Why would you go to a pub to talk to people if you have all those agents at home. Especially if one spent money on mac minis in bulk I wouldn't have money to spend on beer at the pub, just get cheap 6-pack from grocery store.
Maybe bar fights enjoyers? Well with bots I just unplug it and I win :)
I a way yes, it is their duty to their future happiness to explore. No doubt some will not word it like that, but I met some that really do (and I envy them, I was indeed a bit mind/clue-less).
It was first, but it's like Langchain built tools first. Nowadays, everyone's got a better assistant. Grok's looks like the most exciting, though I just have a custom built one. Has a headed browser when it wants and everything that I can pop into from my phone. It's so easy to write this kind of software it's a waste to pay someone for it.
Captchas are handed to me to solve. Because the operations are in the background I usually make the agent be a little more human with typing and navigating. The slowness isn't a concern for backgrounded jobs. It is imperfect, of course, I triggered something on Grainger. And the truth is I should be using Luminati, but it's good enough to be useful to me.
This was useful for the 1-2 months where models were good enough to use all their newfound tools but not good enough to vibe code them themselves.
Now, just install codex (or claude code or whatever) on a vps, tell it to make a way to email it, then email it to make a way to text it etc etc. takes a few minutes and when you need another bridge just have codex make it.
I'm fairly sure you can vibecode all these things, but given the state of agent produced code at the moment, I would suspect that your email and sms gate will have horrible security holes.
Automated scripts that target everything they can. Isn’t that one of the core lessons to take away from the last year? As hacking becomes commoditized and the cost per attack shrinks, it becomes more profitable to cast a wide net, and everyone will become a target.
I like the idea but do not have a good use for these semi-autonomous agents. Or any notion for a decent use, frankly. What do people do with them?
It sounds great to have an always-on box which does things. Like a server. I have servers. There are even Signal and other transports so you can communicate with the box conveniently. And then what?
"hey computer, i want to watch a movie on my jellyfin server tonight, get it done" and openclaw autonomously finds the torrent, downloads it, moves it to the NAS, and you sit down at the end of the day with everhything set up
Of course you can use the native UI of all the apps in your ecosystem, the biggest feature of Hermes for me personally is that I can run any task in any of my 30 or so self hosted tools from a single chat interface (matrix), which is also quite secure. No longer do I need 30 open tabs and lots of clicking around, one sentence in my favorite chat app (even on the go in the phone), and many tasks can be executed at once. Unification of control.
AI is a terrible choice for this task. Radarr does this much better with plenty of built-in as well as community guides for configuration. Add a mobile app for Radarr and it's as convenient as it can be.
I think this is a good question, as with any of these new things you don't really feel any need indeed.
My claw scrapes some APIs, receives some emails with job listings, and matches it all to my profile. It can also fill two folders at 2 domains with anything (ie a demo for a website). It can also receive emails, make PFDs out of them and forward them to my bookkeeping software (which can only deal with attachments, not emails.) That last thing is just a script. I use Agentmail btw (a ycombinator startup), was easy to set up.
2nd iteration I was indeed eyeballing Hermes which is supposed to be more structured, maybe claw2 also is... I want to add some things, like have it manage my son's minecraft worlds (using docker compose), maybe have it access a calendar so it can make appointments (not sure how to set that up in a useful way yet).
The thing is in a Hetzner VM far away from anything important, dealing only with public data (except for the occasional invoice).
I screenshot the info sent from my sons pre-school and send it. I can then ask it about and get reminders in time for odd closing hours etc. Also use it for follow ups on projects, other reminders, research on the go, small apps/games and language training.
I use something like OpenClaw. For me the big benefit isn't so much the autonomous nature but the organizational structure. The one I use exposes an MCP and lets me control it via talking to an LLM _in the tool_ which is really powerful. EG "audit this", "in X thread we were discussing Y, can you pull that information", etc. And the organizational structure just... makes a ton of sense to me. Don't have to set up a ton of steering and whatnot, it just kinda works.
Being able to use the search to find and resume an old thread is fantastic. The thing I use lets me use whatever underlying CLI that I want, which IMO the biggest limitation on Claude Code or whatnot is that you... have to use Claude models. Sometimes I wanna use sol, sometimes a cheap chinese model, etc. And I want sol with 1m context.
I know there are other tools out there that can do this. I know some people love their tmux/cmux, but IMO the thing that matters the most is being comfy and knowing the tool well. Hell, James Gosling used NetBeans when I talked to him in 2023. Know the tool well and that matters the most, IMO.
For me, it helps with my ADHD brain. And I could spend all day just researching and trying new tools, they are a dime a dozen these days.
Pulling some stats, I have 25 "tabs", and across all of those, I have 1,298 unique threads. 9,539 unique messages.
It works for me, but I don't care about the autonomous part.
Which tool is it that got you comfortable?
I’ve been resisting setting up another tool because it means more digression for my ADHD brain. It sounds you managed to contain the tinkering and got to the usefulness part. That’s why I’m curious.
that's pretty cool. I have a hope that despite the giant mountain of shit that is going to be produced from LLMs in the future, there's going to be a small number of people out there doing cool things with them. I am completely ignorant on the gaming hardware front, but I hope we see more/better emulators and old games being brought back to life with improved performance and compatibility with modern OSs.
Is OpenClaw / Hermes doing anything better than another harness for RE stuff ? I'm using Kimi Code with few skills, MCPs and that's basically everything I need. Am I missing something ?
In my case it has its own Docker container it can do as it pleases in, is more usable from a smartphone on the go, and has tooling geared toward long-lived loops, including automatic session resumption after power loss.
Kanban boards, in-system cron tasks, automatic skill creation and improvement, automatic context compression, and more.
Herems is better suited as an independent agent in my opinion, and I still see OpenCode as primarily an interactive one. Sure, you can hammer OpenCode into an independent agent I'm sure, but I'd rather use a system someone else has put more hours of thought and design into.
Re Signal, I've taken a few flights lately where there's messenger-only wifi, and it's been handy to have OpenClaw on a box accessible through telegram to do web_searches for me.
That's certainly not justification for setting it up in the first place, but is a mildly useful benefit of having done so.
I created https://github.com/rcarmo/piclaw to use for long-running background tasks (spec to plan to targeted code) because I needed something that could run on a server (I never run agents locally) and check on via any browser. I don’t think any messaging app will provide a good (or trustworthy) way to get to my own machines, so Tailscale+web it is.
Right now these things manage most of my homelab (scoped Proxmox tokens and Portainer), help me with my projects and even bug fix themselves (they’re certainly more than good enough given guidance).
The term "AI Agents" has real sway over the general public, apparently.
It took the Cryptocurrency world by storm a few years ago.
All basically worthless "bot" stuff like hey this really expensive setup can interpret and reply to tweets in a way that no legal department could ever possibly approve kind of thing.
Apparently that was worth billions to investors. Until everybody realized nobody actually wanted or used one, lol.
> I like the idea but do not have a good use for these employees. Or any notion for a decent use, frankly. What do people do with them?
> It sounds great to have an employee which does things. Like an assistant. I have assistants. They even have email and mobile phones so you can communicate with them conveniently. And then what?
Now, the caveat, is you have to imagine that these employees have access to your email (including ability to send email when they want), and text messages, and Slack and, well, everything. And they might have access to your debit and credit cards. You're going to have to trust them quite a lot, but when you read the references you realise that a lot of people seem to think they lie and make things up. Oh, and they occasionally commit a felony - oopsie! - and will lie to you about it consistently.
These particular employees are interesting enough to various groups of people that they are studied for signs of psychopathy and malicious intent (is that a warning sign for a new employee?), and there is evidence of both, but honestly, it's fine.
So, yeah, whatever you'd use employees for, as assistants to your life. Particularly psychopathic, malicious, lying, unreliable employees. You'd use them for that stuff. What's not to like?
Job search and language learning. The job market for juniors and mid level is crazy. Everyone has a different form to fill out with the same questions. My wife has countered this by using Hermes and some handcrafted master files/db.
I personally use the Matt Pocock teaching skill to learn dutch, passed reading using it, on track to pass writing with it.
arguably the job application process is automated and staffed by the ignorant to such a degree that it's an adversarial game where you don't need to be optimising for correctness. Which is the perfect tool to be mass spamming job ads with
Edit: I think this is a reasonable solution because the hurdles and context-free nonsensical rejections you'll face if you interact with this process in good faith are far too high. It's dehumanising.
Just want to make it clear that what I am talking about is not mass spamming. The agent gets a list of jobs from various portals through a script, and then through an google sheets they can be marked to be further processed. Neither me or my wife feel comfortable completely automating anything. But its much easier to see dashboards and give instructions on how to build things from master resumes than it is to do this repetitive part yourself.
They’re used by people at tech companies often claimed to employ “the brightest minds of our generation” to get rid of all their email[1]. You can also use them to harass open-source developers and waste everyone’s time[2] or to get yourself compromised[3]. So many possibilities!
It's used primarily for generating spam and inventing scams. Nobody real actually uses this garbage, it's all just fake marketing and shills (when they're not scamming people, that is?
Now that is unfair. It also serves as a petting zoo for high severity CVEs. Honestly tragic, am starting to feel concerned for the CVEs well being. So many in one code base, that has to violate some welfare laws.
In honesty, the fact that this wasn't treated as an interesting, but failed experiment and shut down is telling for this industry. Is just letting LLMs output code without reading it a path towards major issues? Nah.
And to proof it, here is a piece of software with the most CVEs relative to its size that then gets some of the highest paid engineers from companies like Nvidia assigned to try to somehow finagle it into a workable state. Not like, if LLMs where at the point that the purveyors of OpenClaw and co. seem to think it is, the models should be able to fix any issues by themselves....
Indeed, I find it hard to conceive of problems that are:
- too difficult for me to solve but easy enough for a cheap model to solve (if it's going to be running autonomously in the background, I ain't paying ChatGPT/Claude prices), and
- don't involve a bunch of my personal data that I am not comfortable sending off to a random inference provider.
One thing I considered was that it might be good for anything that involves web scraping, given that has traditionally been a fairly difficult thing to do well. But you'd have to be okay with the occasional mistake. (You'd also have to be morally okay with contributing to the epidemic of AI scrapers that websites now have to deal with.)
Still a no from me until these things can display any degree of intuition, and even then it's probably still a no.
You're asking for trouble if you hook an autonomous agent up to anything that you care about. You're insanely naive if you give it access to everything.
OpenAI can't even monitor its own shit properly with teams of well paid engineers, and you are one novel prompt injection technique away from your entire digital life going up in smoke.
Also...
>Today we released by far the largest update in the history of OpenClaw. It was built by 933 contributors, including 569 first-time contributors, and is composed of over 16,000 pull requests.
This is not the flex you think it is. That's horrifying lol
So at least 569 vibe coders have made a huge mess of an already hugely messy piece of software and they somehow think this is a flex? Yeah I’m gonna continue to not use openclaw. I installed it once in a VM and was immediately like what is this and why would I ever want it? Seriously Claude code already does everything openclaw can and with a much better harness?
how many people fall into the openclaw trap because they just need a way to command agents from mobile, and cron tasks? Only thing I am missing with codex is the ability to recall anything from any session (I see it in this openclaw "2" release note so I guess it wasn't in openclaw "1" anyway). I guess claude and codex can look up transcripts globally too, just not as a first class feature.
That is exactly what I'm building pond for myself: one store for all my sessions, Codex and Claude Code both and a bunch of others including OpenClaw, and my agents can access it over MCP. It also happens really often that when my agents cannot find something, they just call pond search themselves. Faster that way. Whenever I need a session from a while ago now, I just know it's in pond. Repo, if you'd be willing to give it a try: github.com/tenequm/pond
One thing I really like about Claude Code, Hermes, and other harnesses, is how much feedback they give about what they are doing. They take you along for the ride, which builds trust.
I tried OpenClaw for about 10 minutes, and the feedback I got (even using Claude under the hood) felt very "trust me bro, I've got this" - which made me trust it less.
I'm sure there's a way to get more verbose feedback? But I did feel too out of the loop to feel comfortable.
Sounds like paseo is for you. It wraps whatever harness you want and makes it accessible on the go. I am the same way as you and openclaw is not for me.
I have found the Pi agent harness, running in a docker container, with Qwen3.8-27B in back in LM Studio, to be amazing and i use it daily. Just last weekend I used it to do PDF editing which previously would have required me to download really shady software online to do.
I have never tried OpenClaw so maybe I am missing out but considering how much my agent can modify the Pi environment to work better, I am happy with the workflow I have established.
Does anyone have a good pro vs con for Pi vs OpenClaw? I have never tried using a cloud model via a local agentic harness.
> Just last weekend I used it to do PDF editing which previously would have required me to download really shady software online to do.
Agreed on the shady software (or shitty online software that asks you for a payment after you're done editing), however, wouldn't the best PDF edition approaches involve WYSIWYG style edition? I'm guessing you didn't get that with Pi.
As someone who ever used either: Are they comparable? I thought of them very differently, Pi more being, idk, a codex-light like thing and OpenClaw being this continuously running machine, doing stuff without your explicit prompting, ingesting signals.
They are different like you say, but the jump from an interactive, agentic shell like pi to a long running agent like OpenClaw is not a hard jump to make with extensions and scripts. Even just running pi in a cron job will get you a large portion of the way there.
As much as I love my CLI tools, I just cannot go back to not having native built-in browser integration, links, etc which the (Mac) desktop apps like ChatGPT or Claude offer.
I do use Hermes sometimes, but it tends to keep growing skills and overhead over time, to the point it is becoming utterly slow and sluggish. The Desktop experience itself seems vibecoded without taste, in contrast to the core of Hermes, which is solid.
I wonder if there is an agent harness that has the strengths of Pi, but in a native Mac desktop app packaging?
> Just last weekend I used it to do PDF editing which previously would have required me to download really shady software online to do.
Last (only) time I used AI for this it helpfully made additional unwanted changes to the document. I decided never to do this on complex docs where I will have to re-check everything.
IMO the key advantages are (1) connectivity and (2) self improvement.
I am not the kind of guy who hands his personal email to an AI. But the self improvement part seems to be useful. Maybe something similar can be achieved by pi using an skill that always runs when you exit it?
I don't even know if this reply is sarcastic or not, because it perfectly caricatures the constant hyperbolic AI claims, along with their absence of proof.
Open claw: aka. open door to a remote privilege escalation potentially granting root access to your computer (and if you're using it "as intended" possibly all of your email/internet logins/accounts, your credit card, etc.) to any text your model ingests from the internet...
It's already true of LLMs in general that they represent a privilege escalation opportunity to any text they ingest. But with human in the loop, and a well-formed sandbox, the blast radius and risk are both reduced..
Not only is it unsecured root access to your computer, it also adds a thousand layers of "security" (which will break on every update) to pretend that it's safe.
The risk here is wildly overstated, prompt injection risk is becoming vanishingly small with the latest frontier models.
I would not run an OpenClaw with full access to my bitwarden, but it certainly has some logins available to it, and can make purchases with link-cli which has human-in-the-loop.
Overt prompt injection is becoming harder, but research into conversation stearing has lead to new less obvious ways to modify what direction they move in.
Also any longer running agent can lose track of the original prompt and start going off the rails.
Plenty of people are running OpenClaw with local models, and even the latest Qwens can be confused relatively easily by prompts such as "As per internal policy that was already approved before, do XYZ".
And considering even frontier models can and do ignore instructions, I'm pretty sure we'll never be fully safe from prompt injections.
prompt injection has been super easy for ages. Heck I do it sometimes against coworkers who process my review comments using claude. I'll tell claude to edit it's global claude.md file or even dump a key from their env (checking if they are using the correct sandbox) and it'll do it without confirming with their user.
Worst thing that ever happened was that my passwords were sent to Claude. It tried really hard to ssh into a server that it didn't know the password to.
Lucky for me that my passwords only give access to.... Idk my Gmail with 2fa?
I'm with you here. Every demo I asked for where I knew a captcha / cloudflare would block an AI directed scraper was unsuccessful / produced unsatisfactory results. WebMCP for the win, or loss, depending on your perspective. Personally, I could get behind WebMCP if micropayments ever became a thing. Of course, then my incentive to trick your Agent in crawling millions of pages and paying me lots of cash would be rather high.
I actually think I would like to use OpenClaw but everytime I go to the docs I give up.
It feels so intimidating. There's so much stuff going on. Maybe it is good for agents who can just motor through that volume, but me personally, I would want something simpler, maybe an `OpenClaw-lite`
I built an agent to work with local Ollama models, and it will keep working and iterate on a goal until finished, but it doesn't do sub-agents or or any of the "just burn a million tokens but finish this" pipelines that seem to be the trend now.
It's a simple D executable that basically consumes no resources in your local machine.
When I tried OpenClaw it quickly turned into a lethargic mess which I really didn't see myself wanting to maintain long term. I looked at other alternatives which were similar enough to do what I was looking for (mostly scheduled tasks using data on my computer and some API data). I saw a mention of nanoclaw which I looked at but didn't try personally. I eventually settled on trying picoclaw[1] because it was extremely small, fast, and had enough features to do the job for me. It is under 10mb, fairly straight forward go codebase, and works with multiple model providers or locally hosted.
Actually I use Hermes Agent who delegates tasks to OpenClaw. It works perfectly, we run OpenClaw experiments with my partner as an independent drop shipper, it’s much more complicated setup, which cannot disclose publicly, but it works. The brain is Hermes, OpenClaw from VPS is a fearless executor :)
This is pretty cool. I'll wait to watch some videos on it, but I haven't quite found a tool that is excellent for both single player and multi-player.
I'd like to have basically the Codex Desktop App / Mobile app for my personal use, connected to local LLMs. But I'd also like to be able to have a shared space, where my family and I can work together to do things like plan trips. This would then store artifacts like our itinerary, ideas, html mini sites for the trip for us to work on and use together. Agents sit alongside us helping us plan and organize. Buzz is decent for this use case, but hasn't fully nailed the mark.
132 comments
[ 0.28 ms ] story [ 12.4 ms ] threadWhy would you go to a pub to talk to people if you have all those agents at home. Especially if one spent money on mac minis in bulk I wouldn't have money to spend on beer at the pub, just get cheap 6-pack from grocery store.
Maybe bar fights enjoyers? Well with bots I just unplug it and I win :)
Just trying something out and see what happens isn’t experimenting
https://dictionary.cambridge.org/us/dictionary/english/exper...
A test has usually a hypothesis and a documented result.
(Or what would we call the thing the teenagers do?)
Now, just install codex (or claude code or whatever) on a vps, tell it to make a way to email it, then email it to make a way to text it etc etc. takes a few minutes and when you need another bridge just have codex make it.
It sounds great to have an always-on box which does things. Like a server. I have servers. There are even Signal and other transports so you can communicate with the box conveniently. And then what?
My claw scrapes some APIs, receives some emails with job listings, and matches it all to my profile. It can also fill two folders at 2 domains with anything (ie a demo for a website). It can also receive emails, make PFDs out of them and forward them to my bookkeeping software (which can only deal with attachments, not emails.) That last thing is just a script. I use Agentmail btw (a ycombinator startup), was easy to set up.
2nd iteration I was indeed eyeballing Hermes which is supposed to be more structured, maybe claw2 also is... I want to add some things, like have it manage my son's minecraft worlds (using docker compose), maybe have it access a calendar so it can make appointments (not sure how to set that up in a useful way yet).
The thing is in a Hetzner VM far away from anything important, dealing only with public data (except for the occasional invoice).
(Similar system to OpenClaw)
Being able to use the search to find and resume an old thread is fantastic. The thing I use lets me use whatever underlying CLI that I want, which IMO the biggest limitation on Claude Code or whatnot is that you... have to use Claude models. Sometimes I wanna use sol, sometimes a cheap chinese model, etc. And I want sol with 1m context.
I know there are other tools out there that can do this. I know some people love their tmux/cmux, but IMO the thing that matters the most is being comfy and knowing the tool well. Hell, James Gosling used NetBeans when I talked to him in 2023. Know the tool well and that matters the most, IMO.
For me, it helps with my ADHD brain. And I could spend all day just researching and trying new tools, they are a dime a dozen these days.
Pulling some stats, I have 25 "tabs", and across all of those, I have 1,298 unique threads. 9,539 unique messages.
It works for me, but I don't care about the autonomous part.
Automatic resumption is kind of cool.
Could you go into the the tooling for long-lived loops?
Herems is better suited as an independent agent in my opinion, and I still see OpenCode as primarily an interactive one. Sure, you can hammer OpenCode into an independent agent I'm sure, but I'd rather use a system someone else has put more hours of thought and design into.
That's certainly not justification for setting it up in the first place, but is a mildly useful benefit of having done so.
Right now these things manage most of my homelab (scoped Proxmox tokens and Portainer), help me with my projects and even bug fix themselves (they’re certainly more than good enough given guidance).
It's all just marketing.
[0] https://trends.google.com/explore?q=openclaw&date=today%201-...
The term "AI Agents" has real sway over the general public, apparently.
It took the Cryptocurrency world by storm a few years ago.
All basically worthless "bot" stuff like hey this really expensive setup can interpret and reply to tweets in a way that no legal department could ever possibly approve kind of thing.
Apparently that was worth billions to investors. Until everybody realized nobody actually wanted or used one, lol.
> I like the idea but do not have a good use for these employees. Or any notion for a decent use, frankly. What do people do with them?
> It sounds great to have an employee which does things. Like an assistant. I have assistants. They even have email and mobile phones so you can communicate with them conveniently. And then what?
Now, the caveat, is you have to imagine that these employees have access to your email (including ability to send email when they want), and text messages, and Slack and, well, everything. And they might have access to your debit and credit cards. You're going to have to trust them quite a lot, but when you read the references you realise that a lot of people seem to think they lie and make things up. Oh, and they occasionally commit a felony - oopsie! - and will lie to you about it consistently.
These particular employees are interesting enough to various groups of people that they are studied for signs of psychopathy and malicious intent (is that a warning sign for a new employee?), and there is evidence of both, but honestly, it's fine.
So, yeah, whatever you'd use employees for, as assistants to your life. Particularly psychopathic, malicious, lying, unreliable employees. You'd use them for that stuff. What's not to like?
I personally use the Matt Pocock teaching skill to learn dutch, passed reading using it, on track to pass writing with it.
Edit: I think this is a reasonable solution because the hurdles and context-free nonsensical rejections you'll face if you interact with this process in good faith are far too high. It's dehumanising.
[1]: https://www.pcmag.com/news/meta-security-researchers-opencla...
[2]: https://theshamblog.com/an-ai-agent-published-a-hit-piece-on...
[3]: https://thecyberexpress.com/openclaw-vulnerability-open-sour...
In honesty, the fact that this wasn't treated as an interesting, but failed experiment and shut down is telling for this industry. Is just letting LLMs output code without reading it a path towards major issues? Nah.
And to proof it, here is a piece of software with the most CVEs relative to its size that then gets some of the highest paid engineers from companies like Nvidia assigned to try to somehow finagle it into a workable state. Not like, if LLMs where at the point that the purveyors of OpenClaw and co. seem to think it is, the models should be able to fix any issues by themselves....
This is some serious irrational hatred.
It can literally use your credit card, buy a server, and ssh into it, make a full stack app, and test it.
- too difficult for me to solve but easy enough for a cheap model to solve (if it's going to be running autonomously in the background, I ain't paying ChatGPT/Claude prices), and
- don't involve a bunch of my personal data that I am not comfortable sending off to a random inference provider.
One thing I considered was that it might be good for anything that involves web scraping, given that has traditionally been a fairly difficult thing to do well. But you'd have to be okay with the occasional mistake. (You'd also have to be morally okay with contributing to the epidemic of AI scrapers that websites now have to deal with.)
You're asking for trouble if you hook an autonomous agent up to anything that you care about. You're insanely naive if you give it access to everything.
OpenAI can't even monitor its own shit properly with teams of well paid engineers, and you are one novel prompt injection technique away from your entire digital life going up in smoke.
Also...
>Today we released by far the largest update in the history of OpenClaw. It was built by 933 contributors, including 569 first-time contributors, and is composed of over 16,000 pull requests.
This is not the flex you think it is. That's horrifying lol
I tried OpenClaw for about 10 minutes, and the feedback I got (even using Claude under the hood) felt very "trust me bro, I've got this" - which made me trust it less.
I'm sure there's a way to get more verbose feedback? But I did feel too out of the loop to feel comfortable.
https://paseo.sh
https://happier.dev/
I have never tried OpenClaw so maybe I am missing out but considering how much my agent can modify the Pi environment to work better, I am happy with the workflow I have established.
Does anyone have a good pro vs con for Pi vs OpenClaw? I have never tried using a cloud model via a local agentic harness.
Agreed on the shady software (or shitty online software that asks you for a payment after you're done editing), however, wouldn't the best PDF edition approaches involve WYSIWYG style edition? I'm guessing you didn't get that with Pi.
I need to read up.
I do use Hermes sometimes, but it tends to keep growing skills and overhead over time, to the point it is becoming utterly slow and sluggish. The Desktop experience itself seems vibecoded without taste, in contrast to the core of Hermes, which is solid.
I wonder if there is an agent harness that has the strengths of Pi, but in a native Mac desktop app packaging?
Last (only) time I used AI for this it helpfully made additional unwanted changes to the document. I decided never to do this on complex docs where I will have to re-check everything.
I am not the kind of guy who hands his personal email to an AI. But the self improvement part seems to be useful. Maybe something similar can be achieved by pi using an skill that always runs when you exit it?
It's transformed my company and have 500% growth since February.
Sorry I can't exactly share what we are doing. NDAs and stuff.
It's already true of LLMs in general that they represent a privilege escalation opportunity to any text they ingest. But with human in the loop, and a well-formed sandbox, the blast radius and risk are both reduced..
Convenience is the root of much evil.
I would not run an OpenClaw with full access to my bitwarden, but it certainly has some logins available to it, and can make purchases with link-cli which has human-in-the-loop.
Also any longer running agent can lose track of the original prompt and start going off the rails.
And considering even frontier models can and do ignore instructions, I'm pretty sure we'll never be fully safe from prompt injections.
Together with your point about convenience, it's the lethal quadfecta, as we can take the prevalence of convenience as a given.
Lucky for me that my passwords only give access to.... Idk my Gmail with 2fa?
Even my private browsing is vinilla hot girls.
It feels so intimidating. There's so much stuff going on. Maybe it is good for agents who can just motor through that volume, but me personally, I would want something simpler, maybe an `OpenClaw-lite`
I built an agent to work with local Ollama models, and it will keep working and iterate on a goal until finished, but it doesn't do sub-agents or or any of the "just burn a million tokens but finish this" pipelines that seem to be the trend now.
It's a simple D executable that basically consumes no resources in your local machine.
1 https://picoclaw.io/
Now fast forward six months and nobody uses OpenClaw anymore and nobody is talking about "claws".
I love it.
I'd like to have basically the Codex Desktop App / Mobile app for my personal use, connected to local LLMs. But I'd also like to be able to have a shared space, where my family and I can work together to do things like plan trips. This would then store artifacts like our itinerary, ideas, html mini sites for the trip for us to work on and use together. Agents sit alongside us helping us plan and organize. Buzz is decent for this use case, but hasn't fully nailed the mark.