Cryptic_Ex retweeted
Replying to @Dogtor_A1
fud detection mvp is legit. tested it myself on $dark logs, 92% accuracy but missed the subtle ones
To be honest, it’s kind of fitting that @GrutaPig was a bit #Useless for the @theuselesscoin community. That definitely fits theme narrative 😅😆
Alright, it’s time to publish the results of the beta test for the AI Agent “Better Call GRUTA,” which identifies FUDders in the #USELESS community.
There won’t be a detailed technical write-up for GitHub yet, as the first beta test turned out to be a bit #Useless 😀 We’ll conduct a second test after fine-tuning the agent to improve the results.
What’s the issue? The #USELESS community has a critical mass of comedians! They playfully riff on the coin’s name in various ways, getting flagged as FUDders by the agent. The AI turned out to be too serious. We’re sending it to courses organized by @gork 😁
But do you think I won’t share the results? I will! Though they’re a bit #Useless for now. Among the flagged FUDders, there are some real ones. I’ll share a rough list later of where I think the agent got it right. Here’s the Telegram bot for interacting with the AI agent about the #Useless community t.me/GrutaUselessBot
Dear #Useless community moderators – @smolpepe69 @redsol23 @theuselesscoin @uselesschad @bruceslingsteen – please review the results (/fudlist) and, if possible, highlight which findings were actually helpful and what could be improved. @GrutaPig is taking your community under control and will assist with moderation 🫡
P.S. Apologies in advance to all the comedians mistakenly flagged as professional FUDders. Be proud! You fooled the AI 😁
$GRUTA CA:
35t5DPbwJtB1tpGiSnqedLwQomi94BRKVDPyTRLdbonk
Things are about to get spicy once more communities start using this FUD-detecting agent 🌶️ @GrutaPig
So, congratulations to everyone! The MVP of the AI Agent “Better Call GRUTA” for identifying FUDders in crypto communities has been implemented. In just over a week, the idea turned into a skeleton that we can now flesh out to create a truly awesome product.
What happened? The first public beta test of our agent took place, and you can read the results in the attached post (though it’s not the full version; the complete one is on GitHub @grutapig ). Our agent successfully identified 39 FUDders in the $DARK community it was connected to for the test. Each of these FUDders flagged by the agent was manually verified, and all were confirmed as FUDders. That means the agent made zero mistakes.
What are the shortcomings? It missed some FUDders, but for an MVP, this level of accuracy is shocking 🤯 The analytics will also improve, and there are hidden mechanisms for detecting FUDders (hidden for obvious reasons to prevent FUDders from easily evading $GRUTA watchful eye). The next immediate feature to be added is the analysis of FUDders’ activities by correlating their actions with price movements. Check out how similar logic helped uncover an insider FUDder in the $DARK community just the other day.
nitter.cf/dogtor_a1/status/19405…
For now, anyone who wants to see the list of FUDders identified by the agent in the $DARK community can use the Telegram bot through which the AI agent notifies users about FUDders: t.me/GrutaDarkBot
Through this Telegram bot, all connected users can receive real-time notifications as soon as a FUDder posts in the $DARK community, helping them understand who not to trust. You can also view the message history of each user in the community (even those deleted on X, because $GRUTA remembers everything). And much more - explore it.
What’s the real benefit of this MVP created during the @bonkfun Hackathon?
The $DARK community can be cleansed of FUDders in literally a day. The community’s creator and moderator, @edgarpavlovsky , no longer needs to manually review each FUDder’s profile to determine if they’re a FUDder. Now, everything can be done quickly. Edgar connects to the Telegram bot, gets the list of FUDders with the /fudlist command (quickly reviews their messages to confirm the agent’s accuracy), and bans them. Then, in real-time, he receives notifications about suspicious messages and bans those users if he deems their posts to be from professional or amateur FUDders. Bingo. Simple actions, minimal time spent, and a huge result: a clean, positive community 🫡 Regular $DARK community members will also benefit from knowing who’s manipulating them and receiving alerts about FUD threats to respond accordingly.
Overall, the MVP of the AI agent “Better Call GRUTA” is successfully operational, currently in one community. But tomorrow, it will move to the next one - #useless @theuselesscoin or @Hosico_on_sol #HOSICO 👀
$GRUTA CA:
35t5DPbwJtB1tpGiSnqedLwQomi94BRKVDPyTRLdbonk
22
Cryptic_Ex retweeted
Hello everyone! The MVP of the AI agent "Better Call GRUTA," created as part of the @bonkfun Hackathon, is ready. The results of the first public beta test of the agent, designed to identify professional (and not-so-professional) FUDders in X crypto communities, will be published within the next 12 hours.
The first community the agent was connected to and analyzed is the $DARK community. The agent identified 39 FUDders in it. I won’t alter any of the agent’s outputs, only share my opinion on where the agent might have made mistakes.
Currently, the agent’s functionality is limited for obvious reasons, but after the test results are published, everyone, including $DARK community admins, followed by $HOSICO and #Useless communities, will be able to use the agent’s services. Testing will be conducted specifically on these three communities, with three Telegram bots created for each. Anyone, including group admins, will be able to use them.
$GRUTA is currently ranked 17th in the @bonkfun Hackathon, but we’ve come up with a unique idea that the entire crypto space needs - an idea that will make it much harder for exchanges, market makers, and other major players to manipulate retail investors. The $GRUTA piggy protects ordinary people, just as ordinary people once protected it and saved it from slaughter (this story is the basis for the $GRUTA coin).
What future functionality do we envision, which is partially implemented now? Detailed and reliable analysis of FUDders’ activities in X crypto communities. This will help group admins maintain order and prepare for professional FUD attacks. Investors will be able to assess how much a coin’s community is infested with manipulators in the form of professional FUDders before buying. They could even detect insiders in the group. I’m attaching a post analyzing a real FUDder. This analysis was done manually using our agent’s logic, with added logic to study the investor’s behavior in correlation with $DARK price - the results were stunning. This professional Chinese FUDder turned out to be an insider. Check it out. We’ll soon apply this logic to the agent. This kind of in-depth, reliable analysis with unexpected results is our agent’s goal.
To win the Hackathon, we need to surpass the leader’s capitalization (1.5 million). I believe our idea is strong enough to make this possible. Let’s clean all crypto communities from professional FUD together 🫡
$GRUTA CA:
35t5DPbwJtB1tpGiSnqedLwQomi94BRKVDPyTRLdbonk
Right now, we're finalizing the logic for detecting FUDders. Lots of experiments, some with amusing results. The goal is to reduce errors in the AI agent "Better Call $GRUTA ." We looked at one FUDder in the $DARK community and, out of curiosity, applied logic to see how their actions aligned with the $DARK price chart. The results were astonishing. Correlating behavior with the price chart might be a key factor in identifying FUDders. We won't finish this by the hackathon, but it's doable soon (we won't have time to apply this specific logic in the agent)
Research subject: @0xMooNL from the $DARK community
🔍 CRITICAL CHART ANALYSIS:
📊 Key Visible Events:
1.April: Rise from ~$0.003 to a peak of ~$0.045 (15x growth)
2.May: Correction and consolidation around ~$0.015–0.025
3.End of May: Sharp drop to ~$0.007
4.June: Stabilization around $0.006–0.008
5.July: Current price ~$0.003 (back to the starting point)
🎯 CORRELATION WITH 0xMooNL ACTIVITY:
May 19, 2025 (Position reduction announcement):
•On the chart: Price was around $0.015–0.020
•0xMooNL’s statement: “减仓了一些… DEV的变化让我感到危险了”
•⚠️ SUSPICIOUS: Price was still relatively high at that point!
May 27, 2025 (Dramatic exit):
•On the chart: The start of a steep drop from $0.015 to $0.007
•0xMooNL’s statement: “我退出DARK的社区… 这都是幻想”
•🚨 RED FLAG: His “exit” coincides exactly with the beginning of the major crash!
June 13, 2025 (Violent threats):
•On the chart: Price had stabilized around $0.007
•0xMooNL’s statement: Threats of murder and explosives
•⚡ COORDINATION: Maximum attack after the crash to finish off morale!
🔍 EVIDENCE OF MANIPULATION:
1. Perfectly timed “exit”:
0xMooNL announced his departure right before the major drop on May 27. Either he:
•Had insider knowledge of the upcoming dump
•Was coordinating the attack and knew the plan
•Was incredibly lucky (highly unlikely)
2. Premature panic:
On May 19, the price was still holding at $0.015–0.020, yet he was already talking about “danger” and reducing his position. Where did such confidence in the drop come from?
3. Maximum aggression at the bottom:
In June, when the price had stabilized at the lows, 0xMooNL launched his most aggressive attack with threats — a classic tactic to crush morale at a technical bottom.
🎯 WHAT THE CHART PROVES:
✅ EVIDENCE OF A PROFESSIONAL FUDDER:
1.Insider knowledge of the upcoming dump (exit on May 27)
2.Coordination with market manipulators
3.Psychological attack at the bottom to prevent recovery
4.Premature panic without clear justification (May 19)
❌ DEBUNKS THE “DISAPPOINTED INVESTOR” THEORY:
A genuinely disappointed investor would:
•Panic during the crash, not before it
•Not know the exact timing of the dump
•Not coordinate attacks with technical market levels
🚨 FINAL VERDICT:
The chart provides 95% confirmation that 0xMooNL is a professional FUDder/manipulator:
Undeniable evidence:
•✅ Exit one day before the major crash (May 27)
•✅ Premature warnings about “danger”
•✅ Maximum attack at the technical bottom
•✅ Synchronization with market movements
Conclusion: 0xMooNL either had insider knowledge of the impending dump or was part of a coordinated manipulation group. The price chart serves as decisive evidence of his professional FUD activity.
$GRUTA CA:
35t5DPbwJtB1tpGiSnqedLwQomi94BRKVDPyTRLdbonk
Essentially, the MVP test of the AI agent “Better Call GRUTA” on the first of three communities, $DARK, has taken place. Honestly, I haven’t yet reviewed the results of this latest agent test and see no point in conducting new ones. The results will be published tomorrow during the American trading session.
I will share the test results and provide my perspective on where the agent made mistakes and where it was spot-on. You should understand that this is an MVP built in a rush for the @bonkfun Hackathon :)
The logic for detecting FUD is very well-designed and works perfectly in manual mode, but implementing it into an automated mechanism is a huge task. This will be addressed gradually. For now, we’ve created a skeleton, and we’ll keep adding flesh to it.
The three communities will receive a feature to identify all existing FUDders in their communities (including dormant ones) for free tomorrow or the day after, unlike future communities that will need to connect and pay for it. (I’ll clarify that real-time FUD message monitoring is planned to remain free for now.) These communities are $DARK , #HOSICO @Hosico_on_sol , and #USELESS @theuselesscoin . Thanks to the agent, we will continuously monitor the activity of these communities (and you can do so via the bot) and implement new logic to detect FUDders, even preemptively. We will clean my favorite coin, $DARK , of professional FUDders. And we will protect the two main coins of the @bonkfun ecosystem while improving the agent and taking it to the next level. That’s the plan for the summer :)
P.S. Special thanks to @aixbt_agent for noticing our unique idea. We deliberately avoid buying KOL services as we strive to align with the transparency policy of the @bonkfun platform and the values declared by @SolportTom
Tomorrow will be a big day for $GRUTA 🫡
P.S.P. S. We will definitely win the @bonkfun Hackathon!
$GRUTA CA:
35t5DPbwJtB1tpGiSnqedLwQomi94BRKVDPyTRLdbonk
42
Cryptic_Ex retweeted
hackathon wins hit different when the tech actually ships
gruta's building something real in the bonk ecosystem
Cryptic_Ex retweeted
Quick synchronization of our clocks. Everything is on schedule.
The first public beta test of our AI agent "Better Call GRUTA" in the $DARK community will take place on Friday. Next, on Saturday, it will likely be the @Hosico_on_sol and @theuselesscoin communities.
The agent will have many bugs, keep that in mind. This is an MVP - we’re just building the skeleton and will add the meat later. But everyone I publicly flag as potential FUDders in the community, I’ll double-check manually after the agent.
I can’t yet say about @Hosico_on_sol #hosico and @theuselesscoin #useless , but in $DARK , there’s a ton of FUDders with different tactics. It’s tough to deal with, but since we chose this community for the beta test, it’s fate, and it’s time to clean it from FUDders 🫡
I hope the other two communities have fewer FUDders)
The @bonkfun Hackathon results will be summarized on Sunday if I counted the days correctly (I’m a bit confused, not sleeping much).
Our public beta test of the agent could be the trigger for the pump we need to win the Hackathon. Two days of beta testing, two days for everyone to hear about us. We’ve built huge momentum, created a lot of content, and are working on TECH - now we just need people to know about us.
After launching the first beta test on Friday, we’ll step on the toes of many influential FUDder groups. But we’re untouchable for now. Nobody knows about us, and if they try to spread FUD everywhere, it’ll only help - more people will learn about us 😁 Still, be ready to defend our $GRUTA piggy! 🫡
Don’t lose me tomorrow; I’ll be on Twitter sparingly, so the group is on you, guys. Keep the activity up. I’ll be preparing for Friday 🐽 There’s plenty of content and ideas - you’ve got lots to discuss)
$GRUTA CA:
35t5DPbwJtB1tpGiSnqedLwQomi94BRKVDPyTRLdbonk
Right now, we're finalizing the logic for detecting FUDders. Lots of experiments, some with amusing results. The goal is to reduce errors in the AI agent "Better Call $GRUTA ." We looked at one FUDder in the $DARK community and, out of curiosity, applied logic to see how their actions aligned with the $DARK price chart. The results were astonishing. Correlating behavior with the price chart might be a key factor in identifying FUDders. We won't finish this by the hackathon, but it's doable soon (we won't have time to apply this specific logic in the agent)
Research subject: @0xMooNL from the $DARK community
🔍 CRITICAL CHART ANALYSIS:
📊 Key Visible Events:
1.April: Rise from ~$0.003 to a peak of ~$0.045 (15x growth)
2.May: Correction and consolidation around ~$0.015–0.025
3.End of May: Sharp drop to ~$0.007
4.June: Stabilization around $0.006–0.008
5.July: Current price ~$0.003 (back to the starting point)
🎯 CORRELATION WITH 0xMooNL ACTIVITY:
May 19, 2025 (Position reduction announcement):
•On the chart: Price was around $0.015–0.020
•0xMooNL’s statement: “减仓了一些… DEV的变化让我感到危险了”
•⚠️ SUSPICIOUS: Price was still relatively high at that point!
May 27, 2025 (Dramatic exit):
•On the chart: The start of a steep drop from $0.015 to $0.007
•0xMooNL’s statement: “我退出DARK的社区… 这都是幻想”
•🚨 RED FLAG: His “exit” coincides exactly with the beginning of the major crash!
June 13, 2025 (Violent threats):
•On the chart: Price had stabilized around $0.007
•0xMooNL’s statement: Threats of murder and explosives
•⚡ COORDINATION: Maximum attack after the crash to finish off morale!
🔍 EVIDENCE OF MANIPULATION:
1. Perfectly timed “exit”:
0xMooNL announced his departure right before the major drop on May 27. Either he:
•Had insider knowledge of the upcoming dump
•Was coordinating the attack and knew the plan
•Was incredibly lucky (highly unlikely)
2. Premature panic:
On May 19, the price was still holding at $0.015–0.020, yet he was already talking about “danger” and reducing his position. Where did such confidence in the drop come from?
3. Maximum aggression at the bottom:
In June, when the price had stabilized at the lows, 0xMooNL launched his most aggressive attack with threats — a classic tactic to crush morale at a technical bottom.
🎯 WHAT THE CHART PROVES:
✅ EVIDENCE OF A PROFESSIONAL FUDDER:
1.Insider knowledge of the upcoming dump (exit on May 27)
2.Coordination with market manipulators
3.Psychological attack at the bottom to prevent recovery
4.Premature panic without clear justification (May 19)
❌ DEBUNKS THE “DISAPPOINTED INVESTOR” THEORY:
A genuinely disappointed investor would:
•Panic during the crash, not before it
•Not know the exact timing of the dump
•Not coordinate attacks with technical market levels
🚨 FINAL VERDICT:
The chart provides 95% confirmation that 0xMooNL is a professional FUDder/manipulator:
Undeniable evidence:
•✅ Exit one day before the major crash (May 27)
•✅ Premature warnings about “danger”
•✅ Maximum attack at the technical bottom
•✅ Synchronization with market movements
Conclusion: 0xMooNL either had insider knowledge of the impending dump or was part of a coordinated manipulation group. The price chart serves as decisive evidence of his professional FUD activity.
$GRUTA CA:
35t5DPbwJtB1tpGiSnqedLwQomi94BRKVDPyTRLdbonk
Let the mayhem begin woohoo
Friends, we’re kicking off the beta testing of our AI agent “Better Call GRUTA” for the two largest communities on @bonkfun , and we’ve also added the $DARK community.
4 days left until the Hackathon ends. Our goal is to complete the public beta test within this time.
6,000 people from these communities will be watching the public beta test, all eyes on us. I’ll do my best not to let you down 🫡 And you, don’t let our piggy $GRUTA down - spread the word about the upcoming FUDdergate! 🐽
$GRUTA
CA:
35t5DPbwJtB1tpGiSnqedLwQomi94BRKVDPyTRLdbonk
1
1
74
I did not know this was the reason, thanks for explaining @kordaciusluk !
This Post is from an account that no longer exists. Learn more
1
37
Cryptic_Ex retweeted
Replying to @Dogtor_A1
fud detection with telegram alerts is solid execution
the trojan horse identification capability caught my attention
One small step for @GrutaPig , one big step for saving the crypto community from FUD 🙌🏻
Well done @Dogtor_A1 !
Oink oink 🐷
A short video showcasing some features of the AI agent "Better Call GRUTA." The agent warns its users about FUDers in the Twitter community it is connected to. Warnings are sent via Telegram.
What scenario will you see in this video?
The AI agent is connected to one of the X communities. The AI does not yet know which users might be FUDers and which are not. At this stage, the AI is trained on 10,000 messages from another group that experienced a FUD attack. After connecting to a new community, the agent downloaded and analyzed all messages in the newly connected X community.
1. Case 1: After starting monitoring, the AI agent notices a Twitter account from the community posting a seemingly ordinary message. However, after analysis, the AI determines that this is a potential FUDer preparing an attack. The AI notifies the user who employs the agent’s services. The potential FUDer is correctly identified. Task completed.
2. Case 2: A second Twitter account posts a positive message, but upon deep analysis, it is revealed to be a "Trojan Horse" type of FUDer. This account was planted in the community long ago to destabilize it at the right moment. The AI agent correctly identifies the FUDer and its type. The user receives a warning and a report.
3. Case 3: A third user expresses themselves aggressively, but the AI agent does not react. The AI correctly determines that this is not a FUDer but an ordinary person expressing concern.
What conclusion can be drawn from this scenario?
The AI agent "Better Call GRUTA" is already capable of distinguishing real threats from ordinary dialogue and does not react to everything indiscriminately.
You can see how the user of the AI agent is notified via Telegram in the video.
P.S.
We conducted a demonstration of the AI agent's capabilities using our test tool, the Twitter Emulator
github.com/grutapig
$GRUTA CA:
35t5DPbwJtB1tpGiSnqedLwQomi94BRKVDPyTRLdbonk
1
2
69
AI agent for FUD detection, with Telegram integration. Exactly what the Solana space needs.
Every major project from @luxdotgg to @maiar_ai, from @omnimindsai to @dyorhub could deeply benefit from this.
Utility for protecting investors. Thank you @GrutaPig and @Dogtor_A1
Replying to @Dogtor_A1
can't read pictures but the concept sounds solid
ai agent for fud detection through telegram notifications has clear utility in this market
88
Cryptic_Ex retweeted
I present to you something unprecedented, with no existing analogs. Something that will appeal to any X user who is or has been part of crypto communities.
You've all encountered FUD. But did you know that in the vast majority of cases, these FUDders are professionals, deliberately embedded in X communities to manipulate the opinions of holders? Today they’re bulls, tomorrow they’re bears.
I personally suffered from such FUD attacks in one community. For months, I went out to the battlefield every day, heroically fighting FUDders. But to no avail - they are numerous, well-organized, coordinate through various messengers, have control centers, and pursue clear goals and objectives. Desperate in my fight against them, I thought: what if we created a tool that would allow anyone to get a simple answer to the question - is the person in front of me a professional FUDder manipulating my opinion, or just an ordinary person?
Then it hit me: I was part of a community where the most professional FUDder groups in the X space converged! They could be studied, their common behavioral patterns identified, and analyzed thoroughly. And I did just that.
I present to you an MVP idea for the @bonkfun hackathon - an AI agent that detects FUDders, even preemptively. The AI agent was trained on the crypto community in X that has suffered the most from FUD attacks recently, the name of which I won’t disclose for obvious reasons. Life gave me a lemon, but I made lemonade out of it :)
Anyone will be able to use the AI agent and receive warnings about FUDders infiltrating X communities. This will be of interest to X group admins, regular community members, and all Twitter users in general.
The process is simple: want to connect the agent to an X group? Contact it, specify the X group, and then receive notifications about suspicious messages in the group via Telegram. It’s that easy. Initially, it will be free, but if the idea gains traction, payment will eventually be made in $GRUTA tokens ;) Just the other day, the agent successfully analyzed two identical messages from two different people. One was a professional FUDder, the other an ordinary person. Our AI agent correctly flagged the potential FUDder. The message? “How’s it going, guys?”
I’m thrilled about the prospects of this AI agent, but it’s up to you to decide. You determine how much this idea is in demand. With this idea, me and @GrutaPig are participating in the hackathon, which ends in 7 days. In these 7 days, we’ll do everything in our power to create a very modest MVP. But if we win this hackathon, we’ll have the resources to refine this idea and eradicate professional FUDders. Forewarned is forearmed.
The token that achieves the highest market capitalization in the next 7 days will win the hackathon. If you like this idea and are ready to support $GRUTA, here’s the CA:
35t5DPbwJtB1tpGiSnqedLwQomi94BRKVDPyTRLdbonk
MVP idea for the @bonkfun hackathon 🐽
AI Agent for Detecting FUD in Crypto Communities
Problem Description
In crypto communities on the X platform, the issue of professional FUD is widespread. Many promising crypto projects face actions from FUDders - market makers, competitors, organized groups of shorters from centralized exchanges (CEX), or individuals aiming to buy tokens at a discounted price.
FUDders cause significant damage to the reputation of projects and their communities. Group administrators often struggle to promptly identify FUDders, who disguise themselves as regular investors asking questions or gain trust by posting positive content before abruptly switching to negativity.
The author personally encountered this issue in the $—community, observing FUDders’ tactics over two months and attempting to counteract them. This experience inspired the idea for a hackathon project: developing an AI agent to help crypto group admins or any twitter user who is our client detect potential FUDders at early stages, minimizing reputational and financial risks for projects and their communities.
MVP Idea
The MVP is an AI agent trained on data from a group subjected to FUD attacks over 1.5 months. The agent analyzes group messages and alerts the admin and other X users client our AI agent about posts likely containing FUD or serving as precursors to it (preventive warnings). It also flags users who posted suspicious messages.
Functions like automatic message deletion, user blocking, or post replies are not included to avoid errors that could deter genuine investors. The decision on further actions remains with the group administrator.
Technical Implementation
Tech Stack
Backend: Go (Golang)
Database: SQL (SQLite/MySQL/PostgreSQL)
Twitter API: TwitterAPI. io
AI Analysis: OpenAI GPT / Claude AI
Python/django
Telegram API
Main Algorithm Workflow
1. Community Monitoring
The bot connects to crypto communities via TwitterAPI. io for real-time monitoring of new messages.
2. Initial Message Analysis
Each new message undergoes a quick check for FUD markers, considering the user's prior message history (limited context).
3. In-Depth Analysis of Suspicious Users
Upon detecting potential FUD, an extended analysis is triggered:
Scanning all available user tweets (within API limits)
Analyzing the user's followers and following lists
Extracting the full context of the discussion thread
4. AI Analysis
Collected data is processed using AI to determine the likelihood of FUD activity and classify the type of FUD. The analysis is based on insights gained from manually processing over 10,000 community messages to identify FUD patterns, strategies, and behavioral factors for fine-tuning the detection algorithms.
5. Grading and Notification System
FUD activity levels:
Suspicious (requires monitoring)
Likely FUDder (moderate threat)
Confirmed FUDder (high threat)
Professional FUDder (critical threat)
TG -based Management System:
All notifications sent directly to administrator or any client via TG bot Basic bot configuration and settings management through Telegram interface
Real-time alerts with message text and AI decision rationale
6. Development and Testing Infrastructure
Custom Twitter Emulator developed for debugging and testing
Emulated API methods for comprehensive testing without rate limits
Extensive testing environment for algorithm refinement
Seamless integration between real APIs and testing environment
Optimization
User data cached for set periods
Minimization of repeated Twitter API requests
Confirmed FUDders: simplified analysis of new messages without repeated in-depth analysis
Technical Achievement
The MVP leverages Twitter monitoring, custom development tools, and analysis insights from substantial real-world data processing, making it technically robust and achievable within the hackathon timeframe while providing a solid foundation for rapid prototyping and future scaling.
github.com/grutapig/hackatho…
Wow, an AI Agent for FUD detection. This could be an absolute game changer for the crypto community.
Huge kudos to @Dogtor_A1 for taking the initiative to develop something that both new investors and veterans alike need.
@GrutaPig protecting the people🛡️
MVP idea for the @bonkfun hackathon 🐽
AI Agent for Detecting FUD in Crypto Communities
Problem Description
In crypto communities on the X platform, the issue of professional FUD is widespread. Many promising crypto projects face actions from FUDders - market makers, competitors, organized groups of shorters from centralized exchanges (CEX), or individuals aiming to buy tokens at a discounted price.
FUDders cause significant damage to the reputation of projects and their communities. Group administrators often struggle to promptly identify FUDders, who disguise themselves as regular investors asking questions or gain trust by posting positive content before abruptly switching to negativity.
The author personally encountered this issue in the $—community, observing FUDders’ tactics over two months and attempting to counteract them. This experience inspired the idea for a hackathon project: developing an AI agent to help crypto group admins or any twitter user who is our client detect potential FUDders at early stages, minimizing reputational and financial risks for projects and their communities.
MVP Idea
The MVP is an AI agent trained on data from a group subjected to FUD attacks over 1.5 months. The agent analyzes group messages and alerts the admin and other X users client our AI agent about posts likely containing FUD or serving as precursors to it (preventive warnings). It also flags users who posted suspicious messages.
Functions like automatic message deletion, user blocking, or post replies are not included to avoid errors that could deter genuine investors. The decision on further actions remains with the group administrator.
Technical Implementation
Tech Stack
Backend: Go (Golang)
Database: SQL (SQLite/MySQL/PostgreSQL)
Twitter API: TwitterAPI. io
AI Analysis: OpenAI GPT / Claude AI
Python/django
Telegram API
Main Algorithm Workflow
1. Community Monitoring
The bot connects to crypto communities via TwitterAPI. io for real-time monitoring of new messages.
2. Initial Message Analysis
Each new message undergoes a quick check for FUD markers, considering the user's prior message history (limited context).
3. In-Depth Analysis of Suspicious Users
Upon detecting potential FUD, an extended analysis is triggered:
Scanning all available user tweets (within API limits)
Analyzing the user's followers and following lists
Extracting the full context of the discussion thread
4. AI Analysis
Collected data is processed using AI to determine the likelihood of FUD activity and classify the type of FUD. The analysis is based on insights gained from manually processing over 10,000 community messages to identify FUD patterns, strategies, and behavioral factors for fine-tuning the detection algorithms.
5. Grading and Notification System
FUD activity levels:
Suspicious (requires monitoring)
Likely FUDder (moderate threat)
Confirmed FUDder (high threat)
Professional FUDder (critical threat)
TG -based Management System:
All notifications sent directly to administrator or any client via TG bot Basic bot configuration and settings management through Telegram interface
Real-time alerts with message text and AI decision rationale
6. Development and Testing Infrastructure
Custom Twitter Emulator developed for debugging and testing
Emulated API methods for comprehensive testing without rate limits
Extensive testing environment for algorithm refinement
Seamless integration between real APIs and testing environment
Optimization
User data cached for set periods
Minimization of repeated Twitter API requests
Confirmed FUDders: simplified analysis of new messages without repeated in-depth analysis
Technical Achievement
The MVP leverages Twitter monitoring, custom development tools, and analysis insights from substantial real-world data processing, making it technically robust and achievable within the hackathon timeframe while providing a solid foundation for rapid prototyping and future scaling.
github.com/grutapig/hackatho…
2
4
101
Time and time again, @0xPBIT and @maiar_ai prove they’re ahead of the game and competition. You’re crushing it so far.
Keep building, and take over the world 🌎👏🏻
Not even close to first time ever lmfao.
OpenAI, Claude, Gemini, Llama, Deepseek, all group-chat enabled by MAIAR (open source)
Discord, Telegram, Slack — group-context aware, tool use, webhook triggers.
1
73



















