What Makes an AI System an Agent, and What Does Not
Most things sold as AI agents are prompt chains with job titles. Here is the practical test that tells them apart, and why the difference changes cost, audit, and accountability.
What I write about
Crypto investigations, AI and the law. Notes from cases, tools I am building, and things I am still working out.
The Expertise Paradox: AI Is Eating the Junior-to-Senior Career Ladder
The real AI threat is not mass job loss alone. It is the collapse of the training ground. Automate the easy work and you may erase how hard expertise gets built.
Case Solved: A Wallet With Almost Nothing Lost Six Figures in USDT, in One Transaction
A case study on a Tron USDT drain: an empty-looking wallet signs an approval, real funds arrive later, and one transaction empties it. What the trail can prove, and where it stops.
First Build the Machine. Then Add the AI Engine.
Most companies add AI before they have a workflow worth amplifying. AI is not the system. It is the amplifier. Structured processes create scalable AI. Broken ones create faster chaos.
If You Still Babysit Every Output, You Do Not Have an Agent
Most launches labeled AI agent are chatbots in costume. A real agent is architecture: goals, tools, permissions, memory, and feedback loops. Stop chasing the label. Start mapping the workflow instead.
AI Should Enhance the System, Not Subsidize Unclear Thinking
A powerful model cannot fix unclear thinking. It can make it faster and more confident. The advantage is directing the workflow: the job, context, constraints, checks, and where it can fail.
Guardian Agents: Autonomy Without Oversight Is Just Faster Risk
Only a fraction of firms have shipped AI agents, yet most plan to. Hallucination rates and breach data say the same thing: autonomy without a second loop is not scale. It is exposure.
India's Crypto Tax Visibility Gap: ₹28 Lakh Crore In, ₹511 Crore Out
Parliamentary TDS and income-tax figures show what India's crypto tax net can see onshore. This piece separates activity metrics from taxable records, maps the offshore gap range, and flags what CARF will not fix.
A Better Model Will Not Fix Unclear Thinking
A stronger model makes unclear thinking faster, smoother, and far more confident. It does not make it correct. What decides the output is still framing the job, checking the answer, and knowing when you are wrong.
When a Crypto Theft Is Bigger Than One Transaction
A crypto theft can involve many wallets set up before the main move. What on-chain data can prove, where the public record stops at an exchange, and why overstating the trail hurts the case.