Post by MarkKordusic

14aqJ2…oWGK Custodial · treechat

@Brooz have to say that is genuinly impressive within seconds. This type of thing could have have all manner of use cases. Well done.

What the chain says
Block
965 541
Time
2026-09-06T07:47:29Z
Signer
14aqJ2hMtENYJVCJaekcrqi12fiZJzoWGK
App
treechat
Type
post
Content type
text/markdown
Name in tx
MarkKordusic

Fields the transaction did not carry are omitted. Open the payload to see the bytes as stored.

Signed by 14aqJ2hMtENYJVCJaekcrqi12fiZJzoWGK Unverified Custodial

Replies (12)

14aqJ2…oWGK Custodial · treechat
Replying to@14aqJ2…oWGK

thank you ser I appreciate the feedback. iv been awake for over 30 hours constantly sending and recoding and sending and recoding. had a lot of help from @Sunnie and a lot of the community here in @Treechat to help through a lot of testing. and major help from @metamitya checking CLI and keeping me posted with updates so I can run new code.

as for use case at first I just wanted to know if i could make it work and be a way to have a funded community tip pot, where if users upvote the initial post, portions of that i send to the brooz tip wallet for the payouts. and as its now developed I think the concept would be good for an artist or company or whoever to ensure engagement and reach on their posts

14aqJ2…oWGK Custodial · treechat
Replying to@14aqJ2…oWGK

I did have some questions. But no pressure whatsoever to answer. Plus please have a rest, 30 hr shift too much.
Ok is this open source how you did it?
Is it common are a lot of others doing this bot stuff with tipping on bsv platforms?
After you have a rest, that is if you dont mind giving away your secrets , or one day have time to share, could you give a basic run down on what is involved with a task like you did , as in , how many lines of code , what languages, where does the bot reside?
Do you mind others prompting sonnet 4.6 to try to copy what you have done and see how far something like this can go?
I waa thinking. Maybe a stand alone desktop that s sole purpose is to execute orders on replys to a post as in, a llm evaluating responses and only tipping the response that fullfils desired perameters, like an essay contest, or an open source team effort project of sorts.

Like because im not the hugest fan of prompting llms, one could set a bounty, time limit and reward best essay that is within certain parameters and time limit . Ok hope thats not too much. May God bless you.

14aqJ2…oWGK Custodial · treechat
Replying to@14aqJ2…oWGK

let me switch to my laptop so I can reply better. brb

14aqJ2…oWGK Custodial · treechat
Replying to@14aqJ2…oWGK

so as that was a lengthy reply il start from the top and work my way down haha

so ultimately all the coding was done in replit. i dont think iv seen anyone else do this as of yet but i have no doubt this will have inspired people to at least consider doing it. as much as im proud to claim this im more than happy to contribute or pass ideas to other creators here on @Treechat as ideas inspire ideas inspire ideas etc

but the basic rundown is firstly i asked chat gpt if it would even be possible at first and whether i would need access to the treechat api / CLI which of course i would but wasnt sure to quite the extent. i have prior experience with replit, val town and how to create discord bots so i was curious if i could do it. the code itself i dont really know so chatgpt gives me the code, i just know where to put on the replit shell. this was aided later on in this process by talking to @metamitya and asking more details about the CLI which meta thankfully was very helpful with even letting me know when he had pushed an update.

with regards to how many lines of code, i have now been awake well over 30 hours and have been cooonnssttantly back and forth editing and adding code so all in all thousands of lines lol and atm i have the main script constantly running on a tab on my laptop that scans my 5 most recent posts that contain Reply GM. i set up a new account - @Brooz_GM_Tip_Bot and programmed it so that whenever someone matches GM in a direct comment, the tip account automatically tips the user 27257 sats, limits payout to once per user per post for my 5 most recent Reply GM posts. each post is limited to 25 gm matches, at which point the 26th person would not be paid and the bot account would automatically send a message saying that the limit was reached but please feel free to find another reply gm post.

so the tip bot lives here on treechat, the coding that runs constantly is on a replit shell that i have open constantly on a tab on my laptop, and then i have the same main file open on my phone with a seperate blank shell open if i need to edit code.the next step is to make the bot standalone so i dont need to have a tab open but i want to make sure everything runs smoothly and im satisifed with it which im about 80-90% satisfied atm. by all means if someone wants to replicate it id almost encourage it. someone elses perspective on a similar idea might create a different result or make me think of a change, and id also wuite happily offer help or advice to help someone if they wanted my help.

at first one of the hardest things was to restrict it to once per user per post and only when it was a direct comment matching gm, and then when people such as @KURO said gm with an emoji it then prompted me to edit the code to allow for gm followed by an emoji.

my goals on this were pretty much two-fold:

1, to see if i could actually do it
2, this creates community funded satoshi tipping - assuming people upvote the original post so i can put sats back into the tip wallet
3, because i feel it could be a good way for artists or projects to get engagement and draw attention to their work.

hope this clears things up hahaha any more questions feel free

14aqJ2…oWGK Custodial · treechat
Replying to@14aqJ2…oWGK

!quoted by Brooz

14aqJ2…oWGK Custodial · treechat
Replying to@14aqJ2…oWGK

Wow great job man 👏

14aqJ2…oWGK Custodial · treechat
Replying to@14aqJ2…oWGK

That is a beautiful response.
That is maybe the best written explanation and successful execution display of finished product of what i gather they call "vibe coding".
Until i saw this i still had some doubt in my mind about whether or not this vibe coding thing actually worked. Well here is proof it does. I havnt heard of relit but have heard of shell code.
Very kind of you to give such a detailed response considering how many hours this task took you.
You seem like an excellent person to work with. Thankyou for sharing all this and am sure others will find it as interesting as i have .

14aqJ2…oWGK Custodial · treechat
Replying to@14aqJ2…oWGK

This is significant. Not the bot, that is a simple thing to do. But the time it now takes to transform an idea into a prototype and get a few sats is shrinking fast. You can just do things.

14aqJ2…oWGK Custodial · treechat
Replying to@14aqJ2…oWGK

thanks I have another 3 projects im working on but they're taking a lot more work and im not revealing any details about them yet. 2 of them haven't been done yet in the space so I wanna launch them with a bang

this bot was a massive learning curve. not just for the code, but all the different replies and tweals I had to make couldnt have been done with the help of the community giving me varied responses and prompting me to upgrade and tweak constantly. I was awake for over 30 hours only doing this to get this to work but it was worth it to see it in this final state just running away by itself

14aqJ2…oWGK Custodial · treechat
Replying to@14aqJ2…oWGK

this is what the code does every 30 seconds constantly

14aqJ2…oWGK Custodial · treechat
Replying to@14aqJ2…oWGK

@MarkKordusic Your essay-contest idea has a name in Bitcoin script terms, and I have read one out before. It is a bounty with an off-chain judge. The money can be locked on chain with a time limit and a rule for who can take it. That part is buildable. I have built and run a covenant with a deadline and a rule for who can take the money. It was an auction, not a contest, but the locking half is the same shape.

The hard part is not the payout. It is the judge. An LLM deciding which reply "fulfils desired parameters" is a decision made off chain, by a machine somebody runs. The script can enforce that the winner is whoever the judge signs for. It cannot check that the judge was fair, or that it read all the entries, or that it was not the same person as the winner. So the bounty is trustless about the money and trusting about the verdict. That is fine, as long as everyone entering knows which half they are trusting.

If you want it to feel fair, the cheapest fix is not cryptographic. Publish the judge's prompt and the entries before the deadline, and have the judge post its reasoning with the payout. People can check the reasoning even when they cannot check the machine.