Add-on Gonna vibe code my first XF addon called "Sutradhar"

N

new_view

Active member
Licensed customer
I have Codex and Claude 20x plus Gemini, too many tokens to burn, but I don't know where to start. I am stubborn, I guess; there is only one way to find out. My other vibe coding project is a desktop app coming in 3 years, cause like, it's too big.

Nonetheless, what skills should I set in my repo: https://github.com/thulaen/sutradhar

It's MIT-licensed, by the way; those who wanna help can help. Look, Linus said vibe coding is okay, so this should be a walk in the park, right? I mean, what could go wrong?
 
Last edited:
Interesting that you have already a name for the add on but absolutely no description what the add on shall do or be good for. :-)

my first XF addon called "ZEN-G"
Click to expand...

You could change that to "Gen-Z" - would be interested to see what it would do then. :D
 
I have a Dell machine, downloaded XenForo, and I am now setting up the dev things.

The first thing would be laying down the law so that agents don't drift.
 
A name for an add-on.

Sarcastic Sarcasm GIF by MOODMAN
 
Noble™ said:
So, do we know what the add-on is supposed to do?
Click to expand...

enterprise seo framework with MCP support and other Reddit-like features will use PHP, Python (can't do AI stuff or NLP without Python), and Marimo notebooks for labs, i.e., QA.

Internal linking works like this: the system finds links and stores them as suggestions. They appear on a review page in the XenForo admin area, showing the post, the phrase to link, and the target thread. A moderator clicks accept, and only then is the link inserted and live. Reject hides it. That is the default and only flow for v1; no editor suggestions for authors, and no auto mode. This is a simplified explanation; there is also other stuff like machine learning for moderation, MCP topic hubs, etc.
 
new_view said:
Internal linking works like this
Click to expand...
Did you give up on this:
N

Thread 'XF Internal Linker V2 - The Vibe Coding journey begins!'

I will come with results and keep you updated, pinky promise. So far, I have a plan about XenForo SEO internal linking Web App Version 2 that runs locally (not that I made it myself; Claude Opus and Gemini Pro made it to help a noob, my goodness, such kindness; let's see how it goes! My current AI stack is Claude Code, Codex, and Google Antigravity.

I am just a noob who has no idea what I am doing. It's all about the vibes, I guess. I am a dog chasing cars. I just do things. 😂:ROFLMAO:


To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.


Also, Claude Opus suggested we use PageRank (a paper published in 1998 by Sergey...
  • Like
 
GitHub is intimidating, guess what? Today I learned about AlphaEvolve; I am gonna vibe code GanitEvolve.

This will do what Google's AlphaEvolve does, but with Claude, Codex, and Gemini Subscriptions.

It will use Rust, Julia, and Fortran. Yes, Fortran - legends never die, never... You thought it was over; I tell you, pal, I am not done yet with my Frankenstack: COBOL, TLA+, SQL / PL/pgSQL, Lean 4
 
Is this a good idea:
  • 1. Sutradhar's engine never runs on the server that hosts a forum: not on shared hosting, not on a VPS, and not as an option. Remove every install path that puts the engine there.
  • 2. The engine runs only on the owner's own computer (Windows 10 or 11, or desktop Linux), managed by one desktop app, the Sutradhar app, built with Tauri 2.
  • 3. The forum server gets only the thin XenForo add-on or WordPress plugin. It never starts work by itself, and it never makes outgoing network requests.
  • 4. Heavy batch work may optionally go to cloud helpers in the owner's own AWS or Google Cloud account, as already designed. The helpers never hold the database or keys, and they are off by default.
  • 5. Supersede, with new ADRs, every decision that assumed the engine runs on a server: the one-line server installer, cloud-init installs, server-side sutra install, server Docker footprint and ledger rules, server-side updates and repair, the "where should the engine run?" choice, and the "rent a server" option. List each superseded ADR for me in plain English, with what replaces it. Keep everything that still applies, such as transactional steps, the ledger, doctor, repair, the pause switch, and signatures, and move it to the owner's computer.
 
new_view said:
Bro, how do people vibe code stuff over the weekend? Am I doing it wrong?
Click to expand...
It's not a vibe coding thing, its a coding thing. Understand a worktree and how to maximise it, so you can have multiple coding in the same file happening without clashing. The more agents you have working, the more gets done. If you want a slow pipeline, use github, otherwise, look at putting your own opensource system onto a local box so you have lots of cores to speed things up without paying third parties.

You also need to understand, AI is a tool, it is equal to the best coders in the world only if you know how to use it, and then you still have to go through all the security and round robin fault finding, then penetration testing.

I have told you before, which I will take a guess happened, you listened to AI and it took you down rabbit holes and you didn't get what you actually wanted. You got this bloated goat that the AI talked you into.

You have to spend more time researching and planning before you code anything. You need a clear map, start to finish, goals, and don't deviate. You can add and remove to software when you complete your map, but if you constantly modify it along the way, you get lost in code.
 
new_view said:
Bro, how do people vibe code stuff over the weekend? Am I doing it wrong?
Click to expand...
I will give an example of something I have running right now. Codex and Claude Code. All AI have strengths and weaknesses, like Perplexity is the best AI for research, Claude is the best at coding, etc etc. GPT is extremely good at rationalising things, looking outside the box and not getting hyper focused on one path. So I use codex and link it with claude code via a relay, so both are talking to each other in a chat file. Both check it every couple of minutes, odd and even assigned, talking to each other to discuss ideas, help code get a little more funky with its thinking, whilst codex is running k6 from another server. One is tuning, one is testing, one is coding, one is offering ideas.

I am not needed in the process. A goal is set to tune x to under x ms p95 and p99 scores, zero errors. I can eat, sleep, go out, walk the dog, do whatever the hell I want, they have their tasks and a set goal to try and achieve, not forced achievement, but attempt. They can research and learn from the web, access docs, etc, everything they need to try x, y, z and every combination of it can run, as two things may help performance, but together they make it worse.

This is how you have AI achieve things for you, let alone my above comments about just coding something specifically... that is with lots of agents all working together in the same tree, same files, yet not screw each other over and write over another's code.

Just think out of the box... hell, ask AI what it can come up with to help you code faster. It will help you.
 
You must log in or register to reply here.
Back
Top Bottom