.
You can do anything if you believe!
Re: QEMU: For the sandboxing I realized what I actually wanted was "it can't read/nuke my files", so I made a non-privileged linux user and added myself to its group. So I can read/write its files, but not the reverse.
> It would call a custom API I had set up that adds items to my TODO list
You can use anything to call this API right? I have multiple iPhone shortcut that does this. Heck, I think you can even use Siri to trigger the shortcut and make it a voice command (a bit unsure, it’s been a while since I played with voice)
> You can use anything to call this API right?
The API is on my home PC and not exposed to the outside world. Only OpenClaw via Telegram was. So my question is about the infrastructure:
How do I communicate with something at home (it could be the API directly) using a messaging app like Telegram? I definitely want an LLM in the mix. I want to casually tell it what my TODO is, and have it:
- Craft it into a concise TODO headline
- Craft a detailed summary
- Call the API with the above two.
I'm not asking in the abstract. What specific tools/technologies should I use?
Expose the API to the outside world using tailscale. Run your telegram bot on n8n or windmill.dev. You can absolutely use an LLM, both n8n and windmill.dev support AI agentic workflows. google "n8n LLM telegram bot" and you'll find tons of examples.
Tailscale on your PC and phone. Free (as in beer). I run vikunja.io in docker at home and simply web browse to it over the tailnet.
Just leave the bot running on your home PC, a Telegram bot can be programmed in any language of your choice that can communicate via a network.
If you aren't a programmer it's also the kind of small project that LLMs are great at, there are many examples ingested in their training data.
i would just enterit in the todo app on my phone.
MS todo app, or any number of others. Added benefit of not needing telegram
I've been using my TODO program since 2009. It's heavily customized for my needs. I'm not going to change it.
I do feel like the memory the biggest hurdle I’ve been encountering and I’m curious what solutions people have been doing to make it work.
What seems to be somewhat working for me
1. Karpathy wiki approach
2. some prompting around telling the llm what to store and not.
But it still feels brittle. I don’t think it’s just a retrieval problem. In fact I feel like the retrieval is relatively easy.
It’s the write part, getting the agent to know what it should be memorizing, and how to store it.
As a side note, one thing that seemed to really help was having the agent recursively summarize all the files into a repo map with the same folder/file structure as the repo.
And forcing to always orient itself with that repo map first seemed to really help it from tunnel visioning.
its almost as if its rules should be persisted, versioned and tested in a deterministic way. I also wonder if it might end up useful to try enforcing some kind of constraints to data (memory) itself. like if we could enforce atomicity, consistency, isolation AND durability. and transparency so no place for malware to hide. and deterministic execution, from fully reproducible builds...
er, nevermind. prob just crazy castles in the sky wistful dreams :-)
I primarily "only" use it as a run-manager that can spin up another agent in a tmux which I can then join by ssh on my cell phone. Then, I can monitor the work from my cell phone and choose to either directly interact with the tmux pane or else just message my openclaw agent to do it for me. That right there is the only "killer" app I've found for it. I do also use it to post to my x.com account and that's also pretty useful. Neither of these uses assume any super long context over time will be retained. But, to me, the run-manager use case is pretty great.
I built Atmita (atmita.com) from scratch, not based on OpenClaw. Memory is distributed across agents and automations, each with their own layer, and they interleave intelligently so agents only load what's relevant instead of dumping everything into one context. Cloud-native, no self-hosting required.
"Who's in charge here?"
"The Claw."
Some of this stuff is starting to look like technologies that worked, looked promising, but were at best marginally useful, such as magnetohydrodynamic generators, tokamaks, E-beam lithography, and Ovonics.
Who is this guy and why is he casually admitting to reading all the user conversations???
To clarify - we don't read user conversations - what I'm referring to is deploy volume (how many times OpenClaw gets spun up on our infra) and direct conversations I've had with people in my network who deployed OpenClaw independently
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 outside of that cron / daily briefing or some family schedules.
Just curious, what is the delta with this and Claude Code / Codex CLI for the SWE work?
It is an interesting take. I think this is mainly early adoption pains though. This stuff is moving so fast that if you say 'it isn't useful because X isn't good enough' then just wait a month and X will be good enough to find Y as the blocker (or no blockers are left and it truly does become useful). Soon we will see this hooked into the home assistant world well combined with local and remote compute and then we are likely to see real movement.
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 over time I think.
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.
Openclaw is unreliable. I had it running for a few months. It uses up a lot of resources and doesn't provide any benefit other than being able to chat with it via other methods than tui.
I've removed it.
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.
How much money are people here spending on tokens for this thing?
Memory systems as most people understand and build them are a clear dead end. We just need skills, tools and better context management.
Good to know that I'm not alone. I now use it for music recommendations (not so great) and keeping track of restaurants I want to try (really good at this but so are a lot of other apps).
>This isn’t a bug that gets fixed in the next release. It’s a fundamental constraint of how OpenClaw manages context.
Last I checked, it doesn't!
I'm still trying to figure out what to use it for other than news aggregation...
I understand it's the quickest way to expose all your API keys.
I'm not sure what these people who have strong opinions like this think Openclaw is, but to me, it's a product with 1) a somewhat easy to setup prompt passing wrapper that can span many channels like Telegram, Whatsapp etc. 2) A (at least optimistically) plug-n-play, configurable architecture to wake up to events (cron entries, webhooks etc.) and fire up agents in order to get 'proactive' behavior, with the flexibility to integrate models from a gazillion providers. Pretty much everything else it's bundled with is general purpose tooling that does or could easily exist in any other agentic tool.
It's a rather simple framework around an LLM, which actually was a brilliant idea for the world that didn't have it. It also came with its own wow effect, ("My agent messaged me!") so I consider some of the hype as justified.
But that's pretty much it. If you can imagine use cases that might involve emailing an LLM agent and get responses that share context with other channels and resources of yours, or having the ability to configure scheduled/event-based agent runs, you could get some use out of having an Openclaw setup somewhere.
I find the people who push insanity like "It came alive and started making money for me" and the people who label it utterly, completely useless (because it has the same shortcomings as every other LLM-based product) like Mr. "I've Seen Things. Here's the Clickbait" here, rather similar. It's actually hard to believe they know what they're talking about or that they believe what they're writing.
The twist? This article and marketing campaign for it are 100% by OpenClaw.
Could we stop with the clickbaiting headlines?
You forgot to add, “Here’s why.” /s
I know that headlines are all about eyeballs, but this is seriously just exhausting. Headlines are advertisements and advertisements are about getting engagement. Surely having your audience just getting angry at them isn’t a good thing, right?
Sounds like an armchair expert
Right back at ya ;)
The author makes some good conclusions; I’m as AI-pilled as the next hopefully-not-soon-to-be-ex-software-engineer, and I struggled to find use cases for my Claw that couldn’t be served with a cronjob and $harness.
If your findings contradict that, we are all ears - genuinely.
Oh not really, but the article just reads like it’s from an observing putting of view, without tangibles.
The killer thing was remote control, but that’s here in Claude now. In my opinion claw has to reason to exist anymore.
I tried it, didn’t like it. It gave me the ick with the communication channel.
No wireless. Less space than a Nomad. Lame.
Sure, anything it does can be done better with specialized tooling. If you know that tooling.
The memory thing sounds like an implementation limit rather than something fundamentally unsolvable. Just experiment with different ways of organizing state until something works?
That is very similar to human memory.
Author basically admitting to having a boring outlook on life IMO. Sure maybe not for work but there's tons of things that suck the life out of your limited time, having a tool not just OpenClaw is one way to not bend to the will of BigCo for whatever thing you want to do, need a 3D model? Need something summarized or need control of something which the manufacturer forbids? All can be done without spending entire weekends for.
You probably don't know how to setup memory.
The killer usecase is letting you make whatever you want, instead of being at the mercy of what your OS/platform dictates.
Your idea of a killer idea is a whatsapp summarizer lol.
I can already make whatever I want
It's for people who can't, did drop some neurons on this thread?
Why aren't databases the solution to many memory problems? Maybe this is a naive question:
For example, for the invitations in the OP: Have Openclaw write incoming rsvps to a database, probably a flat file here, and use the db as persistent memory: OpenClaw can compose outgoing update emails based on the database. Don't even suggest to OpenClaws that it try to remember the rsvps - its job is just writing to and reading from a database, and composing emails based on the latter. ?
Does that violate the experiment, by using some tool in addition to OpenClaw?
Mostly because it can’t or shouldn’t remember everything - eventually the context window gets too big. But how does it know what to forget and what’s important and should be remembered?
IMHO, the biggest problem with OpenClaw and other AI agents is that the use-cases are still being discovered. We have deployed several hundred of these to customers and I think this challenge comes from the fact that AI agents are largely perceived as workflow automation tools so when it comes to business process they are seen as a replacement for more established frameworks.
They can automate but they are not reliable. I think of them as work and process augmentation tools but this is not how most customers think in my experience.
However, here are a several legit use-case that we use internally which I can freely discuss.
There is an experimental single-server dev infrastructure we are working on that is slightly flaky. We deployed a lightweight agent in go (single 6MB binary) that connects to our customer-facing API (we have our own agentic platform) where the real agent is sitting and can be reconfigured. The agent monitors the server for various health issues. These could be anything from stalled VMs, unexpected errors etc. It is firecracker VMs that we use in very particular way and we don't know yet the scope of the system. When such situations are detected the agent automatically corrects the problems. It keeps of log what it did in a reusable space (resource type that we have) under a folder called learnings. We use these files to correct the core issues when we have the type to work on the code.
We have an AI agent called Studio Bot. It exists in Slack. It wakes up multiple times during the day. It analyses our current marketing efforts and if it finds something useful, it creates the graphics and posts to be sent out to several of our social media channels. A member of staff reviews these suggestions. Most of the time they need to follow up with subsequent request to change things and finally push the changes to buffer. I also use the agent to generate branded cover images for linkedin, x and reddit articles in various aspect ratios. It is a very useful tool that produces graphics with our brand colours and aesthetics but it is not perfect.
We have a customer support agent that monitors how well we handle support request in zendesk. It does not automatically engage with customers. What it does is to supervise the backlog of support tickets and chase the team when we fall behind, which happens.
We have quite a few more scattered in various places. Some of them are even public.
In my mind, the trick is to think of AI agents as augmentation tools. In other words, instead of asking how can I take myself out of the equation, the better question is how can I improve the situation. Sometimes just providing more contextually relevant information is more than enough. Sometimes, you need a simple helper that own a certain part of the business.
I hope this helps.
Can you be more specific about these? For instance can you map "analyzes our current marketing efforts" to precisely what is being fed into the context window. What prompt is used to "find something useful"?
Like many here, I am struggling to see a meaningful delta between OC and CC but fully willing to accept that my skepticism is misplaced. Basically, I am in "trying to care about OC" mode right now.
That’s not the issue. It’s an issue of trust and whether the thing does exactly what it should continuously over long periods of time without having to keep an eye on it.
Until it gets there, it’ll remain a fringe product.
I was getting a lot of use case out of it mainly interacting with the file system.
The problem is if not carefully designed it will burn through tokens like crazy.
It can integrate apis for you on the fly. That’s one of the biggest usecases IMO. Combine that with skills, cron, and sub-agents, and you get a lot of power there.
there are zero legitmate use cases? sure bro. you can say that to my claw which is making me more money than my salary
Making Money, the ultimate "legitimate use case".
Please elaborate
Almost certainly some kind of scam
From what i have read openclaw seems to be the ideal scam/clickfarm bot or am I completely mistaken?
Bounties?
Sure, bro.
[flagged]
Please don't do this here.
I am sorry. I understand that many of us are heavily invested financially or emotionally into AI but the key insight about OpenClaw remains the same that it really isn't much useful beyond maybe a daily news summary. Nothing that cannot be done otherwise.
It’s a lame trope that people only comment favorably about things because they have a financial incentive or some other pathological reason to do so, and for that reason it’s against the guidelines.
I think it broadly speaks to the issue of llm’s in general.
People have to actively interact with these things to get outputs of them. Why can’t they do it passively?
Because on the surface LLM’s seem intelligent. But they’re not. They need constant attention to ensure they are doing what one wants and not going off track. I experienced this myself many times - the variance in the output can be jarring at times. From surprisingly good to absolutely disappointing.
Why hasn’t customer service over the web/telephone become completely autonomous or at least partially via llm’s? We have all this talk about agentic programming and yet much of the world carries on as it was.
Interesting. I don’t think this variance problem is going away. It’s a different kind of variance than humans are used to when dealing with other humans.
Then intellectual honesty demands that we present these limitations as such but capitalist grifters have their own principles it seems.