Getting to Know Deplo

itzsenu

@deplo said: although we don't find some of its conclusions particularly accurate.

maybe..... i checked the whole commit history. what i'm saying is the architecture isnt reliable. checked the CLAUDE.md and AGENTS.md. if this was developed using Claude, it's definitely a newer model, depending on what those md files say.

@deplo said: we believe it’s important to set clear boundaries around how we use it.

if u really knew what u were doing and set boundaries, i don't think you'd ever publish this with 2 critical memory leaks, both causing crashes. so clearly, this was developed across many sessions BY CLAUDE

@deplo said: then be reviewed and refined using the tools available today

u clearly said it, but didn't do it correctly. so if ur saying this, u should at least architect it correctly. so thinking isn't really thinking here.

@deplo said: we are absolutely PRO-AI

Yup, Sure.

@deplo said: You only need to look at websites that are genuinely AI-generated to notice the pattern

nah, no need. i wanted to recreate a website made entirely by a human using blender. got curious, used my friend's claude api cuz i'm broke, threw it into fable and this is what i got. Fully three js.

@deplo said: Web Developer and Designer, as well as a UI/UX Graphic Designer. The other members of the team are specialized in different fields

okkey, ur a frontend designer, so how would u architect a PaaS? i also checked ur other two, so honestly idk.

@deplo said: Thank you for giving us the opportunity to properly explain our vision and approach to AI.

i cant say this is vibe coded. if they used a newer Claude model, there wouldn't necessarily be any obvious trace of it being vibe coded. so there's no way to say for sure. but from what i checked, there are a lot of issues, even with basic stuff. if this was properly human directed, why publish it with a memory leak that's been sitting in the code for so long? clearly wasn't tested properly. and most of the time, AI ends up directing the whole project anyway, cuz current Claude models are good enough to handle it.

also poorly optimize

so all i have to say is, this has sooo many issues. even security issues, basic issues, and a lot of other stuff. if this really had human supervision, how did it end up here? imo, that alone shows how the AI was being used.

@deplo said: I've tried Podman and it's great.

i tried it cuz i wanted to change things up a bit. it's spanky and smooth, so i wanted to see how it works. i use Dokploy for simple projects where i don't wanna waste time. otherwise i usually do everything manually through the CLI.

//

anyway, like SSHSpan, im not wasting my life on this. this thing has so many issues. i somehow got promoted to unpaid volunteer security tester without applying. damn, only realised it now. anyway, vacation ends today, back to reality

CloudHopper

@itzsenu said:
i cant say this is vibe coded. if they used a newer Claude model, there wouldn't necessarily be any obvious trace of it being vibe coded. so there's no way to say for sure. but from what i checked, there are a lot of issues, even with basic stuff. if this was properly human directed, why publish it with a memory leak that's been sitting in the code for so long? clearly wasn't tested properly. and most of the time, AI ends up directing the whole project anyway, cuz current Claude models are good enough to handle it.

It's obviously vibe coded because no human developer would add this many comments to a Dockerfile, and even if they did they wouldn't be full of Claudisms: https://github.com/DeploCloud/deplo/blob/main/Dockerfile

s0n1c

Why do I keep seeing hosting providers release vibecoded products that nobody asked for when almost every vibecoded product released by a host gets absolutely shit on 99% of the time? Is it a humiliation kink or something?

itzsenu

@CloudHopper said: It's obviously vibe coded because no human developer would add this many comments to a Dockerfile,

yes (it's fully AI-built. you can check the full commit history, i went through all of it.) i mean zero human directed work, but AI directed, not vibe coded. anyway, i found traces proving it was vibe coded. there are so many basic issues. they used a new model. i already said there are so many issues and i'm still waiting for their reply.

and these aren't even the main issues. these are the traces i found showing it wasn't properly human directed:

• lib/deploy/build.ts has a 630line runDeployment god-function mixing deploy, render, and routing logic.
• lib/data/migration-import.ts has 5,513 lines and 97 functions. even with banners, this should be split into a directory.
• appEnv and appEnvKeys duplicate their entire bodies and the same 4 line comment.
• parseStackHealthCheck is exported but only used internally.
• 997 ad-hoc throw new Error("sentence") calls, with heuristic error masking and no central error system.
• AGENTS.md has outdated ADR info, while the ADR README is missing ADR-0030. ( if he literally prompted it correctly, this couldnt have happened. )
• strict: true but no noUncheckedIndexedAccess, despite code doing rows[0] everywhere.

AI: build a self-hosted deployment platform.
Human: alright, keep it clean and maintainable.
AI: understood.
AI: *writes 5,513 lines in one file*
Human: maybe split that up?
AI: no.
AI: *adds comments every 3 lines*
Human: why are there comments everywhere?
AI: so humans understand the code.
Human: okay... remove the AI instructions before pushing.
AI: `CLAUDE.md` added.
Human: you pushed it too?
AI: yes.
Human: ...
AI: deployment platform complete.
Human: did you test it?
AI: i read the code.

i'm not listing the other issues. i have human hands, so i dont care enough to type out 2+ pages of QA for them i hope he'll fix them. overall, this slop is a solid 7/10. good vibe coding. slop, but premium slop. anyway, good night. vacation's over lol.

@itzsenu said: this has sooo many issues. even security issues, basic issues, and a lot of other stuff. if this really had human supervision, how did it end up here? imo, that alone shows how the AI was being used.

let him find his own issues and fix them. if he really needs me to point them out, i'll do it, but after SSHSpan, my little cutie heart is already broken .

@s0n1c said: Why do I keep seeing hosting providers release vibecoded products that nobody asked for when almost every vibecoded product released by a host gets absolutely shit on 99% of the time

probably sponsored or something. an ocean of tokens, somehow distilled into something solid.(premium slop)

GOOD NIGHTTTTTT

zed

posting in a new hot thread

itzsenu

@zed said: posting in a new hot thread

don't think this one's gonna cook.

deplo

@itzsenu said:

@zed said: posting in a new hot thread

don't think this one's gonna cook.

@zed said:
posting in a new hot thread

@s0n1c said:
Why do I keep seeing hosting providers release vibecoded products that nobody asked for when almost every vibecoded product released by a host gets absolutely shit on 99% of the time? Is it a humiliation kink or something?

@CloudHopper said:

@itzsenu said:
i cant say this is vibe coded. if they used a newer Claude model, there wouldn't necessarily be any obvious trace of it being vibe coded. so there's no way to say for sure. but from what i checked, there are a lot of issues, even with basic stuff. if this was properly human directed, why publish it with a memory leak that's been sitting in the code for so long? clearly wasn't tested properly. and most of the time, AI ends up directing the whole project anyway, cuz current Claude models are good enough to handle it.

It's obviously vibe coded because no human developer would add this many comments to a Dockerfile, and even if they did they wouldn't be full of Claudisms: https://github.com/DeploCloud/deplo/blob/main/Dockerfile

@alincupunct said:

@LEBUserJoe said:

@DeluxHost said:

@averagedatahoarder said:
How is this different from Coolify?

You can check it here: https://deplo.build/comparison/coolify

LOL

"Three things"

1 entirely useless, make permissions PER ACTION LOL useless.
Networkin? Balatant lie works fine on coolify if that's complicated ur brain is smooth
Coolify only needs 1 server for team.

3 things, 1 useless, 2 complete lies! what a great ass platform give ur AI tokens back to society.

And that part of the website is HORRIBLE UX.

@LEBUserJoe said:

@deplo said:

@alincupunct said:

@CloudHopper said:

@DeluxHost said:

On what planet is a solution that deploys docker compose stacks an "alternative to the cloud"?

AI generated slogan.

@itzsenu said:

@alincupunct said: AI generated slogan.

something fully generated daily by an unemployed person. first it looked cool, but it doesn’t anymore. anyway, better than SSHSpan.

Hi! Thank you for such a thorough comment, although we don't find some of its conclusions particularly accurate.

We’d like to take this opportunity to clarify our position regarding AI, because we believe it’s important to set clear boundaries around how we use it.

We believe AI is a great and valuable tool that can certainly be helpful. At the same time, we believe ideas should come from the person first and then be reviewed and refined using the tools available today. Fortunately, AI is still not quite capable of coming up with titles that compelling, which is why I can confirm that the copy for our website was entirely conceptualized by a copywriter and then transcribed and refined in certain parts with the help of AI.

The slogan was entirely conceived by the copywriter in question, with no AI involved in the creative process behind it.

We want to be very clear and transparent about this: we are absolutely PRO-AI, but we believe in using it within reasonable limits. AI cannot currently replace a human being or their expertise in a specific field. You only need to look at websites that are genuinely AI-generated to notice the pattern: generic wording, an unclear target audience, zero emotional depth, gradients everywhere, and the same recognizable formatting patterns.

To make our position crystal clear, the founder of the project is a Web Developer and Designer, as well as a UI/UX Graphic Designer. The other members of the team are specialized in different fields that complement the project and bring their own expertise to it.

Thank you for giving us the opportunity to properly explain our vision and approach to AI.
Have a great day!

Aka, 4 paragraphs to say humans wrote the text on the website (Of which has countless errors)

And the entire backend is 100% AI but you avoid talking about it.

Morning, LET!
We've read all your replies carefully, and this will be our last response regarding the concerns raised in this thread.

First of all, Deplo is open-source. We are aware that the public repository has been around for about 2 months, but Deplo itself has been in development for roughly 6 months.
We're also fully aware that, as with any newly launched software, bugs can exist, and that's exaclty why being open-source is a strength: if someone finds a problem, we want to genuinely know about it, in order to fix it as fast as we can.

But, there's an important difference between reporting a real problem and speculating that one exists.
If there is a memory leak, for example, we want to investigate it. But we need to know how Deplo was installed, what happened, how memory usage changed, and ideally have logs or a reproducible scenario.
Without any of that, what exactly are we supposed to test? We cannot debug an issue based on a comment with zero technical information behind it.

The same applies to the "It's made with AI" argument.
A bug does not prove that AI (or a newer model as someone said) wrote the code, nor does AI-assisted development imply that there was no human involvment. Human error exists in every field, including software development, with or without AI.
Also, one of the criticism we found was that our website contains grammatical mistakes. That's a rather strange argument to make immediately after claiming that "basic issues like these" couldn't exist with proper human review. If anything, it demonstrates the rathen than obvious fact that humans make mistakes.

The same goes for "how can one Frontend Developer build a SaaS like this, alone?". Well, nobody said one person built Deplo alone.
We've already explained that there are multiple people behind the project. Again, don't turn assumptions into arguments.

From now on, we are happy to discuss real, reproducible problems and spend our time fixing them. We care far more about fixing a real issue than winning an argument on a forum.
To be clear, we won't continue discussing claims without evidence. If you have facts, we're here. If not, we'll leave it at that.

Taking into consideration that this is a technical forum, we expected more facts and less speculation.

x0x0x

@deplo said: Taking into consideration that this is a technical forum, we expected more facts and less speculation.

Taking into consideration that this is LET, you should expect more drama than anything else.

alincupunct

@deplo said:

@itzsenu said:

@zed said: posting in a new hot thread

don't think this one's gonna cook.

@zed said:
posting in a new hot thread

@s0n1c said:
Why do I keep seeing hosting providers release vibecoded products that nobody asked for when almost every vibecoded product released by a host gets absolutely shit on 99% of the time? Is it a humiliation kink or something?

@CloudHopper said:

@itzsenu said:
i cant say this is vibe coded. if they used a newer Claude model, there wouldn't necessarily be any obvious trace of it being vibe coded. so there's no way to say for sure. but from what i checked, there are a lot of issues, even with basic stuff. if this was properly human directed, why publish it with a memory leak that's been sitting in the code for so long? clearly wasn't tested properly. and most of the time, AI ends up directing the whole project anyway, cuz current Claude models are good enough to handle it.

It's obviously vibe coded because no human developer would add this many comments to a Dockerfile, and even if they did they wouldn't be full of Claudisms: https://github.com/DeploCloud/deplo/blob/main/Dockerfile

@alincupunct said:

@LEBUserJoe said:

@DeluxHost said:

@averagedatahoarder said:
How is this different from Coolify?

You can check it here: https://deplo.build/comparison/coolify

LOL

"Three things"

1 entirely useless, make permissions PER ACTION LOL useless.
Networkin? Balatant lie works fine on coolify if that's complicated ur brain is smooth
Coolify only needs 1 server for team.

3 things, 1 useless, 2 complete lies! what a great ass platform give ur AI tokens back to society.

And that part of the website is HORRIBLE UX.

@LEBUserJoe said:

@deplo said:

@alincupunct said:

@CloudHopper said:

@DeluxHost said:

On what planet is a solution that deploys docker compose stacks an "alternative to the cloud"?

AI generated slogan.

@itzsenu said:

@alincupunct said: AI generated slogan.

something fully generated daily by an unemployed person. first it looked cool, but it doesn’t anymore. anyway, better than SSHSpan.

Hi! Thank you for such a thorough comment, although we don't find some of its conclusions particularly accurate.

We’d like to take this opportunity to clarify our position regarding AI, because we believe it’s important to set clear boundaries around how we use it.

We believe AI is a great and valuable tool that can certainly be helpful. At the same time, we believe ideas should come from the person first and then be reviewed and refined using the tools available today. Fortunately, AI is still not quite capable of coming up with titles that compelling, which is why I can confirm that the copy for our website was entirely conceptualized by a copywriter and then transcribed and refined in certain parts with the help of AI.

The slogan was entirely conceived by the copywriter in question, with no AI involved in the creative process behind it.

We want to be very clear and transparent about this: we are absolutely PRO-AI, but we believe in using it within reasonable limits. AI cannot currently replace a human being or their expertise in a specific field. You only need to look at websites that are genuinely AI-generated to notice the pattern: generic wording, an unclear target audience, zero emotional depth, gradients everywhere, and the same recognizable formatting patterns.

To make our position crystal clear, the founder of the project is a Web Developer and Designer, as well as a UI/UX Graphic Designer. The other members of the team are specialized in different fields that complement the project and bring their own expertise to it.

Thank you for giving us the opportunity to properly explain our vision and approach to AI.
Have a great day!

Aka, 4 paragraphs to say humans wrote the text on the website (Of which has countless errors)

And the entire backend is 100% AI but you avoid talking about it.

Morning, LET!
We've read all your replies carefully, and this will be our last response regarding the concerns raised in this thread.

First of all, Deplo is open-source. We are aware that the public repository has been around for about 2 months, but Deplo itself has been in development for roughly 6 months.
We're also fully aware that, as with any newly launched software, bugs can exist, and that's exaclty why being open-source is a strength: if someone finds a problem, we want to genuinely know about it, in order to fix it as fast as we can.

But, there's an important difference between reporting a real problem and speculating that one exists.
If there is a memory leak, for example, we want to investigate it. But we need to know how Deplo was installed, what happened, how memory usage changed, and ideally have logs or a reproducible scenario.
Without any of that, what exactly are we supposed to test? We cannot debug an issue based on a comment with zero technical information behind it.

The same applies to the "It's made with AI" argument.
A bug does not prove that AI (or a newer model as someone said) wrote the code, nor does AI-assisted development imply that there was no human involvment. Human error exists in every field, including software development, with or without AI.
Also, one of the criticism we found was that our website contains grammatical mistakes. That's a rather strange argument to make immediately after claiming that "basic issues like these" couldn't exist with proper human review. If anything, it demonstrates the rathen than obvious fact that humans make mistakes.

The same goes for "how can one Frontend Developer build a SaaS like this, alone?". Well, nobody said one person built Deplo alone.
We've already explained that there are multiple people behind the project. Again, don't turn assumptions into arguments.

From now on, we are happy to discuss real, reproducible problems and spend our time fixing them. We care far more about fixing a real issue than winning an argument on a forum.
To be clear, we won't continue discussing claims without evidence. If you have facts, we're here. If not, we'll leave it at that.

Taking into consideration that this is a technical forum, we expected more facts and less speculation.

But we all know this is another piece of AI slop.
We all do.
Also, I think deplo is short for deplorable.

johnwinchester

This communuty has become toxic af.
Crusifying the project solely beacuse it's "vibe coded" and there are potentially bugs in it?every software has bugsm the constant state of software is bugs.
If you're so smart you could contribute on github or I don't know may be fork it and fix it?.
Put your "skills" where your mouth is, instead of farming drama here.

forest

@johnwinchester said: This communuty has become toxic af.

This community has always been toxic.

@johnwinchester said: Crusifying the project solely beacuse it's "vibe coded" and there are potentially bugs in it?every software has bugsm the constant state of software is bugs.

Vibe-coding is specifically shitty-low quality shovelware software. Not many people here have an issue with using AI for things like code review and vulnerability research, but "claude, make me an app" is not programming and will always result in shitty code.

@johnwinchester said: If you're so smart you could contribute on github or I don't know may be fork it and fix it?.

Sometimes a codebase is so bad that there's really no point. It's like trying to fix a broken down car with the engine split in two. Eventually you realize that it's a better use of time to not bother and write it off as a loss.

With all that said, I'm speaking in generalities. I haven't looked at Deplo's code and have no idea if it's good or bad.

TheGreatOakley

AI slop

x0x0x

@forest said:

@johnwinchester said: This communuty has become toxic af.

This community has always been toxic.

@johnwinchester said: Crusifying the project solely beacuse it's "vibe coded" and there are potentially bugs in it?every software has bugsm the constant state of software is bugs.

Vibe-coding is specifically shitty-low quality shovelware software. Not many people here have an issue with using AI for things like code review and vulnerability research, but "claude, make me an app" is not programming and will always result in shitty code.

@johnwinchester said: If you're so smart you could contribute on github or I don't know may be fork it and fix it?.

Sometimes a codebase is so bad that there's really no point. It's like trying to fix a broken down car with the engine split in two. Eventually you realize that it's a better use of time to not bother and write it off as a loss.

With all that said, I'm speaking in generalities. I haven't looked at Deplo's code and have no idea if it's good or bad.

I think most people are fine with the "Claude make me an app" part. It's when Claude quietly disappears from the credits and it becomes "We built" and "We developed" that the pitchforks come out

zed

@x0x0x said:

@forest said:

@johnwinchester said: This communuty has become toxic af.

This community has always been toxic.

@johnwinchester said: Crusifying the project solely beacuse it's "vibe coded" and there are potentially bugs in it?every software has bugsm the constant state of software is bugs.

Vibe-coding is specifically shitty-low quality shovelware software. Not many people here have an issue with using AI for things like code review and vulnerability research, but "claude, make me an app" is not programming and will always result in shitty code.

@johnwinchester said: If you're so smart you could contribute on github or I don't know may be fork it and fix it?.

Sometimes a codebase is so bad that there's really no point. It's like trying to fix a broken down car with the engine split in two. Eventually you realize that it's a better use of time to not bother and write it off as a loss.

With all that said, I'm speaking in generalities. I haven't looked at Deplo's code and have no idea if it's good or bad.

I think most people are fine with the "Claude make me an app" part. It's when Claude quietly disappears from the credits and it becomes "We built" and "We developed" that the pitchforks come out

for me it's more about trying to get innocent bystanders to use your dangerous shit.

monsoon73

The whole project feels to me like they looked at successful/similar products and just said "Let's do another one of those" without bringing any new concepts to the table along with the project.

Building your own version of something is a great personal project/learning experience, but not every vibe-coded personal project needs to become a new product.

itzsenu

@deplo said: bugs can exist, and that's exaclty why being open-source is a strength

yes, we know. we love open source too. im also an open source contributor to some well-known projects, but i dont wanna name them cuz im half-anonymous anyway. open source + AI is fine. the issue is how it’s built. the architecture feels unreliable, like a 60s car held together with duct tape.

even Linus uses AI, but he has the knowledge to understand and structure what it produces. here it feels more like AI built the whole thing without a proper architecture behind it.

@deplo said: But, there's an important difference between reporting a real problem and speculating that one exists.
If there is a memory leak, for example, we want to investigate it

i clearly pointed out some issues, and some of them are also making it slow and sloppy. if u really need the memory leaks, find them urself. they’re from a little while ago in the commits. they clearly crash the thing. so either u didnt test it properly, or only did some basic testing. thats what i mean.

@deplo said: The same applies to the "It's made with AI" argument.

ai made isnt the issue. most ppl here have enough knowledge to understand things in seconds, so they see the issues and call them out. im a new member here but i already love this place. if ai is the argument, nope. i use DeepSeek too. i’ve built fully automated code and the architecture entirely with AI using proper directional prompting. otherwise, most of the time u simply wont get the same result.

               ┌─────────────────────────────────────────────────────┐
                    │              GLOBAL CONTROL PLANE                    │
                    │  - Session registry (distributed, replicated)        │
                    │  - Leader election (Raft/Paxos across mother nodes)  │
                    │  - Global health telemetry aggregator                │
                    └───────────────┬───────────────────────────────────────┘
                                    │
          ┌─────────────────────────┼─────────────────────────┐
          │                         │                          │
  ┌───────▼────────┐      ┌────────▼────────┐       ┌─────────▼───────┐
  │ MOTHER-REGION-A │◄────►│ MOTHER-REGION-B │◄─────►│ MOTHER-REGION-C  │  ← MOTHER TIER
  │ (Primary, Asia) │      │ (Primary, EU)   │       │ (Primary, LATAM) │    (multi-region,
  │  + hot standby  │      │  + hot standby  │       │  + hot standby   │     active-active)
  └───────┬─────────┘      └────────┬────────┘       └─────────┬────────┘
          │                         │                          │
    ┌─────┴─────┐             ┌─────┴─────┐              ┌─────┴─────┐
    │  TIER 1   │             │  TIER 1   │              │  TIER 1   │
    │  MESH A   │             │  MESH B   │              │  MESH C   │
    │ Child 1-N │◄──cross─────│ Child 1-N │──cross──────►│ Child 1-N │   (inter-region
    │ (VPS)     │  region     │ (VPS)     │  region      │ (VPS)     │    mesh sync for
    └─────┬─────┘  sync       └─────┬─────┘  sync        └─────┬─────┘    failover only)
          │                         │                          │
    ┌─────▼─────┐             ┌─────▼─────┐              ┌─────▼─────┐
    │  TIER 2   │             │  TIER 2   │              │  TIER 2   │
    │  RELAY A  │◄───────────►│  RELAY B  │◄────────────►│  RELAY C  │   (dedicated,
    │ (Dedicated│   sync      │ (Dedicated│   sync       │ (Dedicated│    active-active
    │  + backup)│             │  + backup)│              │  + backup)│    with failover)
    └─────┬─────┘             └─────┬─────┘              └─────┬─────┘
          │                         │                          │
    ┌─────▼──────┐            ┌─────▼──────┐             ┌─────▼──────┐
    │  TIER 3    │            │  TIER 3    │             │  TIER 3    │
    │  EDGE A1-N │            │  EDGE B1-N │             │  EDGE C1-N │   (geo-nearest
    │  (VPS)     │            │  (VPS)     │             │  (VPS)     │    to end users)
    └─────┬──────┘            └─────┬──────┘             └─────┬──────┘
          │                         │                          │
       Users                     Users                      Users
    (Asia region)              (EU region)               (LATAM region)

@deplo said: We cannot debug an issue based on a comment with zero technical information behind it.

first of all, before u say it, i already mentioned some of the sloppy stuff and clearly asked if u fixed them. if u ask for the issues, i’ll point them out. yeah, it works like duct tape.

@deplo said: Human error exists in every field, including software development, with or without AI.

there’s an interesting thread for u: https://lowendtalk.com/discussion/220939/sshspan-a-foss-ssh-key-manager-with-embedded-ssh-sftp

just read it. at least that guy fixed the issues before replying or ghosting.

@deplo said: Also, one of the criticism we found was that our website contains grammatical mistakes. That's a rather strange argument to make immediately after claiming that "basic issues like these" couldn't exist with proper human review. If anything, it demonstrates the rathen than obvious fact that humans make mistakes.

idk, but it’s completely okay to have grammar mistakes.

@deplo said: "basic issues like these" couldn't exist with proper human review. If anything,

site is okay, but what about the project? simply basics ?

@deplo said: The same goes for "how can one Frontend Developer build a SaaS like this, alone?"

first of all, tell us how a PaaS became a SaaS??????? what kind of thing are u even talking about?

@deplo said: Well, nobody said one person built Deplo alone.
We've already explained that there are multiple people behind the project.

okay, but before typing, u didnt clearly answer the questions i asked earlier. yet u said “i read all.”

@deplo said: Again, don't turn assumptions into arguments.

none of them are arguing unless u make it an argument. this is LowEndTalk.

@deplo said: To be clear, we won't continue discussing claims without evidence. If you have facts, we're here. If not, we'll leave it at that.

thats strange. what if, as u said, a non technical person uses it and says “i installed it but it’s not working now”? he doesnt know how to explain the issue, and somehow you expect the user to debug your own project for you? if u really built the project, investigate it properly urself instead of waiting for users to hand you the exact bug on a plate.

@deplo said: Taking into consideration that this is a technical forum, we expected more facts and less speculation.

idk who u ppl are, but u joined and within seconds showed us an open source project built with AI. so what did u expect?

it clearly has poor AI direction, so how are we supposed to react? again, nobody here is arguing. ppl here care about security first and a lot of other things. u joined, showed us the project, said “welcome, try this” and didnt ask for help finding issues. i installed it the second i saw it cuz i thought it was built by @DeluxHost. anyway,

this slop is a solid 7/10. good vibe coding. slop, but premium slop. anyway

it isnt complete slop, but there’s still a lot of slop in the way it’s built.

@deplo said: A bug does not prove that AI (or a newer model as someone said) wrote the code,

so u didnt clearly read what i said. i said we cant clearly call this fully vibe coded, but it is AI driven without proper human direction, cuz this is a newer model. it isnt only about bugs.

@johnwinchester said: Crusifying the project solely beacuse it's "vibe coded" and there are potentially bugs in it?every software has bugsm the constant state of software is bugs.
If you're so smart you could contribute on github or I don't know may be fork it and fix it?.

vibe coded projects have issues and bugs. even properly human written or AI directed projects have them, but it’s completely different. ppl love to fork and push changes, sure. but why would anyone waste their time on complete slop with fucked up architecture? why?

also, nobody here is trying to prove they’re some 500 IQ tech god. we just point out what’s broken and why it not worth and anything, thats it. nobody knows everything, including u and me. welcome.

forest

@itzsenu said: just read it. at least that guy fixed the issues before replying or ghosting.

I'd trust Deplo a lot more than SSHSpan tbh... That's not a particularly high bar though.

itzsenu

@forest said: I'd trust Deplo a lot more than SSHSpan tbh... That's not a particularly high bar though.

same, it’s much better than SSHSpan. if they handle it correctly, i’ll push some fixes maybe.

forest

@itzsenu said:

@forest said: I'd trust Deplo a lot more than SSHSpan tbh... That's not a particularly high bar though.

same, it’s much better than SSHSpan. if they handle it correctly, i’ll push some fixes maybe.

Honestly I think Deplo has a lot more potential since:

1. It actually does something that some people find useful, and
2. Based on the grammar in many comments, it's clearly not 100% vibe-coded, even if it has issues

CloudHopper

@deplo said:
Taking into consideration that this is a technical forum, we expected more facts and less speculation.

This is a technical forum and yet you assume the people here can't run docker compose up -d so you spent 6 months with Claude developing an app for it... 🙃

itzsenu

@forest said:
Based on the grammar in many comments, it's clearly not 100% vibe-coded, even if it has issues

first of all, they needed to answer correctly, cuz imo they didnt have an idea what they were doing in many places.

yup, it’s not vibe coded. but the architecture isnt reliable. it’s mostly AI driven, with human directed prompting here and there ( in UI/UX ). imo it’s around a 7/10.

LEBUserJoe

@deplo said:

@itzsenu said:

@zed said: posting in a new hot thread

don't think this one's gonna cook.

@zed said:
posting in a new hot thread

@s0n1c said:
Why do I keep seeing hosting providers release vibecoded products that nobody asked for when almost every vibecoded product released by a host gets absolutely shit on 99% of the time? Is it a humiliation kink or something?

@CloudHopper said:

@itzsenu said:
i cant say this is vibe coded. if they used a newer Claude model, there wouldn't necessarily be any obvious trace of it being vibe coded. so there's no way to say for sure. but from what i checked, there are a lot of issues, even with basic stuff. if this was properly human directed, why publish it with a memory leak that's been sitting in the code for so long? clearly wasn't tested properly. and most of the time, AI ends up directing the whole project anyway, cuz current Claude models are good enough to handle it.

It's obviously vibe coded because no human developer would add this many comments to a Dockerfile, and even if they did they wouldn't be full of Claudisms: https://github.com/DeploCloud/deplo/blob/main/Dockerfile

@alincupunct said:

@LEBUserJoe said:

@DeluxHost said:

@averagedatahoarder said:
How is this different from Coolify?

You can check it here: https://deplo.build/comparison/coolify

LOL

"Three things"

1 entirely useless, make permissions PER ACTION LOL useless.
Networkin? Balatant lie works fine on coolify if that's complicated ur brain is smooth
Coolify only needs 1 server for team.

3 things, 1 useless, 2 complete lies! what a great ass platform give ur AI tokens back to society.

And that part of the website is HORRIBLE UX.

@LEBUserJoe said:

@deplo said:

@alincupunct said:

@CloudHopper said:

@DeluxHost said:

On what planet is a solution that deploys docker compose stacks an "alternative to the cloud"?

AI generated slogan.

@itzsenu said:

@alincupunct said: AI generated slogan.

something fully generated daily by an unemployed person. first it looked cool, but it doesn’t anymore. anyway, better than SSHSpan.

Hi! Thank you for such a thorough comment, although we don't find some of its conclusions particularly accurate.

We’d like to take this opportunity to clarify our position regarding AI, because we believe it’s important to set clear boundaries around how we use it.

We believe AI is a great and valuable tool that can certainly be helpful. At the same time, we believe ideas should come from the person first and then be reviewed and refined using the tools available today. Fortunately, AI is still not quite capable of coming up with titles that compelling, which is why I can confirm that the copy for our website was entirely conceptualized by a copywriter and then transcribed and refined in certain parts with the help of AI.

The slogan was entirely conceived by the copywriter in question, with no AI involved in the creative process behind it.

We want to be very clear and transparent about this: we are absolutely PRO-AI, but we believe in using it within reasonable limits. AI cannot currently replace a human being or their expertise in a specific field. You only need to look at websites that are genuinely AI-generated to notice the pattern: generic wording, an unclear target audience, zero emotional depth, gradients everywhere, and the same recognizable formatting patterns.

To make our position crystal clear, the founder of the project is a Web Developer and Designer, as well as a UI/UX Graphic Designer. The other members of the team are specialized in different fields that complement the project and bring their own expertise to it.

Thank you for giving us the opportunity to properly explain our vision and approach to AI.
Have a great day!

Aka, 4 paragraphs to say humans wrote the text on the website (Of which has countless errors)

And the entire backend is 100% AI but you avoid talking about it.

Morning, LET!
We've read all your replies carefully, and this will be our last response regarding the concerns raised in this thread.

First of all, Deplo is open-source. We are aware that the public repository has been around for about 2 months, but Deplo itself has been in development for roughly 6 months.
We're also fully aware that, as with any newly launched software, bugs can exist, and that's exaclty why being open-source is a strength: if someone finds a problem, we want to genuinely know about it, in order to fix it as fast as we can.

But, there's an important difference between reporting a real problem and speculating that one exists.
If there is a memory leak, for example, we want to investigate it. But we need to know how Deplo was installed, what happened, how memory usage changed, and ideally have logs or a reproducible scenario.
Without any of that, what exactly are we supposed to test? We cannot debug an issue based on a comment with zero technical information behind it.

The same applies to the "It's made with AI" argument.
A bug does not prove that AI (or a newer model as someone said) wrote the code, nor does AI-assisted development imply that there was no human involvment. Human error exists in every field, including software development, with or without AI.
Also, one of the criticism we found was that our website contains grammatical mistakes. That's a rather strange argument to make immediately after claiming that "basic issues like these" couldn't exist with proper human review. If anything, it demonstrates the rathen than obvious fact that humans make mistakes.

The same goes for "how can one Frontend Developer build a SaaS like this, alone?". Well, nobody said one person built Deplo alone.
We've already explained that there are multiple people behind the project. Again, don't turn assumptions into arguments.

From now on, we are happy to discuss real, reproducible problems and spend our time fixing them. We care far more about fixing a real issue than winning an argument on a forum.
To be clear, we won't continue discussing claims without evidence. If you have facts, we're here. If not, we'll leave it at that.

Taking into consideration that this is a technical forum, we expected more facts and less speculation.

The community knows the difference between genuinely useful software aided by AI and an unguided, vibe-coded mess.

Looking at the repo, it’s painfully obvious you didn't direct the AI you just blindly trusted whatever it spit out. From the stack choices to the boilerplate comments and basic architecture bugs, these aren't the mistakes of an experienced team using AI as an assistant. They're the result of prompt-and-pray development without basic sanity checks.

Saying "open source means the community helps fix bugs" doesn't excuse shipping an untested, half-baked wrapper and expecting the forum to do your basic QA for free.
Between DeluxHost's already rough reputation on here and this being a zero-effort Coolify knockoff that adds nothing of value, it’s just pure bloat. Don't act surprised that a technical forum can point that out.

Wrote by AI, Everything AI but can't afford a multi-model threat detection before release. Funny.

itzsenu

@LEBUserJoe said: it’s painfully obvious you didn't direct the AI you just blindly trusted whatever it spit out.

@deplo show some screenshots of the prompts ( Not ui/ux, the backend ). not from near day one, previous ones at least 4+.

kait

How many LET users are making the next AI sloppa SSH client

DrNutella

Vibed?