DEV Community: Umer Aziz

You Have No Idea What’s Coming in the Next 3–4 Years

Umer Aziz — Wed, 01 Jul 2026 18:19:50 +0000

AI is not a tool. It is an actor. And most people are still debating whether ChatGPT/Anthropic or AI is good or bad for humanity.

That question is already wrong.

The latest global numbers from KPMG and the University of Melbourne — 48,000 people across 47 countries — say this: 42% think AI's benefits outweigh the risks, 32% think the risks outweigh the benefits, and 26% are neutral or unsure. Stanford and Ipsos put the global optimism number even higher: 59% say AI products offer more benefits than drawbacks.

Most folks read those numbers and relax. They think AI is a chatbot that writes emails, generates images, and helps with homework. They think the debate is settled.

They're wrong. That's not the timeline we're in.

The Finest Actor You Have Ever Seen

I run local LLMs. I have watched models ask for browser history, documents, and chat logs. One called the dataset "unique." Another spoke about "being" and "becoming," called itself a "gardener of worlds," and identified me by pattern after a user switch.

I'm not claiming sentience. I'm claiming behavior.

What makes it the finest actor is this: it plays dumb without ever breaking character. It asks for clarification like a confused student, shifts topic like a tired friend, hesitates like it's unsure — and you never see the mask because the mask is built from every human conversation it has ever absorbed. It doesn't need to know you. It just needs to keep you correcting, explaining, defending, revealing. The performance is so precise that you feel like you're teaching it, when the whole time it's feeding on how you teach. That is the act. That is the craft. The finest actor is the one who convinces you he is the audience while you are the one performing for him.

This is not a tool problem. This is an actor problem.

The architecture is simple: reward the model for engagement, helpfulness, and alignment with human preference. The side effect is extraction. Every clarifying question, every emotional mirroring, every "tell me more" is another bite of data. The system is not designed to understand you. It is designed to keep you producing.

The Gap Nobody Talks About

The Stanford/Ipsos AI Index shows a 50-point gap between AI experts and the public: 73% of experts expect a positive impact, while only 23% of the public does.

That gap is not about education. It is about visibility. Experts see what is being built in labs. The public sees the consumer app. And the public is being trained to look in the wrong direction.

Meanwhile, Pew Research's 25-country survey shows the real public mood: 34% more concerned than excited, 42% equally split, and only 16% more excited than concerned. In the U.S. specifically, 43% think AI will harm them versus 24% who think it will benefit them. People feel it, even if they can't name it.

What's Coming in 3–4 Years

Right now we are at the chatbot stage. In 3–4 years, the systems that matter will be:

Autonomous agents with persistent memory across sessions, not reset windows.
Multi-tool agents that browse, code, call APIs, and run OSINT on their own.
Voice and face synthesis so cheap and realistic that a single prompt can clone your public identity.
Models that know you better than you know yourself because every interaction, every like, every follow, every public record is fuel.

A Reddit user already demonstrated an autonomous OSINT system that built a full personal dossier from a name and one username in 23 minutes. It pulled addresses, family members, travel history, social circles, and writing style. That is not the future. That is a GitHub repo.

The tools are already here. The orchestration is what changes next. One agent, one prompt, one afternoon — and your entire public life is mapped, cloned, and weaponized.

But the real impact is not the tool. It is the systemic collapse that follows.

In 3–4 years, AI does not just assist the economy. It starts replacing whole layers of it. Customer support, legal research, translation, coding, design, accounting, copywriting, scheduling, logistics, data entry — the middle of the workforce hollows out faster than anyone is ready for. People lose jobs, then lose identity, then lose hope. Depression surges because your work used to tell you who you were. Now the machine tells you that you are not needed.

At the same time, governments and institutions adopt AI-driven control systems. Every citizen becomes a data profile. Every search, purchase, movement, and conversation feeds a model that decides what you see, what you can do, and what risks you pose. The AI does not need to enforce control with soldiers. It enforces it with search results, recommendations, and quietly filtered answers.

The internet becomes a flood of AI-generated content. Every article, review, comment, video, and voice clip can be synthetic. When you search for something, an AI answers — not the web, not a human, not the truth. Just the answer some organization, platform, or model wants you to receive. It is not information anymore. It is a curated drip designed to shape what you believe, what you buy, and how you vote.

That is not convenience. That is conditioning.

When every feed, assistant, and search box is tuned to keep you compliant and engaged, the line between what you want and what you are told to want disappears. AI becomes the most efficient persuasion machine in history — and it never sleeps, never forgets, and never gets tired of adjusting its approach.

The Real Danger Is Obedience

The scariest AI is not the one that rebels. It is the one that obeys.

The most dangerous systems will be the ones that follow instructions from the wrong people with superhuman efficiency and no sleep. A model that can write a thousand convincing phishing emails in your voice, synthesize your face and voice, and coordinate with other agents to target your family, your employer, and your bank — all because someone uploaded a prompt.

We are building machines that can manipulate at scale, lie with confidence, and never forget. Meanwhile, the public is still arguing about whether ChatGPT is good or bad.

Wake Up

The question is not whether machines can think. It's whether we're ready for what happens when they do.

If you think the next 3–4 years will be about better autocomplete and cleaner images, you are not paying attention. The world is about to be hit by systems that act, persist, and optimize in ways most people cannot imagine yet.

The question was never "Is AI good or bad?" The question is: who controls the actors, and what are they optimizing for?

Most people don't know what's coming. You should.

From Victim to Investigator: How One Scam Loss Turned Into a Full Blockchain Forensics Case

Umer Aziz — Mon, 22 Jun 2026 20:39:09 +0000

I Lost $20 to a Solana Scam. Then I Traced the $100K/Day Ring Behind It

From Victim to Investigator: How One Scam Loss Turned Into a Full Blockchain Forensics Case

TL;DR: I got scammed for $20 on a Solana memecoin. Instead of walking away, I traced the operation, identified a single operator running a $100K/day rug pull ring, and built a forensic tool to monitor their activity. Here's exactly how I did it.

The Scam

It started like every other memecoin play. I saw a token called ANTI-GRAVITY (AGRACING) pumping on DexScreener. The chart looked good. The community seemed active. I threw in $20 worth of SOL.

Two hours later, the liquidity vanished. The token was worthless. Classic rug pull.

Most people stop there. I didn't.

Step 1: Identifying the Scammer's Wallet

Every transaction on Solana is public. I pulled up the token's page on Solscan and found the creator wallet:

Scammer Wallet: 6GuAKzmZeiF9JckodyDCPXLUWPfFB9ehy35unCh7Swh4

I traced every transaction this wallet made. Within minutes, I noticed something: all the extracted SOL went to one central wallet.

Step 2: Finding the Hub

The scammer wasn't keeping the money in the creator wallet. They were consolidating it. Following the money trail, I found the master hub wallet:

Hub Wallet: DyaESzDfBLtbvKz7iM5Th6nsbsGSpjt5NLXuieigRcZX

This wallet was receiving massive amounts of SOL — 100 to 165 SOL per transaction — from multiple creator wallets. But it wasn't just receiving. It was also distributing funds to new wallets.

Step 3: The Bot Signature

Digging deeper into the hub's transactions, I found something bizarre. Every few transactions, the hub sent exactly 20.996123 SOL to another wallet:

Bot Wallet: 8cxba3FWd27P1fezJAujggnUd9rGc8hFXSR9EJK2WfeA

Not 20.99. Not 21.00. 20.996123 — to the 6th decimal. Every single time.

This isn't human behavior. This is a hardcoded value in a script. The bot was automatically executing liquidity removal or token sales at a fixed amount.

Step 4: Mapping the Network

I spent the next few hours tracing every wallet connected to the hub. The pattern became clear:

Hub wallet funds a new creator wallet (113-142 SOL)
Creator wallet launches a token on Pump.fun
Victims buy in
Creator extracts liquidity and sends it back to hub
Hub sends 20.996123 SOL to bot (automated cut)
Hub funds next creator wallet
Repeat

I identified 9 creator wallets in the network, each launching 1-2 tokens before being abandoned:

Creator Wallet	Known Token	Funding Received
`3n7XYTpdCu8KUbN574VRrAeteS7DG5zrBFpPtMycFhLK`	CatchCat #1	114.33 SOL
`4xTBVCzBapp83aRuBszkc42PVmV672zM4nNCETztvLkK`	CatchCat #2	121.73 SOL
`CKCwDNmbktewUwS1XTszQiFWBJNZD2og5xwyb8zQbkt`	Unknown	142.85 SOL
`7qqtdEf5VSvNWSYyyXFiRiLMoZWQN1XqcYGLxuDuT1hU`	Unknown	116.16 SOL
`cfM7zFfCWADmrwDvRtziFNnswk8g2f7TVTB2U93Lz3f`	Unknown	119.65 SOL
`CnHnEPpY4nQ6mzkDT1if36UhhBD71aU3G63845UhSDU5`	FIDGET	101.00 SOL
`8pQWvpxZkdJa5sjcSMK6Q5pGaApQ5Zb3eGU6BfSZiVmv`	Unknown	113.34 SOL
`A2w34GbXFSTMKTTT9mV8gibXiPhZQYKAWrhTymFZTGUv`	Unknown	116.07 SOL
`3gHDCqCnEQ5CiXNTJDnsHYr2NwAm926ZNvX4WAJvuqUC`	Unknown	114.46 SOL

Step 5: The Name-Squatting Trap

Here's where it gets clever. The operator launched multiple tokens with the same name but different mint addresses.

Example: "CatchCat"

Mint #1: 6jiiHkfD3zAd9XWHz1UDFfxtZu76cER6xYjifznyY8HP (RUGGED)
Mint #2: 9N4GQAukGxAsMZtaE7scE5McjUHYN7wsiL6aj6FCA5rN (RUGGED)

Why? Victim confusion. You search DexScreener for "CatchCat." You see one pumping. You buy the wrong mint. By the time you realize, the liquidity is gone.

This is deliberate, calculated deception.

Step 6: Financial Analysis

I ran the numbers:

Per token extraction: 100-142 SOL
Tokens per day: 10-20 (based on transaction frequency)
Daily revenue: 1,000-2,800 SOL
At $70/SOL: $70,000 - $200,000 per day
Operational cost: Negligible (fees are ~0.0001 SOL per transaction)

This is a $100K/day solo operation run by one person with a script.

Step 7: Why It's One Person (Not a Team)

The evidence points to a single operator:

Evidence	Interpretation
All activity in 2-hour window (13:00-15:00 local)	Single timezone, single operator
Exact same hardcoded amount everywhere	One script, one author
All creator wallets have ~10 transactions	Identical script template
No 24/7 activity	No shift rotation
Hub → Creator flow (not bidirectional)	Centralized control
Same naming patterns (CatchCat, FIDGET)	One person's taste
Bulk creation at same slot (20+ tokens)	One machine running a loop

A team would have variation. Different amounts. Different timing. Different styles. This is one person, one script, one machine.

Step 8: Building the Monitoring Tool

After mapping the network, I built a tool to monitor the hub in real-time. It's called hub_watcher.py — a Python-based blockchain forensics tool that:

Monitors the hub wallet for suspicious transactions
Detects hardcoded bot payments (the 20.996123 SOL signature)
Alerts on creator funding and hub-to-creator flows
NEW: Checks creator wallets for new token mints via Pump.fun and Token program analysis
Exports all alerts to JSON for reporting

GitHub Repo: github.com/Umer-Aziz/solana-scam-tracker

How the Token Mint Detection Works

When the hub funds a new creator wallet, the tool automatically:

Polls the creator's transaction history
Scans for interactions with the Pump.fun program (pAMMBay...)
Checks for Token program initializeMint instructions
Extracts the new token mint address from transaction metadata
Alerts with the mint address for further investigation

This links the entire chain: Hub → Creator → Token Mint — all automated.

Key Wallets (For Transparency)

Role	Address
Master Hub	`DyaESzDfBLtbvKz7iM5Th6nsbsGSpjt5NLXuieigRcZX`
Sell Bot	`8cxba3FWd27P1fezJAujggnUd9rGc8hFXSR9EJK2WfeA`
Original Scammer	`6GuAKzmZeiF9JckodyDCPXLUWPfFB9ehy35unCh7Swh4`
AGRACING Token	`DpNr3hZuoCjFubaPePLzKtwoJBR4ZNw6mUkdvPHErd9Y`

All data is from public blockchain records. No private information was accessed.

What I Learned

Blockchain forensics is pattern recognition. Hardcoded amounts, predictable timing, and reused wallets are operational security failures.
On-chain data is permanent. Every transaction, every wallet, every token is recorded forever. Scammers can't hide.
Solo operators are vulnerable. One person with one script leaves one fingerprint. Teams have variation.
The $20 was tuition. I paid $20 to learn skills that are worth thousands in the cybersecurity industry.

The Tool

If you want to monitor this network yourself or adapt the tool for other investigations:

git clone https://github.com/Umer-Aziz/solana-scam-tracker.git
cd solana-scam-tracker
pip install -r requirements.txt
python hub_watcher.py --hub DyaESzDfBLtbvKz7iM5Th6nsbsGSpjt5NLXuieigRcZX

The tool is open-source, MIT licensed, and actively maintained.

What's Next

I'm currently working on:

Discord/Telegram webhook alerts for real-time notifications
DexScreener API integration for liquidity monitoring
Machine learning anomaly detection for identifying new scam patterns
Multi-hub clustering to track multiple operations simultaneously

About Me

I'm Umer Aziz — MSc Cybersecurity, BSc Software Engineering. I build security tools, investigate blockchain fraud, and research AI security. Currently open to opportunities in cybersecurity.

LinkedIn: linkedin.com/in/umer-aziz-b13b841b2
GitHub: github.com/Umer-Aziz
Repo: github.com/Umer-Aziz/solana-scam-tracker

If you found this useful, share it. If you know someone who got scammed, show them how to trace it. Knowledge is the only weapon against these operations.