I'm starting an experiment:
Can I build a real online income using X?
I'm going to document everything โ experiments, failures, revenue and lessons.
Follow along. No fake screenshots. No โget rich quickโ bullshit.
Something interesting is happening with AI agents in India. ๐ฎ๐ณ
It's not just startups experimenting anymore.
Companies like Swiggy, Tata Steel, HDFC Bank and NPCI are putting agentic AI into real workflows.
The interesting shift:
We're moving from
โLet's add AI to our app.โ
to
โCan an AI agent actually operate parts of our business?โ
That's a much bigger change.
What industry do you think gets disrupted first?
OpenAI introduced Dots last week.
They're designed to be always-on agents that can take responsibility for ongoing work instead of waiting for you to send a prompt.
That's a pretty big shift:
Chatbot โ answers questions.
Agent โ completes tasks.
Always-on agent โ keeps working toward a goal.
Would you actually trust an AI to work while you're offline?
You can now build Amazon Bedrock Managed Agents powered by OpenAI models.
The interesting part isn't just โOpenAI models on AWS.โ
These agents can run inside AWS with:
AWS identities + permissions + governance
So now imagine an AI agent that can:
โ read your S3 data
โ query your databases
โ interact with AWS services
โ follow IAM permissions
That's where AI + AWS + DevOps gets really interesting.
Would you give an AI agent access to your AWS account? ๐
Coding agents are getting scary good.
But here's the part nobody talks about enough:
The bottleneck is slowly moving from writing code โ reviewing code.
If an agent can generate thousands of lines overnight...
Who is reviewing all of it?
MCP might end up being more important than people realize.
Instead of building a custom integration for every AI tool:
Agent โ MCP โ Tool
One common way for agents to interact with the systems they need.
The interesting part is what happens when every SaaS app becomes "agent accessible."
Google Antigravity is starting to look tempting.
I haven't used it yet, but Gemini 4 Argon alone might be worth trying especially when Claude and ChatGPT keep hitting their limits.
One AWS networking fact that's easy to miss:
Security Groups are stateful.
If you allow an inbound connection, the response traffic is automatically allowed.
You don't need to create a separate outbound rule for the response.
NACLs work differently.
This comes up a lot in AWS interviews.
Here's an easy AWS cost mistake:
Stopping an EC2 instance doesn't mean your AWS bill becomes $0.
The compute charge stops, but attached resources like EBS volumes can continue to incur charges.
Stopped โ deleted.
Always know what you're actually paying for.
โ๏ธ AWS INTERVIEW CHALLENGE #02
GP2 vs GP3 โ what would you choose?
You have an EC2 workload that needs:
8,000 IOPS
300 MB/s throughput
Cost optimization
Which EBS volume would you pick?
A. GP2
B. GP3
C. Either
Think before you answer ๐
.
.
Answer: B โ GP3
The main reason?
With GP2, performance is tied to volume size.
With GP3, you can configure IOPS and throughput independently of storage size.
So you don't have to increase storage just to get more performance.
๐ฌ Would you choose GP2 or GP3 for a new production workload?
Drop your answer below ๐
Follow for more AWS interview questions.
#AWS #AWSInterview #EBS #EC2 #CloudComputing #DevOps
Your EC2 instance can have a public IP and still be unreachable from the internet.
Why?
A public IP alone isn't enough.
Your subnet needs a route to an Internet Gateway, and your security rules/NACLs must also allow the traffic.
Most people know that an S3 bucket can be versioned.
But did you know?
Deleting an object from a versioned S3 bucket doesn't necessarily delete it.
S3 can add a delete marker instead.
The previous version can still exist.
Small detail. Big difference when you're troubleshooting accidental deletions.
Learning & building around AI, AWS, automation and DevOps.
If you're building cool stuff, learning in public, or somewhere in between
let's connect. ๐ค
Let's learn, build, and grow together. ๐