"Parasitic compute" is strange way to describe "a user running dev workflows on their own GitHub Actions allocation"
when you run ghost, it creates a "ghostbox" - an ephemeral machine on your GitHub account, on your GitHub actions minutes, accessible only by your SSH identity. It's orchestration around GitHub's infra.
Proprietary software built on GitHub is not exactly an unusual category.
I don't believe in releasing source anymore after years of doing it. It's closed source, Rust binary, proprietary but free software offered as a utility. It's the same patterns as used by coding agents and many other CLI tools.
So, your threat model is that I’m really building my business and reputation by creating illegal, criminal malware? Ponder that, is that really a plausible thing to you? You think that about me?
No, I only use open source agents, weird of you to assume that I would make an exception just for agents. My threat model is that trust is earned, not given away to complete strangers who act hostile to simple inquiries.
I was sincerely interested in why you were choosing closed source, you decided to take it as an attack.
lol No, I didn’t take it as an attack (tho that's what it was) i just didn’t answer your question. If you’re not okay with that i guess you’ll need to figure that out yourself.
That’s where i was going with my reply - i wanted you to think more about it. You perceived it as hostile but really I was just asking you some questions, simple ones. It does seem like you’re projecting here, maybe consider that more.
Because even tho I didn't take it as an 'attack' it was an attack, really. Think: What are you saying: you don't trust some software that I wrote. What does that mean? That means you think it's going to do something bad. That means you think I'm going to try to do something bad to you, by this beautiful creative effort that I'm putting out. Like wut? You really are not a builder if you don't understand how that feels, for me, but also -if you don't see the problem with just thinking that's what I'm going to do, like that's crazy. Like you think it's okay to just accuse me of that, just casually, like what? And then you don't think I can saying anything about that - because otherwise I'm taking it as an attack. Which is what it was. So of course I'm going to say something, and I can. And if you don't like that - maybe think more before you accuse people of what you're just projecting.
So, I took it as ignorance, which is what it is, I assume, which is why I asked you questions. Because if I was going to give you my real answer, I don't think you would understand it. That's why I didn't answer your question, because I didn't feel you'd understand (why might that be? Because you just fake accused me of trying to do something I'm not doing?). Is that not obvious to you?
Anyway, what else were you saying? Not weird about agents, it’s so common. I guess you’re a little unusual in your fastidiousness about that. But that’s not a problem.
What about apps on your phone, are you okay with that? Or you have, like, a dumb phone?
If you do want to know my views on open source, maybe you can try your empathy and tell me why you think?
It’s okay if you don’t want to. I’m finding the interaction with you a little boring… lol
Ah, so very considerate of you, well someone's read it, but the short version is: you have 0 right to attack or accuse me in any way. The fact that you want to, just shows you're a bad perosn. You're wrong. You assumed I'm doing something bad, but you don't know me, you tried to blame me for your prejudice, but that's just you, projecting, crazy.
"Why do you want to protect your IP/time/effort rather than giving away your source code? I don't run binaries as a general rule, nothing to do with ghost, which looks cool, btw." is totally fair.
Assuming bad intent, malware, or hidden wrongdoing is not neutral criticism - it is warrantless attack.
No-one knows who you are, and you’re clearly not against using something like GH Actions in a way other than its intended purpose. What’s to say you won’t pivot to running a tiny VM on my machine and making it available to others?
Is that right? Well, some people know. I’m Cris, and you are?
But wait I’ve been building so much, for all this time, but you think what I’ve really being doing is building malware, and there’ve been no consequences, somehow nobody’s noticed and I’ve just “gotten away with it”?
I don’t think anything, and I’m not accusing you of anything, I’m just saying, a lot of folks started with pure motives and got poisoned along the way.
For what it’s worth, it’s not very reassuring that you have a bunch of open source projects but you’ve decided this one is not going to be. Rather than showing I can trust you, it rather makes me wonder what you’re hiding.
The answer may well be nothing, but it’s still strange.
I get you might feel that way about it, but that’s not how it is.
The strange thing is your reaction, don’t you think: If you see a proprietary source product and you think “what’s it hiding?” and if you can’t respect a boundary of not revealing source without projecting an imagined bad onto that, that’s just you, my dude, and I’m not responsible how you react at all.
So you might wanna try to put your mistaken attitude on me, but really you need to own that. And your attitude seems mistakenly entitled.
Also the trust issues are warrantless. And, in reality, if you look at my projects, the most important ones are not “open source”.
You judged too quickly, without context, like many here and arrived at conclusions that are just not warranted.
You shouldn’t be arguing with anyone about that because why you came to those doubts or conclusions is something you have to figure out yourself, it’s not something anyone else can help you with.
> I’m just saying, a lot of folks started with pure motives and got poisoned along the way.
That’s not how I see things. That’s not been my experience of the world. I understand if it’s been yours though. Poor you. I guess in that case my advice is just try to keep in mind that not everyone is gonna have the same kind of negative outlook as you and try to be understanding towards them. There’s a lot of good in the world if you open your eyes to it, I hope you find some.
> if you can’t respect a boundary of not revealing source without projecting an imagined bad onto that, that’s just you, my dude, and I’m not responsible how you react at all
I’m responding to the change, as something worth scrutiny. You used to publish open source projects, now this is closed source. Why?
> So you might wanna try to put your mistaken attitude on me, but really you need to own that. And your attitude seems mistakenly entitled
What mistaken attitude, what am I putting on you, and what is my “entitlement”?
> You judged too quickly
> You shouldn’t be arguing
> my advice is
Please stop dressing up your arguments as some kind of metaphysical commentary on my character. I don’t need advice, I didn’t judge you, and I didn’t plan on arguing.
You built something, some people think it’s cool, a lot of people think it’s problematic. You want to keep it closed source, some people find that worrying.
Keep your faux pity for yourself, engage with me in good faith on the merits of the points I’m making, otherwise we’re done here.
You think I owe you source code, is entitled. You project strange onto change, is low empathy. There's no metaphysics, your unwarranted criticism is a reflection of your character. Don't pretend your weird subjective reaction is anything I need to respond to, nor any reflection of me - it's just you.
You have 0 right to attack or accuse me in any way. That you think you do makes you even more entitled and low empahty. Geez....
Their pattern here of immediately going on the offensive to even the smallest amount of inquiry or criticism is totally normal and not at all suspicious.
Maybe they're just having a bad day. Friendly reminder that you don't have to respond to something as soon as you read it, or even at all.
Yeah, I think that's what the program creates in your github account. I see the source to those files embedded in the executable. (I'm not running the executable, but I downloaded the linux one to my mac to take a look inside.)
And abracadabra - it will return. That's just the crowd madness leading folks from this very thread to abuse the flag/report button on GitHub repo to get it auto-disabled.
I trust it will resurrect once GitHub gets around to inspecting.
Really cool. I've been building a mission control system (multi agent orchestration) that follows very similar patterns of spec driven development, steering, and task management. Having this baked into an IDE is a great idea.
For observability, would be amazing to have session replay or at least session exploration built in. Kinda like git history but tied to tasks and tool use instead of file diffs.
Right on. Good luck! You might also want to play around with https://github.com/simple10/agent-super-spy if you want to see the raw prompts claude is sending. It was really helpful for me to see the system prompts and how tool calls and message threads are handled.
Sub-agent trees are fully tracked by the dashboard. When an agent is spawned, it always has a parent agent id - claude is sending this in the hooks payload. When you mouse over an agent in the dashboard, it shows what agent spawned it. There currently isn't a tree view of agents in the UI, but it would be easy to add. The data is all there.
[Edit] When claude spawns sub-agents, they inherit the parent's hooks. So all sub-agents activity gets logged by default.
I hit a lot of limits on Pro plan. Upgraded to Max $200/mo plan and haven't hit limits for awhile.
It's super important to check your plugins or use a proxy to inspect raw prompts. If you have a lot of skills and plugins installed, you'll burn through tokens 5-10x faster than normal.
Also have claude use sub-agents and agent teams. They're significantly lighter on token usage when they're spawned with fresh context windows. You can see in Agents Observe dashboard exactly what prompt and response claude is using for spawning sub-agents.
I'm not actually reading the jsonl files. Agents Observe just uses hooks and sends all hook data the server (running as a docker container by default).
Basic flow:
1. Plugin registers hooks that call a dump pipe script that sends hook events data to api server
2. Server parses events and stores them in sqlite by session and agent id - mostly just stores data, minimal processing
3. Dashboard UI uses websockets to get real-time events from the server
4. UI does most of the heavy lifting by parsing events, grouping by agent / sub-agent, extracting out tool calls to dynamically create filters, etc.
It took a lot of iterations to keep things simple and performant.
You can easily modify the app/client UI code to fully customize the dashboard. The API app/server is intentionally unopinionated about how events will be rendered. This was by design to add support for other agent events soon.
The hooks approach seems much cleaner for real-time. Did you run into any issues with the blocking hooks degrading performance before you switched to background?
Sort of. It wasn't really noticeable until I did an intentional audit of performance, then noticed the speed improvements.
Node has a 30-50ms cold start overhead. Then there's overhead in the hook script to read local config files, make http request to server, and check for callbacks. In practice, this was about 50-60ms per hook.
The background hook shim reduces latency to around 3-5ms (10x improvement). It was noticeable when using agent teams with 5+ sub-agents running in parallel.
But the real speed up was disabling all the other plugins I had been collecting. It piles up fast and is easy for me to forget what's installed globally.
I've also started periodically asking claude to analyze it's prompts to look for conflicts. It's shockingly common for plugins and skills to end up with contradictory instructions. Opus works around it just fine, but it's unnecessary overhead for every turn.
reply