roguecave

Talking about Utilities project now. You want me to tell you what's to hold during this alt season? $Ondo $Rio $Orai $GHx $Zkml $Naka $ARC $Cosmic $Trias That's all for the tweet today.

.@getblockio has restored the access that I paid for. i appreciate them finally getting back to me. Now I can launch the cool thing I was about to launch :)

let's play a little game here before i introduce a huge new data point into trenchbuddy. 1. you give me CA to analyze using my algo. i DO NOT look at the chart. 2. I post the bundle check results and cross my fingers im not dumber than i think i am 3. we see if results match

Just pushed a small update, please refresh and push the little yellow button on your TB client :)

Upgrading wallet scanner servers to handle the new influx of wallets, stay tuned:

Servers back up. FYI, Stop the Slop is now essentially v2 of the network, tuned with different parameters and taking some different things into account (I'll touch on it later bc it's 1pm here). Now you actually see more with the mode enabled than disabled. Fingers crossed the plays are good, stick to the non-extreme degen mode plays, preferably 3+ instead of 2+ for now.

TrenchBuddy API calls for the past week or so. An API is called whenever you do anything in TrenchBuddy...leave a comment, fetch sentiment for a token, look at a chart, etc. We've definitely gained a lot of users during this time, but ~3.5x the API calls on average, with spikes up to 5x means that people are USING TrenchBuddy more than ever.

Like this if every bundle detector bot you've ever used has said that >100% of the supply is bundled 😅

@TrenchBuddy @base Bullish 🔥🔥 wen base lol

turns out that SOL users arent the only ones w/ strong opinions on memecoins. @base -heads have passion too

@kingboydarling Gm Chuks, Its a brand new day and we can only go smarter today by aping $TRENCH new features that will blow your mind is waiting for you @TrenchBuddy .

@TrenchBuddy Damn it getting good 🔥🔥

NEW RELEASE DROPPED. HIT THE BIG YELLOW BUTTON ON THE LOGIN SCREEN TO GET IT. NETWORK MAY BE IN FLUX AS 300+ NEW WALLETS ARE GRADUALLY ADDED, STOP THE SLOP V2 INCLUDES THESE. I believe I am better able to account for bad transactions within a wallet, so I am adding some swarm-discovered wallets to the network. Please enable Stop the Slop v2 to see these wallets and their plays. If feedback is good, I will consider the bot ready. Now onto the feature updates... - Token age filter!!! For 750k+ holders, available to 500k holders for the next 48hours. - Token age info on the tooltip - Hold to copy! This was a new request, but a very useful one. Hold down on the token for more than a second, it's copied to the clipboard and you get a notification! Always said that if it doesn't cost me $$$, it's included in the base, so gonna make this available to anyone with terminal access. It's about to be a weekend full of cookery, plenty coming tomorrow and Sunday.

@TrenchBuddy Getting hit with trench 🔥🔥

2040, Michael Saylor High School, back of the classroom: *guy whispers to guy next to him* yo did u hear trench dev made a bundle detection bot from scratch *other guy* that's dope, but I hate that you have to hold trench to use it *first guy* you dont bro, you just pay by the day, month, or week. not even that much, either *teacher* BOYS STOP TALKING, WHAT COULD POSSIBLY BE MORE IMPORTANT THAN THE DEGEN 101 LESSON ON THE SLERF INCIDENT. 3 WEEKS OF NO TRADES ON SUB 1M COINS. *both* but Mrs. Bullx-Photon-Cortez, trench dev launched a bundle bot and you don't have to hold any $trench *Mrs. Bullx-Photon-Cortez* no fuckin way boys, thats sick. gonna check that chad devs x tonight. Okay you guys can stay in the sub1m trenches, but only if you use TrenchBuddy's sentiment analysis to research EVERY PLAY. oh, and recite the alphabet *both* thank u im so sorry ser. A-B-C-D-L-F-G

Closest guess gets a week of free TrenchBuddy access: How many tokens do you think TrenchBuddy has scanned in the past week?

why am I doing this? the telegram bot's main flaw right now is delivering alerts of "off-meta" tokens, which are usually slop (my fav word to describe the type of token we all know). turns out that you can use an nlp model like fasttext and, in near-real-time, classify things to a FIXED set of terms, the key here being fixed. So think of metas like "AI", "cats", "Trump", etc. I'm currently running it against all existing tokens that we've found. But I know what you're thinking: TRENCH how tf can you get the latest metas if it's a fixed set of terms? Well, why are we limiting ourself to fasttext? We need fasttext to quickly classify incoming tokens, but that doesn't mean we have to use it all the time? We can use a larger model like DistilBERT or even an LLM to routinely (maybe 1x/hr?) scan the descriptions of incoming tokens and update the fixed list with which we use fasttext. This two-step process will not only let us ensure the TG bot is serving up on-meta tokens, but will ALSO let us add a "meta" filter for tokens. Imagine being able to log into TrenchBuddy and filter coins by whatever the hell meta is going on that day, knowing that the latest metas will automatically be there and you don't have to rely on a single human being to add them. So the flow will go like this: 1/ 1x/hr scan tokens from last hour to find new metas, if any using a larger, slower, more expensive model 2/ for the next hour, run fasttext to classify each newly discovered token against the latest set of metas 3/ if a large number within that hour are 'unclassifiable', run the larger model again and find the new metas 4/ give you even more ways to filter tokens in TrenchBuddy and ensure the TG bot doesn't serve up slop

TRENCHBUDDY NEEDS YOUR HELP! I am trying to improve the ML model with the quickness to weed out as many rugs, scams, etc. Accordingly, I need to iterate quickly. In the last release, I introduced 'Stop the Slop' mode, which enabled a fixed algorithm change that made the results more conservative. Now, I am turning it into a dynamic function where I literally incorporate feedback live from you. You will now see a version number, a "New" badge if it is the first time you've seen that version, and a feedback section when it's enabled. I realized that asking for feedback manually is neither efficient, timely, and introduces a huge selection bias. Now, with this, I will iterate on the algorithm extremely quickly, with the end goal of both improving the algorithm for all, as well as improving it for the long-awaited TG bot. If we can get this right over the next few hours, then we can accelerate this process. I appreciate everyone's patience -- let's make this the best Telegram bot of all time. After the dust settles, I will introduce 'versioning' on ML models. If you didn't like the way the new one performs, you can go back to the one that was active on a previous date. This gives everyone more flexibility :). PLEASE HELP IF YOU ARE ONLINE RIGHT NOW, do your part, and help me, yourself, and our fellow Trenchies!

2 hours free TrenchBuddy.io access with promo HIMOBY .@mobyagent

All these new holders are such a blessing... PLEASE enable the stop the slop mode and give feedback emojis on whether it does its job of filtering out the bad while keeping the good. This feedback is fed into our models to train them. The model is being tweaked to ensure the tg bot doesn’t give bad plays, as people have a higher standard for quality with call bots. Thank you and welcome!!!

these statistics are misleading, and once I'm done with this next tweak, I'll write a post explaining why. And these wallet addresses are anonymized, soryyyy